Research Methods for Criminology and Criminal Justice [Fourth ed.] 1284113019, 9781284113013

Citation preview

Research Methods for Criminology and Criminal Justice FOURTH EDITION M.L. Dantzker PhD University of Texas, Pan American Ronald D. Hunter PhD Georgia Gwinnett College Susan T. Quinn PhD Georgia Gwinnett College

JONES & BARTLETT LEARNING

World Headquarters Jones & Bartlett Learning 5 Wall Street Burlington, MA 01803 978-443-5000 [email protected] www.jblearning.com Jones & Bartlett Learning books and products are available through most bookstores and online booksellers. To contact Jones & Bartlett Learning directly, call 800-832-0034, fax 978-443-8000, or visit our website, www.jblearning.com.

Substantial discounts on bulk quantities of Jones & Bartlett Learning publications are available to corporations, professional associations, and other qualified organizations. For details and specific discount information, contact the special sales department at Jones & Bartlett Learning via the above contact information or send an email to [email protected].

Copyright © 2018 by Jones & Bartlett Learning, LLC, an Ascend Learning Company

All rights reserved. No part of the material protected by this copyright may be reproduced or utilized in any form, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without written permission from the copyright owner. The content, statements, views, and opinions herein are the sole expression of the respective authors and not that of Jones & Bartlett Learning, LLC. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not constitute or imply its endorsement or recommendation by Jones & Bartlett Learning, LLC, and such reference shall not be used for advertising or product endorsement purposes. All trademarks displayed are the trademarks of the parties noted herein. Research Methods for Criminology and Criminal Justice, Fourth Edition is an independent publication and has not been authorized, sponsored, or otherwise approved by the owners of the trademarks or service marks referenced in this product. There may be images in this book that feature models; these models do not necessarily endorse, represent, or participate in the activities represented

in the images. Any screenshots in this product are for educational and instructive purposes only. Any individuals and scenarios featured in the case studies throughout this product may be real or fictitious, but are used for instructional purposes only. Production Credits VP, Executive Publisher: David D. Cella Executive Editor: Matt Kane Associate Acquisitions Editor: Marisa A. Hines Director of Vendor Management: Amy Rose Director of Marketing: Andrea DeFronzo Marketing Manager: Lindsay White Manufacturing and Inventory Control Supervisor: Amy Bacus Composition and Project Management: S4Carlisle Publishing Services Cover Design: Kristin E. Parker Rights & Media Specialist: Robert Boder Media Development Editor: Shannon Sheehan Cover Image: hex-code data © Maksim Kabakou/Shutterstock. Man in handcuffs © nito/Shutterstock Printing and Binding: Edwards Brothers Malloy Cover Printing: Edwards Brothers Malloy

Library of Congress Cataloging-in-Publication Data Names: Dantzker, Mark L., 1958-, author. | Hunter, Ronald D., author. | Quinn, Susan T., author. Title: Research methods for criminology and criminal justice/M.L. Dantzker, PhD, University of Texas, Pan American, Ronald D. Hunter, PhD, Professor and Program Coordinator of Criminal Justice/Criminology, School of Liberal Arts, Georgia Gwinnett College, Lawrenceville, GA, Susan T. Quinn, PhD, Assistant Professor of Criminal Justice and Criminology, Georgia Gwinnett College, Lawrenceville, GA. Description: Fourth Edition. | Burlington, Massachusetts: Jones & Bartlett Learning, 2018. | Revised edition of Research methods for criminology and criminal justice, c2012. | Includes bibliographical references and index. Identifiers: LCCN 2016035879 | ISBN 9781284113013 (pbk.) Subjects: LCSH: Criminology—Research— Methodology. | Criminal justice, Administration of— Research—Methodology. Classification: LCC HV6024.5.D36 2016 | DDC 364.072/1—dc23 LC record available at https://lccn.loc.gov/2016035879 6048

Printed in the United States of America 20 19 18 17 16 10 9 8 7 6 5 4 3 2 1

BRIEF CONTENTS Section I Functions CHAPTER 1 Research: What, Why, and How CHAPTER 2 Research and Ethics CHAPTER 3 The Beginning Basics CHAPTER 4 The Vocabulary of Research Section II Procedures CHAPTER 5 Sampling CHAPTER 6 Introduction to Research Design CHAPTER 7 Research Designs CHAPTER 8 Qualitative and Quantitative Research Designs

CHAPTER 9 Questionnaire Development CHAPTER 10 Data Collection Section III Final Steps CHAPTER 11 Data Preparation and Analysis CHAPTER 12 Inferential Statistics CHAPTER 13 Writing the Research CHAPTER 14 Summing Up

CONTENTS Preface Acknowledgments Section I Functions CHAPTER 1 Research: What, Why, and How What You Should Know! The Nature of Scientific Inquiry Social Science Research and the Real World Science Versus Casual Inquiry The Scientific Method The Relationship Between Theory and Research Theory The Purpose of Research What Is Research? Types of Research Descriptive Research Explanatory Research

Predictive Research Intervening Research Why Research Is Necessary Curiosity Social Problems Theory Testing Factors That Influence Research Decisions How Research Is Done Identifying the Problem Research Design Data Collection Data Analysis Reporting Summary

CHAPTER 2 Research and Ethics What You Should Know! Ethics The Researcher’s Role Belmont Report Ethical Considerations Ethical Ramifications Harm to Others Privacy Concerns

Voluntary Participation Deception The Professionalism of Research Ethical Research Criteria Reasons for Confidentiality and Privacy Summary

CHAPTER 3 The Beginning Basics What You Should Know! Getting Started Picking a Topic The Purpose of the Research Exploring Describing Explaining Become Familiar with the Library Critiquing the Literature Understanding Writing Styles Knowing What to Look For The Research Question Hypotheses Summary

CHAPTER 4 The Vocabulary of Research What You Should Know! The Language of Research Theory Conceptualization Operationalization Variables Hypotheses Assumptions Other Necessary Terms Qualitative Versus Quantitative Research Qualitative Research Defined Merits and Limitations of Qualitative Research Quantitative Research The Research Process What You Have Not Done Before Summary

Section II Procedures CHAPTER 5 Sampling

What You Should Know! Sampling Probability Theory Probability Sampling Simple Random Samples Stratified Random Samples Systematic Samples Cluster Samples Nonprobability Sampling Purposive Samples Quota Samples Snowball Samples Convenience Sample Sample Size Confidence Levels Sampling Formulas A Commonly Used Sampling Formula A Sampling Size Selection Chart Summary

CHAPTER 6 Introduction to Research Design What You Should Know! Empirical Observation

Causality Experimental Research Designs Quasi-Experimental Research Design Quantitative Levels of Measurement Summary

CHAPTER 7 Research Designs What You Should Know! Research Designs Descriptive Research Design Historical Research Design Cross-Sectional Research Design Longitudinal (or Time Series) Research Design Case Study Research Design Determining Correlations and Causations Evaluation Research Summary

CHAPTER 8 Qualitative and Quantitative Research Designs What You Should Know! Qualitative Research Design Field Interviews Structured Interviews Semi-Structured Interviews Unstructured Interviews Focus Groups Field Observation Ethnographic Research Sociometry Quantitative Research Design Survey Research Quantitative Field Observation Summary

CHAPTER 9 Questionnaire Development What You Should Know! Surveys and Questionnaires Rules for Questionnaire Construction Rule One: Start With a List of All the Items One Is

Interested in Knowing About the Group, Concept, or Phenomenon Rule Two: Be Prepared to Establish Validity and Reliability Rule Three: Word the Questionnaire Appropriately for the Target Audience Rule Four: Clearly Identify Who Should Answer the Questions Rule Five: Avoid Asking Questions That Are Biased, Leading, or Double-Barreled in Nature Rule Six: Before Constructing a Questionnaire, Decide Whether to Use Open- or Closed-Ended Questions or a Combination of Both Rule Seven: Keep in Mind That Respondents May Not Have All the General Information Needed to

Complete the Questionnaire Rule Eight: Whenever Possible, Pretest the Questionnaire Before It Is Officially Used Rule Nine: Set Up Questions So That the Responses Are Easily Recognizable Whether the Questionnaire Is SelfAdministered or an Completed in an Interview Rule Ten: Organize the Questionnaire to Keep the Respondents’ Interest, Encouraging Them to Complete the Entire Questionnaire Scales Scaling Procedures Arbitrary Scales Attitudinal Scales Summary

CHAPTER 10 Data Collection What You Should Know!

Survey Research Mail Distribution of Surveys Surveys and the Internet Interviews Face-to-Face Interviews Structured, Semi-structured, and Unstructured Interviews Telephone Interviews Field Observation Secondary Data Sources of Secondary Data Content Analysis Summary

Section III Final Steps CHAPTER 11 Data Preparation and Analysis What You Should Know! Data Preparation Data Coding Data Entry Data Cleaning Missing Data Recoding Data

Data Analysis Statistical Analysis Descriptive Statistics Frequency Distributions Displaying Frequencies Measures of Central Tendency Measures of Variability Summary

CHAPTER 12 Inferential Statistics What You Should Know! Statistical Analysis Overview of Inferential Statistics Measures of Association Statistical Significance Comparative Statistics Crime Rates Crime-Specific Rates Percentage Change Trend Analyses Inferential Statistics Bivariate Analysis Contingency Tables (or Cross-Tabulations)

Bivariate Regression Multivariate Analysis Student t Test Correlation Analysis of Variance Multiple Regression Other Multivariate Techniques Summary

CHAPTER 13 Writing up the Research What You Should Know! The Research Paper The Title Page Abstract The Introduction Methodology Results Conclusions References or Bibliography Tables and Figures Appendices Summary

CHAPTER 14 Summing Up What You Should Know!

Research Ethics Ethical Concerns Getting Started Picking a Topic Reviewing the Literature The Research Question Doing Criminological Research Steps in the Research Process The Language of Research Theory Conceptualization Operationalization Variables Hypotheses Sampling Validity Reliability Data Sampling Probability Theory Probability Sampling Nonprobability Sampling Sample Size Confidence Levels

Introduction to Research Design Causality Experimental and QuasiExperimental Research Designs Research Design Descriptive Research Design Historical Research Design Cross-Sectional Research Design Longitudinal (or Time Series) Research Design Case Study Research Design Correlational and CausalComparative Research Qualitative Research Field Interviewing Focus Groups Field Observation Ethnographic Study Quantitative Research Survey Research Questionnaire Construction

Scales Data Collection Surveys Interviews Field Observation Secondary Data Content Analysis Data Preparation Data Analysis Statistical Analysis Frequency Distributions Other Ways to Describe the Data Inferential Statistics Measures of Association Statistical Significance Bivariate Analysis Multivariate Analysis Writing the Research Summary

References Appendix Institutional Review Board Application

What Is IRB Approval? Who Should Complete the IRB Application Form? How Long Does an IRB Review Take? General Overview of Information in IRB Applications Read This If You Are Using a Published Instrument Read This If You Are Creating Your Own Instrument or Modifying an Existing Instrument Forms and Letters

Glossary Author Index Subject Index

PREFACE The purpose of this text is to assist criminal justice and criminology students in developing an understanding (and hopefully, an appreciation) of the basic principles of social science research. We do not seek to turn you into a research scientist in one short course, but we do hope you will garner a better understanding of the research you read in your studies. Furthermore, this text provides a solid foundation on which to build, should you be interested in doing social science research, whether criminological or criminal justice oriented, in the future. This text will enable you to grasp the importance of scientific research, to read and comprehend the research methodologies employed by researchers, and to develop the basic tools to conduct your own social science research. Whether research is done by a college student completing a project for his or her degree (or just trying to understand an assigned reading) or by a professor meeting requirements or expectations associated with his or her position, it should be enjoyable and not a chore. The first step is to learn the basics for conducting research. A number of

existing textbooks can assist in this task, but many make learning about research—let alone conducting it—appear daunting. This text has made every attempt to ease the task of learning how to conduct research and perhaps even to put the prospect of conducting research in a favorable light. To accomplish these aims, this textbook is divided into three sections: Functions, Procedures, and Final Steps. Each chapter begins with a brief summary of what should be learned from its content. Within the text, realistic examples taken from currently published research are provided to enhance understanding of how specific aspects of research are applicable to criminal justice and criminology. Finally, each chapter ends with questions and/or exercises requiring students to apply what has been learned from the chapter. The text begins by discussing what research is and why and how it is conducted. Chapter 1, Research: What, Why, and How, addresses basic questions: What is criminal justice and criminological research? Why conduct this research? How can this research be completed? In general, this chapter lays the foundation for conducting research. Because criminal justice research often deals with human behavior, the ethics associated with such

research are important. Chapter 2, Research and Ethics, discusses the ethics relevant to conducting social science research. Deciding on what topic to conduct research can often be frustrating. However, there are numerous sources available to assist in making a decision on what to research. Chapter 3, The Beginning Basics, explores what sources to use and the issue of developing the research question, which is often the driving force behind social science research. In Chapter 4, The Vocabulary of Research, students are introduced to the terminology associated with conducting research, such as theory, hypothesis, population, sample, and variables. Furthermore, it briefly explores the processes required for conducting research. It would be great if information could be gathered from a complete population, but this is almost impossible in criminal justice and criminological research. Therefore, sampling is an important aspect of research. Chapter 5, Sampling, examines this concept and related issues. To successfully complete any type of research, it is important to establish a feasible plan or blueprint,

known as the research design. Chapter 6, Introduction to Research Design, discusses the various experimental research designs available for criminal justice and criminological research: experimental research, including classical, pretestposttest, posttest-only, and factorial experiment research designs. A brief mention is made of quasiexperimental research design. The concept of research design is further explored in Chapter 7, Research Designs. This chapter discusses additional research designs available for criminal justice and criminological research, including descriptive, historical, cross-sectional, longitudinal, correlational, and causal–comparative research designs. There is a long-standing debate as to what is more “academic,” qualitative or quantitative research. Chapter 8, Qualitative and Quantitative Research Designs, does not enter the debate but simply explains how both qualitative and quantitative research fit into criminological and criminal justice research. One of the most popular methods of collecting data is the questionnaire. Although a general rule is to use an established questionnaire, many individuals

choose to design their own. Chapter 9, Questionnaire Development discusses the intricacies of designing a questionnaire, including issues of measurement, reliability, and validity. In establishing the research design, a key component is how the data is to be collected. The four primary means for collecting data—survey, interview, observation, and secondary data methods —are identified and explored in Chapter 10, Data Collection. Once the data are collected, the question is what to do with this information. There are a number of statistical techniques from which to choose. This is not a statistics book. However, to assist students in better understanding the role of statistics in the research process, we offer Chapter 11, Data Preparation and Analysis, and Chapter 12, Inferential Statistics. Now that the data are collected and analyzed, the final step of the research process is to write up the findings. For many, this is a daunting task. To help ease the fear and frustration, Chapter 13, Writing up the Research, takes students through a step-bystep introduction to writing the research.

Finally, to briefly assemble all the information offered throughout this text into a handy reference guide, we offer Chapter 14, Summing Up, an extensive, yet simple review of all the main concepts. A final note concerns what this text is not. Research Methods for Criminology and Criminal Justice, Fourth Edition, is not a statistics book. However, it could be used in conjunction with a criminal justice statistics text. The fact is that separate books are often required to provide students the fullest extent of the knowledge required to conduct research and to analyze the data. This text allows students to learn how to conduct the research, leaving the statistics for another course and text. We hope you will find this text as useful as it is intended to be. If nothing else, we hope it will help you feel more comfortable about reading or conducting criminal justice and criminological research. Research Methods for Criminology and Criminal Justice is accompanied by PowerPoint chapter guides, lecture outlines, and a test bank for qualified adopters.

ACKNOWLEDGMENTS As with each book, first edition, or revisions, there are several people who deserve our recognition and gratitude. We greatly appreciate Marisa Hines, our Acquisitions Editor, for her support and confidence. We are very grateful to Amy Rose, our Production Editor, for her excellent work. We also thank the other editorial staff for their efforts. We wish to extend our thanks to the following reviewers: Kenneth Clontz, PhD School of Law Enforcement and Justice Administration Western Illinois University Macomb, Illinois Aric W. Dutelle University of Wisconsin-Oshkosh Oshkosh, Wisconsin Michael Montgomery, PhD Tennessee State University Nashville, Tennessee

We would also like to offer our appreciation to the instructors who have used our book and made it possible to do a fourth edition. Finally, we’d like to thank our families for their love and support throughout this process. Thank you one and all. Mark L. Dantzker Ronald D. Hunter Susan T. Quinn

SECTION I: Functions

CHAPTER 1: Research: What, Why, and How What You Should Know! Research methods for many graduate and undergraduate students can be misleading, confusing, and frustrating. The subject is often taught or presented in a manner where students may think that they are going to be social science researchers, actually conducting research. In reality, however, most students will be consumers of research and will conduct little to no actual research themselves. Students will read studies conducted by others, perhaps hoping to apply or better understand an area, such as policing, corrections, or courts, with which they may be working. As consumers of research, students must be able to evaluate the quality of the research they are reading. However, before being able to conduct or even evaluate research, one should be able to understand the basics of research. After completing this chapter, the reader should be able to:

1. Discuss tradition and authority as sources of human learning and be able to compare and contrast their strengths and weaknesses. 2. Identify and discuss the errors that plague casual observation. 3. Define what is meant by the scientific method. Explain how it seeks to remedy the errors of casual observation. 4. Compare and contrast the inductive and deductive logic processes. 5. Define research and explain the purposes of research. 6. Describe the primary steps in conducting research. 7. Compare and contrast basic, applied, and multipurpose research. 8. Identify and discuss the various types of research. 9. Discuss the reasons for criminal justice and criminological research. 10. Discuss the various factors that influence research decisions.

The Nature of Scientific Inquiry It seems not that long ago that the authors were criminal justice students taking a first course in research methods. Our thoughts were, if we want to be police officers, why do we have to take this

course? This is even worse than criminal theory, another useless course. What does it have to do with the real world in which we want to work? Later police experience in that real world taught us the value of both theory and research in the field of criminal justice. When we subsequently returned to school for graduate studies, the importance of theory and research was more readily apparent. During our careers, we had learned that scientific investigation is very similar to criminal investigation: the use of a logical order and established procedures to solve real-world problems.

Social Science Research and the Real World As police officers, the authors sought to determine whether a crime had been committed (what occurred and when it occurred); who had done it; how they had done it; and why they had done it. We then sought to use that investigatory knowledge to develop a successful prosecution of the offender. Our endeavors in the field taught us that the theory course that we had grudgingly endured had provided the rationale for human behavior on which the strategies of policing, courts, and corrections were based. We also discovered that those theories were not developed in some esoteric vacuum. They were the products of trial-and-error experiments

conducted in policing, the courts, and corrections that had been refined and reapplied to their appropriate subject area. Today’s policedeployment strategies, legal processes, and correctional techniques are all solidly based on prior theory and research. There is an increased focus on evidence-based practices and procedures, which refers to the evaluation of effectiveness of these practices and procedures using sound research practices. For example, a local police department may review studies of police deployment strategies to determine the most effective, as well as cost-effective, strategies. Due to budget constraints, criminal justice agencies are increasingly relying on strategies or elements of strategies that have been previously been proven effective (which may be preferable to throwing darts in the dark and hoping to hit something). The information from this course can help students be a contributing part of this process by developing the skills necessary to evaluate the quality of research studies. These statements related to theory and research can also be applied to social science research in general. Typical real-world conclusions are often flawed because of a number of issues that cause

one’s observations and reasoning to be inaccurate. The scientific method seeks to provide a means of investigation to correct (or at least limit) the inaccuracies of ordinary human inquiry (Adler & Clark, 2007; Bachman & Schutt, 2008; Kline, 2009). How one interprets one’s own observations and what one learns from others is based on tradition and authority. Tradition is the cultural teaching, including customs and beliefs, about the real world. It is based on the experiences of members of society who passed their knowledge on to others. “Poisonous snakes are dangerous. Beware of them!” You do not have to be bitten by a rattlesnake to appreciate its hazard. You have been taught by other members of your culture to respect the threat of poisonous snakes. This example is one of learning accurate information from tradition. Unfortunately, knowledge based on tradition is often erroneous. For example: “Women are not suited to be police officers. They are too weak and too emotional.” A multitude of highly competent and professional police officers have proved this sexist stereotype to be a fallacy. It is also important to note that just because a belief is held by a large number of people does not mean that the belief is accurate. The other source of secondhand knowledge is authority (Kraska & Neuman, 2008; Lavrakas,

2008; Maxfield & Babbie, 2009). Authority refers to new knowledge that is provided from the observations of others whom one respects. The older sibling or cousin who explained the facts of life to you was an authority figure. How accurate his or her explanations were, we leave to you to decide. As you got older, you learned that much of the free advice you received was worth what you paid for it and that a great deal of bought advice also had little value. The importance of knowledge gained from authority figures depends on their qualifications relative to the subject being discussed. Therefore, one goes to a physician for help with health problems and hires a plumber to fix a broken water pipe. These individuals are expected to have the expertise and training to provide information that laypersons do not have. Your criminal justice professor may be able to provide accurate information on policing, courts, or corrections but is less likely to be able to do so on the biology of tree frogs. Like tradition, the knowledge gained from dealings with authority figures can be extremely accurate or highly erroneous.

Science Versus Casual Inquiry Casual inquiry is influenced by the sources of knowledge (tradition and authority) discussed in the previous section. In addition, there are other pitfalls

that create errors in one’s observations. The information gathered during casual inquiry may be flawed because of inaccurate observation, overgeneralization, selective observation, and illogical reasoning (Kraska & Neuman, 2008; Lavrakas, 2008; Maxfield & Babbie, 2009). Inaccurate observation occurs when conclusions are made based on hasty or incomplete observations. As an example, first impressions are made with limited information and are rarely completely accurate. Let us say that you have started a new job and on the first day, one of your new co-workers comes in late looking disheveled. During the course of the day, you notice that the coworker is constantly on the phone. Based on this information, you may decide that the co-worker is lazy and does not care about his job. However, what you did not know is that this co-worker was dealing with a family member who had a serious illness and had been admitted to the hospital. Overgeneralization occurs when conclusions are made about individuals or groups based on knowledge of similar individuals or groups. “All lawyers are liars!” is an example. Despite the preponderance of lawyer jokes and any bad experiences one may have had with an attorney,

one cannot accurately make that conclusion about all attorneys. There are simply too many attorneys (men and women of honesty and integrity and those of questionable ethics) to make such a conclusion without an individual knowledge of the person. Selective observation is when one sees only those things that one wants to see. Individuals may select observations that support what they already believe while ignoring observations that contradict their beliefs. Racial and ethnic stereotyping is an example of negatively biased selective observation. The attitude that “all whites are racists who seek to oppress” may cause the observer to see what he or she believes in the behaviors of all European Americans with whom the observer encounters. Selective observation may also be positively biased. “My darling, wonderful child has never done anything like that.” Such selective observation can lead to major disappointment, such as when “He’s a wonderful man who caters to my every whim” becomes “He’s a selfish jerk who doesn’t ever consider my feelings.” Finally, illogical reasoning happens when one decides that despite past observations, the future will be different. The individual who plays the lottery week after week believing that eventually he has to

win exemplifies illogical reasoning. Depending on the lottery, the chance of winning may be one is several hundred million no matter how often one plays. Another example would be when someone continues to drive even after the car’s gas gauge has reached “empty.” Eventually, the car will run out of gas, and the illogical reasoning will result in a call to a roadside service (and hopefully not a large bill!). Science seeks to reduce the possibility of the previously mentioned errors occurring by imposing order and rigor on observations. The scientific method provides the means of doing so with the goal of increasing the accuracy of the observations.

The Scientific Method The scientific method seeks to prevent errors of casual inquiry by using procedures that specify objectivity, logic, theoretical understanding, and knowledge of prior research in the development and use of a precise measurement instrument designed to record observations accurately (Bryman, 2008; Creswell, 2008; Gavin, 2008; McBurney & White, 2007). The result is a systematic search for the most accurate and complete description or explanation of the events or behaviors that are being studied. Just as a criminal investigation is a search for “the facts” and a criminal trial is a search

for “the truth,” the scientific method is a search for knowledge. The criminal justice and criminological researcher seeks to use the principles of empiricism, skepticism, relativism, objectivity, ethical neutrality, parsimony, accuracy, and precision to assess a particular theoretical explanation. Of the previously mentioned principles, empiricism is defined as seeking answers to questions through direct observation, such as using your sense of sight to observe the impact a bullet fired at target has on the target. Skepticism is the search for disconfirming evidence and the process of continuing to question the conclusions and the evidence that are found. It is particularly important in the research process because without skepticism there would not be the motivation for additional or new research. Objectivity mandates that conclusions are based on careful observation that sees the world as it really is, free from personal feelings or prejudices. Criminal justice and criminological researchers often acknowledge that total objectivity is unattainable, but every reasonable effort is made to overcome any subjective interests that might influence research outcomes. One way to address objectivity is through “intersubjectivity,” which could be accomplished through having more than one person observe a phenomenon and

compare their findings. Ethical neutrality builds on objectivity by stressing that the researcher’s beliefs or preferences are not allowed to influence the research process or its outcomes. A researcher should not modify the analyses to reach a preferred outcome, such as finding a relationship between two variables. Parsimony is the attempt to reduce the sum of possible explanations for an event or phenomenon to the smallest possible number. Accuracy requires that observations be recorded in a correct manner exactly as they occurred. Precision refers to the exactness of the attributes of a variable. For example, saying that a man is young is less precise than saying that the man is 23 years old (Adler & Clark, 2007; Maxfield & Babbie, 2009; Vito, Kunselman, & Tewksbury, 2008).

The Relationship Between Theory and Research As was discussed in a prior section, the practice of criminal justice is based on theories about the causes of crime and how to respond to them. Criminology is an academic discipline that studies the nature of crime, its causes, its consequences, and society’s response to crime. Criminal justice as an academic discipline tends to focus more on the creation, application, and enforcement of criminal laws to maintain social order. Due to the amount of

overlap between the two disciplines, we deal with the two as one discipline within the text (as indeed many criminologists consider them to be). Regardless of the reader’s orientation, theory is integral in the development of research. Likewise, theory that has been validated by research is the basis for practice in the criminal justice system.

Theory Theory explains how something is (Bachman & Schutt, 2008; Bickman & Rog, 2009; Bryman, 2008; Kraska & Neuman, 2008; Lavrakas, 2008; Maxfield & Babbie, 2009). Personal ideologies are of no value in criminological theory unless they can be evaluated scientifically. We define theory as “an attempt to explain why a particular social activity or event occurs.” A theory is a generalized explanation about the phenomenon being studied. From theory, more precise statements (concepts) are developed. These specific measurable statements are hypotheses. It is through observation and measurement that the validity (correctness or ability to predict accurately what it seeks to examine) of a hypothesis is examined. If the hypothesis cannot be rejected, then support for the theory is shown. The method by which the hypothesis is observed and measured is known as research. The relationship

between theory and research may be either inductive or deductive in nature.

Inductive Logic In the stories by Sir Arthur Conan Doyle his detective hero, Sherlock Holmes, continuously assails Dr. Watson, a man of science, about the merits of “deductive logic.” It is through deductive logic that Holmes is said to solve his cases. In actuality, the process that Holmes describes is inductive logic. In this process, the researcher observes an event, makes empirical generalizations about the activity, and constructs a theory based on these activities. After visiting a crime scene (often of a murder), Holmes develops a theory of the crime, specifying how and why it occurred as well as who committed the crime. Only rarely does Holmes engage in the deduction of which he so highly speaks. Another example of inductive logic is Sir Isaac Newton’s alleged formulation of the theory of gravity after observing an apple fall from a tree.

Deductive Logic Deductive logic begins with a theoretical orientation. Based on this theory, the researcher then develops research hypotheses that are tested by observations. These observations lead to empirical generalizations that either support or challenge the

theory in question. Had our hero Holmes followed up his theory construction with such observation, then he would have engaged in deduction. The scientific method is based on deductive logic, which involves theory construction and testing. In criminal justice and criminological research, the distinctions between inductive and deductive logic are often obscured because the two processes are actually complementary parts of a single model. Although described in a circular model (Wallace, 1971), the elements of both inductive and deductive logic may also be viewed as part of a never-ending continuum that begins with theory, which encourages creation of hypotheses, and in turn calls for observations. The result of analyzing observations is generalizations, and the conclusions of the generalizations assist in modification of the theory.

The Purpose of Research The average college student truly believes he or she knows what it means to conduct research. Many have written a “research paper” either in high school or for a college course. Generally, these assignments consist of stringing together journal articles without a plan for collecting or analyzing the information. Realistically, few have ever had the opportunity truly to write a research paper because even fewer have ever conducted scientific research.

What Is Research? Research is the conscientious study of an issue, problem, or subject. It is a useful form of inquiry designed to assist in discovering answers. It can also lead to the creation of new questions. For example, a judge wants to know how much effect her sentencing has had on individuals convicted of drug possession, particularly as it compares to another judge’s sentencing patterns. She asks that research be conducted that focuses on the recidivism, specifically, rearrests, of these individuals. The results indicate that 30% of drug offenders sentenced in her court are rearrested, compared to only 20% from the other judge’s court. Between the two courts, the judge has discovered that her sentencing does not seem to be as effective. This strategy answered the primary question of the research, but it has also created new questions, such as why her methods are not working as well as the other judge’s. Research creates questions, but ultimately, regardless of the subject or topic under study, it is the goal of research to provide answers. One of the more common uses of the term research is a description of what a student might be asked to accomplish for a college class. Often, one hears instructors and students refer to the choosing of a

topic, using several sources, and writing a descriptive paper on the topic as research. If done thoroughly and objectively, this process may actually constitute qualitative research (discussed in detail in Chapter 4). Unfortunately, these “research papers” are too often essays based more on the individual’s ideologies rather than on scientific discovery. For the purpose of this text, the emphasis is on empirical research that yields scholarly results. There are many formal definitions for the term research. This text uses the following: Research is the scientific investigation into or of a specifically identified phenomenon and is applicable to recognizable and undiscovered phenomena. Therefore, in terms of criminal justice and criminology, related research can be viewed as the investigation into or of any phenomenon linked to any or all aspects of the criminal justice system. Using this definition, criminal justice and criminological research is not limited to any one area. Box 1-1 offers just a few of the related topics one might research. Along with the plethora of research topics, there are several methods for conducting research. These include surveys, observation, case studies, and

reviews of official records, among other methods. These methods are discussed in further detail, but first it is important to understand all the underlying characteristics of research. Criminal justice and criminological research is often divided into two forms: basic and applied.

Box 1-1 Applied Research Topics: Some Examples Policing Stress Patrol effectiveness Use of force Job satisfaction Community policing Citizen satisfaction Courts Types of sentencing Plea bargaining Race and sentencing Jury versus judge verdicts Paid versus public defender

Corrections Rehabilitation versus punishment Prisons Probation Electronic monitoring Boot camps Restorative justice Death penalty Other Criminal behavior Victims Drugs Gangs Adult or juvenile criminality

Basic Research Basic research, sometimes referred to as “pure research,” is the conducting of scientific inquiries that may not offer or provide any direct application

or relevance (Drake & Jonson-Reid, 2008; Dunn, 2009). Instead, it is concerned with the acquisition of new information for the purpose of helping develop the scholarly discipline or field of study in which the research is being conducted. This type of research is more often consistent with criminological inquiries. The findings from basic research often have little or no applicable use in the field of criminal justice. However, such research may be used as the foundation for subsequent applied research and criminal justice policy. It is also such research that leads to the development of the criminological theories that guide the actions of lawmakers, police, courts, and corrections. For example, Dantzker and McCoy (2006) explored what process was being used by the largest 17 Texas municipal police agencies for psychological preemployment screening of police officer candidates. This information was later used to help create a survey tool for a larger research study but initially was simply information gathering or pure research.

Applied Research Perhaps the most immediately useful type of research in criminal justice is applied research, which is primarily an inquiry of a scientific nature designed and conducted with practical application as its goal. It is the collection of data and its

analyses with respect to a specific issue or problem so that the applications of the results can influence change. In essence, applied research provides answers that can be used to improve, change, or help decide to eliminate the focus of study. For example, Hernandez (2009) compared several types of self-administered questionnaires among Hispanic juveniles to identify which one was the best predictor of substance abuse. Applied research can be used within criminal justice agencies to evaluate the effective of procedures and to justify funding for procedures or programs found to be effective, such as being able to show that a specific drug court program decreases recidivism.

Multipurpose Research Basic and applied research are vital in the study of crime and justice. Yet, a good portion of the research conducted by criminal justice and criminology academics tends to come under a third area of research best labeled as multipurpose research. Multipurpose research is the scientific inquiry into an issue or problem that could be both basic and applied research. This type of research generally begins as basic, but then the results are found to have an applied purpose. For example, a police chief is interested in the level of job satisfaction among sworn employees of the

department. A job satisfaction survey is conducted that offers a variety of findings related to officers’ satisfaction. From a basic perspective, the data may simply describe how officers perceive satisfaction with differing aspects of their jobs, becoming descriptive in nature. However, these same findings could be used to evaluate the police agency by examining those areas where satisfaction is the lowest and leading to efforts to determine how to improve these areas. This is the applied nature of the research. The result is research that is multipurpose. There is an increasing focus designing research studies that are multipurpose in that the studies are designed to increase the knowledge of a subject area but also have an applied purpose in mind. Whether applied, basic, or multipurpose, research can provide interesting findings about a variety of problems, events, issues, or activities. Regardless of the strategies used, criminological and criminal justice research is necessary for understanding crime and criminality and for developing suitable responses.

Types of Research Before conducting research, one must understand something about research; that is, one must first

study how research is correctly conducted. At some point in one’s college career or during one’s employment, a person may be asked to look into something or research a topic. Often, the individual has no clue where to look, how to begin, or what to look for. Then, once the information is obtained, the person may not understand how the information was found, how to analyze the information, and what it actually means. The primary reason for studying research is to be able to attain a better understanding of why the research was done, what the results actually mean, and how it may be used (Bickman & Rog, 2009; Hagan, 2006; Maxfield & Babbie, 2009). Ultimately, if one does not understand what research is and how it works, one cannot understand the products of research. Therefore, the answer to why one studies research is the same reason as why one conducts research: to gain knowledge. This knowledge may occur in one of four formats or types: (1) descriptive, (2) explanatory, (3) predictive, and (4) intervening knowledge (Bachman & Schutt, 2008; Bryman, 2008; Kraska & Neuman, 2008; Lavrakas, 2008).

Descriptive Research

Knowledge that is descriptive allows one to understand the essence of a topic. Research of this nature helps one gain a better grasp of an issue or problem of which one knows little. For example, women have played some role in criminality in this country. Yet, very little is known about women and criminality, especially with respect to certain types of crime (e.g., organized crime) due to the focus on the criminality of men. Previously, the assumption was the findings for male populations would apply to female populations; however, this assumption is not necessarily true. To understand better what role women have played in crime, a descriptive study might be conducted. Descriptive knowledge is a very common result of criminal justice and criminological research. Although the results might be very informative, what can be done with this knowledge is often limited.

Explanatory Research Explanatory research tries to determine why something occurs, specifically the causes behind the event. This research can be important when trying to understand why certain types of individuals become serial murderers, for example, or what factors contribute to criminality. Knowing the causes behind something can assist in finding ways to counteract the behavior or the problem. For

example, research focusing on gang membership may help to explain why some individuals and not others join gangs. This information could assist in deterring potential future gang members. Ultimately, this type of research may provide answers to the questions of how and why.

Predictive Research Knowledge that is predictive in nature helps to establish future actions. This type of research can be useful to all criminal justice practitioners. For example, if research indicates that a large percentage of juveniles placed in boot-camp environments are less likely to become adult offenders, these results could be used in the future sentencing of juvenile offenders. Conversely, if boot camps are shown to have little or no effect, other alternatives may then be explored. Predictive knowledge gives some foresight into what may happen if something is tried or implemented. Because one of the concerns of criminal justice is to decrease criminal behavior, including recidivism, predictive knowledge could be quite useful in attaining this end.

Intervening Research Intervening knowledge allows one to intercede before a problem or issue gets too difficult to

address. This type of research can be quite significant when a problem arises that currently available means are not properly addressing. Research on the effectiveness of certain community policing programs is a good example of intervening research. It can demonstrate whether a specific type of action taken before a given point provides the desired results. Whether the research is descriptive, explanatory, predictive, or intervening, it is important to understand what research is and how it is valuable. If one fails to study research in and of itself, then all research is of little value. This statement becomes especially true for the criminal justice and criminological academic or practitioner who wants to make use of previously conducted research or to conduct his or her own research. It is important to have a grasp of what research is and why it is conducted, before one can actually conduct research.

Why Research Is Necessary There are a number of specific reasons for conducting criminal justice or criminological research. Ultimately, the reason is because it is of interest to the researcher. Typically, courses are designed around a specific theme; however, with a

research methods course, any subject could be studied. Three primary reasons include (1) curiosity, (2) a desire to address social problems, and (3) the development and testing of theories.

Curiosity Being curious is wanting to know about an existing problem, issue, policy, or outcome. For example, in the early 1990s Dantzker and Ali-Jackson were interested in what effect a course might have on students’ perceptions of policing. A primary reason for this research was the curiosity of one of the researchers who taught police courses and wanted to see whether there were any differences between perceptions at the beginning and at the end of the course. As a student, ask yourself: In what subjects or topics are you interested? What do you want to do after college in the field of criminal justice?

Social Problems The most salient social problem related to criminal justice is crime. Who commits it? Why do they act as they do? How do they do it? These are questions of interest for many criminal justice and criminological practitioners and academics. Concern over the effects of crime on society only adds further reason to conduct related research. This research can help identify who is more likely to commit

certain crimes and why, how to deal better with the offenders and the victims, and what specific parts of the system can do to help limit or even alleviate crime. As a major social problem, crime provides many reasons for research and avenues for exploration.

Theory Testing Linked more closely with pure criminological research, theories provide good cause to conduct research. The relationship between theory and research was discussed previously in this chapter. Theory construction is discussed in detail in Chapter 4.

Factors That Influence Research Decisions Regardless of why the research is conducted, one must be cognizant of factors that can influence why and how research is conducted (Bachman & Schutt, 2008; Lavrakas, 2008). These factors should be identified and carefully considered before starting the research. Three main factors that influence research are (1) social and political, (2) practical, and (3) ethical considerations (Bickman & Rog, 2009; Bryman, 2008; Hagan, 2006; Kraska & Neuman, 2008; Maxfield & Babbie, 2009).

The social and political influences are often specific to the given research. Criminology and criminal justice as social sciences are greatly influenced by social and political events. For example, race and ethnicity, economics, and gender might be influential on research about prison environments. Research on whether a particular law is working might have political ramifications. The inability of the criminal justice system to address problems identified by research may not be caused by the lack of system resources but rather by a lack of social desire or political will. When it comes to conducting research, practicality can play an extremely important role. Economics and logistics are two elements of practicality. How much will the research cost? Can it be conducted in an efficient and effective manner? Would the benefits that are anticipated justify the social, political, and economic costs? Would limited resources be taken from other areas? These are just some of the questions of practicality that could influence the conducting of research and the subsequent uses of that research. Three ethical considerations of importance in conducting research include: (1) invasion of privacy, (2) deception, and (3) potential harm. Within a free

society, citizens jealously protect their rights to privacy. These rights are not just expected by citizens but are protected by law. Deception can have adverse effects not only on the research findings but also on the individuals who were deceived by the researcher. Harm to others, especially to those who did not willingly accept such risks, must be avoided. Because ethics plays an important role in conducting research, a more indepth discussion is offered in Chapter 2. Whatever the reason, researchers must be aware of the influences that have led to the research and those that might affect the research outcomes. Each could be detrimental to the outcome of the research.

How Research Is Done Whether the research is applied or basic, certain basic steps are applicable. There are five primary steps in conducting research: (1) identifying the research problem, (2) research design, (3) data collection, (4) data analyses, and (5) reporting of results. Each of these is given greater attention later in the text, but a brief introduction here is appropriate.

Identifying the Problem

Before starting a research project, one of the most important steps is recognizing and defining what will be studied. Identifying or determining the problem, issue, or policy to be studied sets the groundwork for the rest of the research. For example, it is impractical to embark on the study of crime without focusing on a specific aspect of crime, such as types, causes, or punishments. Therefore, it is important first to specify the target of the research. Doing this step makes it easier to complete the remaining stages.

Research Design The research design is the blueprint, which outlines how the research is to be conducted. Although the design depends on the nature of the research, there are several common designs used in criminal justice and criminology. Keep in mind that researchers may incorporate more than one type of research design into their study. Various designs are presented in this section and discussed in detail in later chapters.

Survey Research One of the most often used methods of research is surveys. This approach obtains data directly from the targeted sources and is often conducted through self-administered or interview questionnaires. This

type of research generally allows for use of a large sample.

Field Research Field research is when researchers gather data through firsthand observations of their targets. For example, if a researcher wanted to learn more about gang membership and activities, he or she might try “running” with a gang as a participant-observer. This design is one of the more time-consuming and limiting designs.

Experimental Research Experimental research is also observational research. Unlike field research, however, experimental studies involve the administration of research stimuli to participants in a controlled environment. Because of ethical and economic concerns, this kind of experimental research is conducted less frequently in criminal justice than are other research strategies.

Case Studies One of the simplest methods of research in criminology and criminal justice is the use of life histories or case studies focusing on individuals. Often these studies require the review and analysis of documents. This type of research might focus on

violent behavior where the researcher investigates the lives of serial murderers to try to comprehend why the persons acted in a particular manner.

Existing Data Another research design is where the researcher evaluates and analyzes official records for relevant data. For example, to determine patterns and influences of robbery, the research design might use data from Uniform Crime Reports. Previously collected data can be a time saver, but the timeliness of the data may come into question. Often it can take 6 months to more than a year for data to be released due to the time it takes to collect the data and make it available for public consumption.

Content Analysis In this research design, documents, publications, or presentations are reviewed and analyzed. A researcher might review existing documents to determine how crime events were publicized in a prior century or may monitor current television broadcasts to assess how the entertainment media influences public perceptions of crime. To identify the qualifications sought for police chiefs, a researcher could review published advertisements for the position of police chief.

These designs offer a variety of options. Other possible design methods are discussed later in the text. Ultimately, the design used depends on the nature of the study.

Data Collection Regardless of the research design, data collection is a key component. A variety of methods (discussed in more detail later in the text) exist. Data collection is closely tied to the research design.

Data Analysis How to analyze and interpret the data is more appropriately discussed in another course, perhaps one focusing on quantitative statistics or qualitative analysis. However, it is an important part of the design and cannot be ignored. The most common means for data analysis today is through the use of a computer and specifically oriented software.

Reporting The final phase of any research project is the reporting of the findings. This process can be completed through various means: reports, journals, books, or presentations. How the findings are reported largely depends on the target audience (e.g., police, lawyers, scholars, government officials). Regardless of the audience or the medium

used, the findings must be coherent and understandable or they are of no use to anyone. There is one last area worthy of a brief discussion. Information has been offered on why and how to conduct research, but when is it inappropriate to conduct research? Often it seems that research is conducted with little concern as to the appropriateness of the research. Failing to consider this might render the findings useless. Therefore, it has been suggested that the prospective researcher be able to answer the following questions with a negative response (Eck & La Vigne, 1994): 1. Does the research problem involve question(s) of value rather than fact? 2. Is the solution to the research question already predetermined, effectively annulling the findings? 3. Is it impossible to conduct the research effectively and efficiently? 4. Are the research issues vague and ill defined? If the answer to any of these questions is yes, the research in question should be avoided.

Summary

Conducting criminological research goes beyond looking up material on a subject and writing a descriptive paper. Before conducting research, one must understand what it is, why it is, and how it might be conducted. For the purposes of this text criminal justice and criminological research are defined as the investigation into or of any phenomenon linked to any or all aspects of the criminal justice system. The type of research conducted can be applied, basic, or multipurpose. A primary reason for conducting research is to gain knowledge, which can be descriptive, explanatory, predictive, or intervening in nature. Studying research is required to improve understanding of the results offered. All research tends to follow five basic steps: (1) recognizing and defining a problem, issue, or policy for study; (2) designing the research; (3) collecting data; (4) analyzing the data; and (5) reporting the findings. Finally, it is important to determine whether it is prudent to conduct the research in question. Research plays a crucial role in criminal justice and criminology. It brings questions and answers, debates, and issues. Knowing what it is, why it is done, and how it can be accomplished is necessary if one is to study crime and criminal behavior.

APPLICATION EXERCISES At the end of each chapter, you will find application and research exercises connected to the chapter. Application exercises provide a real-world scenario and guide you through applying the concepts and processes learned during the chapter. The research exercises are designed to guide you through developing your own research ideas. Some exercises (application or research) will involve brainstorming. In this context, the term brainstorming is used to describe a creative process in which the reader will develop ideas related to a provided topic. After brainstorming, you will evaluate the ideas. Some ideas you will keep and other ideas you will discard. Did you ever throw spaghetti at the wall to see if it was done? Sometimes it is done (it sticks), and other times it is not done (it falls off the wall). The point of this example is not to have you throw spaghetti at a wall but to point out that you will keep some ideas (they will stick) while others will be discarded. You are interested in conducting a study to measure job satisfaction among

correctional officers in your county’s jail facility. 1. In talking to a local sheriff, he tells you that he believes his employees are generally satisfied with the department simply based on his casual observations. You know that casual observations can be flawed for a number of reasons. How would you explain this to him? 2. Having explained to the sheriff how casual observation can be error prone, explain how using the scientific method can reduce errors in casual observation. 3. The sheriff insists that it would be easy to deduce the level of job satisfaction of his employees. Explain the differences between inductive and deductive logic using employee satisfaction as a topic. 4. Describe the purpose of a study of jail employee satisfaction for (1) basic research, (2) applied research, or (3) multipurpose research 5. What type of research design might you use? Why?

6. Briefly describe the steps you would take to complete the research, keeping in mind the purpose of the research.

RESEARCH EXERCISES 1. Develop other ideas that you may be interested in researching (using Box 11 to get started). Write a different topic on each index card (using at least five different cards), then write possible research questions on each index card (add to the research questions throughout the course). 2. In class, compare the information on your notecards (including your ideas and research questions) in groups of no more than three students. Provide feedback to students within the groups regarding the ideas on the notecards, and use the feedback that you receive to modify your own notecards. 3. After incorporating feedback from your classmates, choose one of your ideas and present it to the class. This can be accomplished by writing information on a whiteboard or PowerPoint for smaller classes or making the information available online for students to provide

feedback on the ideas of other students.

CHAPTER 2: Research and Ethics What You Should Know! Conducting research can be simplistic and uncomplicated; however, research in the field of criminology and criminal justice often involves people, which brings along its own set of special challenges. The previous chapter set the foundation for understanding what it means to conduct research in criminology and criminal justice. It is important that prospective researchers be aware of the ethical aspects of research and apply appropriate ethics as they design their research study. After completing this chapter, the reader should be able to: 1. Define what is meant by ethics and explain its importance in criminal justice and criminological research. 2. Present and discuss the various characteristics of ethical problems in criminal justice and criminological research, including

3. 4.

5.

6.

7.

ethical ramification, harm to others, privacy concerns, and voluntary participation. Explain how the researcher’s role influences and is influenced by ethical concerns. Describe the relationship that exists between ethics and professionalism, including a code of ethics. List and describe the four ethical criteria, including avoiding harmful research, being objective, using integrity, and protecting confidentiality. Present and discuss the five reasons why confidentiality and privacy are important research concerns. Describe the impacts of institutional review boards and research guidelines (such as those mandated by the National Institute of Justice) on criminal justice and criminological research.

Ethics Ethics as discussed in this chapter refers to doing what is morally and legally right in conducting research. The role of ethics in research requires the researcher to be knowledgeable about what is being done; to use ethical reasoning when making decisions; to be both intellectual and truthful in the research approach and reporting; and to consider

the consequences, in particular, to be sure that the outcome of the research outweighs any negatives that might occur. Using this approach, ethical decisions are much easier. Criminology and criminal justice are virtual minefields of ethical dilemmas. There is virtually no aspect of these fields in which ethical questions or predicaments do not exist, including research, particularly applied research. The ethical issues encountered in applied social research are subtle and complex, raising difficult moral quandaries that, at least on a superficial level, seem nearly impossible to resolve. These quandaries often require the researcher to strike a delicate balance between the scientific requirements of methodology and the human rights and values potentially threatened by the research (Bryman, 2008; Creswell, 2008). Criminal justice and criminological research almost always involve dealings with humans and human behavior. It is prudent to be aware of the characteristics associated with ethical problems in social research. Although there does not seem to be a consensus as to what these characteristics are and there is no comprehensive list, the following have been identified as recognizable characteristics

of ethical problems (Bachman & Schutt, 2008; Dunn, 2009; Kraska & Neuman, 2008; McBurney & White, 2007): 1. A single research problem can generate numerous questions regarding appropriate behavior on the part of the researcher. 2. Ethical sensitivity is a necessity but is not necessarily sufficient to solve problems that might arise. 3. Ethical dilemmas result from conflicting values as to what should receive priority on the part of the researcher. 4. Ethical concerns can relate to both the research topic and the ways in which the research is conducted. 5. Ethical concerns involve both personal and professional elements in the research. When dealing with humans, ethics plays an important role. It all begins with the researcher’s role.

The Researcher’s Role Contrary to popular belief, the criminologist who conducts research is considered a scientist. Ignoring the distinctions made between a natural scientist, such as a chemist or biologist, and a social

scientist studying sociology or criminology, for example, both are scientists who are governed by the laws of inquiry (Kaplan, 1963). Both require an ethically neutral, objective approach to research. Ethical neutrality requires that the researcher’s moral or ethical beliefs not be allowed to influence the gathering of data or the conclusions that are made from analyzing the data. Objectivity means striving to prevent personal ideology or prejudices from influencing the process. The focus of both ethical neutrality and objectivity is to maintain the integrity of the research. In addition to these concerns, the researcher, whether a nuclear physicist or a criminologist, must also ensure that the research itself does not negatively impact the safety of others. The researcher’s role often coexists and at times even conflicts with other important roles, such as practitioner, teacher, academic, scholar, and citizen. Since the primary focus of each role is different, the meshing of roles can often cause researchers to lose objectivity in their approach to the collection, analysis, and reporting of the data. In particular, there are the concerns over the individual’s morals, values, attitude, and beliefs interfering with completing an objective study. Researchers have to remain alert to this issue because it is easy to slide

in to habits that negatively impact the objectivity of a study. Individuals are raised with certain ideals, identified as morals and values, which are commonly reflected in their attitudes and behaviors. Weak or strong morals and values can affect how research is conducted. For example, individuals raised to believe that success is important, regardless of the costs, might regard the “borrowing” of someone else’s research efforts and passing them off as their own as acceptable, or they might accept the manipulation of data to gain more desirable results. An even more repugnant scenario is one in which the researcher continues with his or her research despite knowing that to do so will cause physical harm or emotional anguish for others. In each of these cases, ethically the decisions are wrong. Because the researcher’s role is intertwined with other roles, ethics becomes even more difficult to manage. Ultimately, it is up to the individual to decide the importance of personal ethics. However, the role of the researcher is just one aspect of ethics in research.

Box 2-1 Responsible Conduct in Research

Responsible conduct on behalf of researchers is of such importance that many institutions and funding organizations require that researchers participate in mandatory training. One such program is Responsible Conduct in Research (RCR) requirements and training. This may be viewed at: https://www.citiprogram.org/index.cfm? pageID=265

Belmont Report The Belmont Report (1979) was released following several serious incidents of unethical research practices. The Nazis used unwilling prisoners as subjects for medical experiments. In the 1940s, the Tuskegee syphilis study used black men to study the impact of untreated syphilis. To allow the continued study of these subjects, they were not provided with an effective treatment that had been developed. In response to these incidents, the Belmont Report specified three basic ethical principles: (1) respect for persons, (2) beneficence, and (3) justice. Regarding the first principle of respect for persons, individuals should be treated as autonomous, in other words being capable of making their own decisions, and special protections should be provided for those whose ability to make decisions is limited, such as children and prisoners.

According to the principle of beneficence, the research should do no harm while maximizing possible benefits and limiting possible harm. Justice refers to the fair distribution of receiving the benefits from the research and bearing the burden of the research. For example, black men were selected for the Tuskegee syphilis study even though this disease impacts other populations. These three principles from the Belmont Report provide the framework for ethical practices in modern research.

Ethical Considerations Conducting research in and of itself can be problematic. Accessibility, funding, timing, and other factors may all impose problems. The reality is that there can be ethical concerns at every step of the research process (Bickman & Rog, 2009). With this in mind, the following considerations should be kept in mind throughout the research process and not just during the design portion.

Ethical Ramifications One of the first things to consider is whether the topic to be studied has innate ethical ramifications. Some topics are controversial by their very nature. For instance, the individual interested in gangs might decide that the best way to collect data is to become a participant observer, which means the

researcher may witness or even be asked to participate in illegal activity. Ethically as well as legally, information regarding criminal activities should be given to the police, but doing so might jeopardize the research. Although it is apparent what decision should be made (the research should be adjusted to avoid such a dilemma or possibly even abandoned outright), the right decision is not always made simply because of how important the research is perceived to be to the individual. Therefore, before embarking on a research topic, the ethical implications of the research itself must be addressed.

Harm to Others Another consideration is what effects the research might have on the research subjects. When the research involves direct human contact, ethics plays an important role. Whether the subjects are victims, accused offenders, convicted offenders, individuals working within the criminal justice system, or the general public, a major consideration is whether the research might cause them any harm. Harm can be physical, psychological, social, economic, or legal. Physical harm most often can occur during experimental or applied types of research, such as testing new drugs due to unanticipated side effects.

Psychological harm might result due to the type of information being gathered. For example, in a study of victims of sexual assault, the research might delve into the events before, during, and after the assault. This line of questioning may be traumatizing to the victim and more psychological harm may be added to the harm that already exists as a result of the assault. Social harm may be inflicted if certain information is released that should not have been. Consider a survey of sexual orientation among correctional officers where it becomes public knowledge as to who is gay. This information may cause those individuals to be treated differently, perhaps discriminated against, causing sociological harm. Economic harm may occur, for example, if an individual admits to illegal drug use during a survey and the subject’s employer receives this information. The employer may fire the subject, resulting in a loss of income and in many cases loss of health insurance. Legal harm is a matter of particular importance when researching criminal justice or criminology topics. Subjects may admit to illegal behavior, which if released, could result in an investigation, conviction, and sentence. It is important that the researcher consider what type of

harm may befall respondents or participants before starting the research.

Privacy Concerns The right to privacy is another ethical consideration. Individuals in the United States have a basic right to privacy. In many cases, research efforts may violate that right. How far should individuals be allowed to pry into the private or public lives of others in the name of research? Ethically speaking, if a person does not want his or her life examined, then that right should be respected. All persons have a right to anonymity. However, there are a variety of documents accessible to the public in which information can be gathered that individuals would prefer to be unavailable, such as arrest records, court dockets, and tax and property records. The ethical question that arises here is whether a person should have the right to consent to give researchers access to certain types of information. Giving consent in general is a major ethical consideration. Anonymity and confidentiality are two oftenconfused words in research. Anonymity means that the subject’s identity is not known. Anonymous data collection could be completed by using a number system to link a survey and consent form, but the number does not link to a specific individual.

Confidentiality indicates that the researcher knows the identity of each research subject but has agreed to not reveal individualized information. Using the previous example, the number on the survey and consent form may be linked to a number on a roster of subjects that could include the subjects’ names and contact information. In data analyses, researchers may indicate that only aggregate data, not individual data, will be released in subsequent reports. Particularly in survey research, it is common for the researcher either to ask for specific consent from the respondents or at least acknowledge that by completing the survey, the respondent has conferred consent. Normally, this process only requires having the individual sign an informed consent form or indicating in the instructions that the survey is confidential and voluntary and that the information will only be used for the purpose of research (Figure 2-1).

Voluntary Participation In the example shared in Figure 2-1, not only did the researchers seek to obtain consent, they also informed prospective respondents that participation was voluntary. Sometimes researchers, including criminologists, require their subjects to sign consent

forms but, particularly within institutional settings such as military organizations, schools, and prisons, neglect to inform them that their participation is voluntary. In fact, in these environments, participation is often coerced since individuals do not feel they can decline to participate. Not all research must use voluntary participation, but compulsory involvement must be based on valid reasons that the knowledge could not otherwise be reasonably obtained and that no harm will come to participants. The general rule in these situations is if there is any doubt as to whether the research could be in any way construed to be intrusive, then consent should be obtained from the subjects. It is also best to assure subjects that their participation is voluntary and they may choose not to participate in the study. Subjects should also be reminded that they can withdraw their participation at any time during the study. Within the academic setting, informed consent and voluntary participation do not seem to be unusual requirements. To ensure that informed consent is provided and to judge the value and ethical nature of the research, many universities have an Institutional Review Board (IRB). IRBs exist as a result of the Code of Federal Regulations, Title 28 Judicial Administration, part 46, which specifies all

aspects of the IRB, including membership, functions and operations, review of research, and criteria for IRB approval (http://www.hhs.gov/ohrp/assurances/index.html). Established primarily for the review of research, usually experimental or applied, dealing directly with human subjects, university IRBs often extend their review over any type of research involving human respondents, survey or otherwise. Although having to attain IRB approval can be frustrating, it is a useful process because it can help identify prospective ethical problems. Also, reviewers may identify problems that have been overlooked by the researcher. It is better to err on the side of caution.

Figure 2-1 Informed consent example. Source: Reproduced from Dantzker, M. L. Psychologists’ Role and Police Pre-Employment Psychological Screening. ProQuest Company, 2010.

The IRB process varies by institution but generally is not difficult. It usually requires the researcher to

submit basic information about the proposed research, often in a format designed by the university. The information required is similar across institutions and typically focuses on study procedures, with a particular interest on possible harmful impact to subjects. Informed consent is valuable because it is important that research subjects are allowed the right to refuse to be part of the research. Although in survey research, consent may not be a major problem (because permission can be written into the documents), it does raise an interesting dilemma for observational research (when the researcher may not want the subjects to know they are being observed). The ethical consideration here is that as long as the subjects are doing what they normally would be doing and the observations do not in any way directly influence their behavior or harm them, it is ethically acceptable to proceed without obtaining informed consent.

Deception Some types of research, particularly field research that requires the researcher to in essence “go undercover,” cannot be conducted if the subjects are aware that they are being studied because subjects may change their behavior. Such research

is controversial and must be carefully thought out before it is undertaken (Vito, Kunselman, & Tewksbury, 2008). All too often the deception is based more on the researcher’s laziness or bias rather than a real need to deceive. For example, a researcher is interested in studying offender behavior within the confines of a prison. Rather than explain to administrators and the subjects what he or she is doing, the researcher proceeds under the guise of an internship or volunteer work. Depending on the type of research, there are always some ethical considerations. What is interesting is that the science of research itself is viewed as ethically neutral or amoral. The ethical dilemmas rise from the fact that researchers themselves are not neutral. This situation fosters the need for regulation in conducting research to ensure that it meets ethical standards (Fowler, 2009; Gavin, 2008).

The Professionalism of Research According to Merriam-Webster (2016b), professionalism is defined as “conduct, aims, or qualities . . . [that] characterize or mark a profession or a professional person”. A profession is defined as “a: a calling requiring specialized knowledge and often long and intensive academic preparation b: a

principal calling, vocation, or employment c: the whole body of persons engaged in a calling” (Merriam-Webster, 2016a). Research in itself is a profession, and when mixed with other professions, there is an even greater need to conduct business in a professional manner. Many professions have support of written codes of ethics for research (e.g., The American Psychological Association). Although criminal justice and criminology do not have a globally applicable code, the Academy of Criminal Justice Sciences developed a code of ethics for its members with a section on researcher ethics (Part III Ethical Standards – B Member of the Academy as Researchers) that highlights the expected ethical behavior of criminologists. Furthermore, although there seems to be no universal code of ethics with respect to research, grant-funded research is more likely to have ethical constraints imposed. For example, a popular source of funding for criminal justice and criminological research is the National Institute of Justice (NIJ). NIJ has developed its own code of ethics to which all grant recipients must agree. The NIJ is very specific in its guidelines, especially with respect to data confidentiality and the protection of human subjects (Figure 2-2).

Ethical Research Criteria Even though there is no universally recognized research code of ethics, there are some specifically identified criteria that, when applied or followed, assist in producing ethical research. These criteria include avoiding harmful research, being objective, using integrity, and protecting confidentiality.

Avoiding Harmful Research The goal of research is to discover knowledge not previously known or to verify existing data. In many instances, this goal can be met without inflicting undue stress, strain, or pain on respondents (e.g., historical or survey research). Unfortunately, at times research can be physically or emotionally harmful. The ethical approach is to avoid any such research regardless of how important its findings might be unless it can be shown that the benefit from the information outweighs the harm. IRBs examine studies overall but often focus on the possible harms and the question of whether the benefit from the study is enough to justify these harms.

Being Objective Biases can be detrimental to a research project and can negatively affect objectivity. For example, assume you do not like drinkers and perceive them

to be weak willed and careless. Your research deals with individuals convicted of driving while intoxicated, and you are interested in their reasons for driving while impaired. The chances are good that if you allow your personal feelings against drinkers to guide you in your research, the results will be skewed, biased, and subjective, which is why it is important to maintain objectivity. Of course, being objective is just one important characteristic of the ethical researcher.

Figure 2-2 NIJ human subjects and privacy protection.

Source: Reproduced from “Human Subjects and Privacy Protection,” National Institute of Justice, 2010.

Using Integrity The last thing a researcher wants is for the results not to meet expectations. Sometimes, because of how important the research is perceived to be, there may be a tendency to manipulate the data and report it in a manner that shows the research was successful—that is, to put a positive spin on an otherwise negative result. This situation is especially possible when the research is evaluative and its results could influence additional funding for the program being evaluated. When faced with this dilemma, because of the desire not to jeopardize the program’s future or to improve future chances for research, the researcher may not report the true findings. This behavior is unethical but may be more commonplace than one would like to believe. The ethical researcher accepts the findings and reports them as discovered.

Protecting Confidentiality One of the biggest concerns in conducting research is the issue of confidentiality or privacy. These two ethical issues are crucial to social researchers who, by the very nature of their research, frequently request individuals to share with them their

thoughts, attitudes, and experiences. Because a good portion of criminal justice and criminological research involves humans, chances are great that sensitive information may be obtained in which other nonresearch efforts might be interested. For example, conducting gang research where street names and legal names are collected, perhaps along with identifying tattoos, scars, and so forth, and voluntary statements of criminal history would be extremely valuable to a police agency. Ethically, that information must remain confidential.

Reasons for Confidentiality and Privacy Overall, five reasons have been identified as to why confidentiality and privacy are important in research (Adler & Clark, 2007; Kline, 2009; Maxfield & Babbie, 2009): 1. Disclosure of particularly embarrassing or sensitive information may present the respondent with a risk of psychological, social, economic, or legal harm. 2. Sensitive information, if obtained solely for research purposes, is legally protected in situations where respondents’ privacy rights are protected.

3. Long-term research may require data storage of information that can identify the participants. 4. The courts can subpoena data. 5. Respondents may be suspicious as to how the information is truly going to be used. The bottom line is that confidentiality and privacy must be maintained. There are two methods of accomplishing this: physical protection and legal protection. Physical protection relates to setting up the data so that links cannot be made between identifying information and the respondents. Reducing who has access can also aid in protecting the data. Legal protection attempts to avoid official misuse. Researchers are aided in this aim by an amendment to the 1973 Omnibus Crime Control and Safe Streets Act, better known as the “Shield Law,” which protects research findings from administrative or judicial processes. As noted previously, funded research through such organizations as NIJ or the National Institutes of Health is overseen by organizational regulations. These guidelines do not completely protect the data, leaving researchers responsible for gathering the data in a manner that best protects the respondents.

By simply meeting the four suggested criteria, a researcher can avoid many ethical problems. However, perhaps the best way to avoid ethical problems is to conduct research using a method that does not compromise ethical standards— research that is legal, relevant, and necessary.

Summary The simple act of research, especially when it involves humans, creates a plethora of possible ethical dilemmas. Because ethics is important to professions, researchers need to be cognizant of several ethical considerations. These include determining whether the topic itself is ethical, what harm or risk is involved to respondents, and how confidentiality and privacy can be protected. There are federal guidelines for protecting individuals’ privacy and for obtaining their consent, which in the university setting is often reinforced through an IRB. The key to ethical research is a professional approach. Some professions have created a code of ethics applicable to research. Although criminal justice and criminology do not have one specific to the discipline, the Academy of Criminal Justice Sciences has established such a code for its members. Four criteria, when followed, alleviate the need for such a code: (1) avoid conducting harmful research; (2) be objective; (3) use integrity in

conducting and reporting the research; and (4) protect confidentiality.

APPLICATION EXERCISES 1. You are currently employed as a data analyst within the research unit of a medium security prison. Due to complaints regarding the negative impact of job stress on correctional officers, the warden has directed the research unit to study two issues: (1) the level of stress among correctional officers and (2) ways to address stress through physical fitness. Does the warden have a vested interest in the results of the study? If so, explain how this interest in combination with the warden’s role as the person in charge of the prison (including the research unit) could impact the implementation and results of the study. 2. Explain how this dual role could impact the voluntary participation of correctional officers, who are employees of the warden. Develop procedures to ensure that correctional officers do not feel coerced into participating in the study or into

providing answers that they think are expected by the warden. 3. Because you work in the prison, you know some of the correctional officers. What are possible issues of conflict with your role as a researcher and staff member? What procedures will you put into place during the study to avoid allowing these dual roles to negatively impact the quality of the research? 4. Researchers who are working with human subjects have to be particularly mindful of the possible harm that could be caused, directly or indirectly, by the research. Explain how your study of correctional officer stress would safeguard against physical, psychological, social, economic, or legal harm. 5. Explain the difference between anonymity and confidentiality. Explain why it is important to ensure confidentiality and privacy regarding the information collected from correctional guards.

RESEARCH EXCERCISES

1. Research the following historical events and describe what occurred: (1) Nazi medical experiments, (2) Willowbrook State School Study, (3) Tuskegee Syphilis Experiment, and (4) Stanford Prison Experiment. For each of the four events, describe the historical context and explain how this setting impacted the course of events. Keeping in mind that these events are historical, what are current expectations regarding ethics in research that you would expect to keep history from repeating itself? 2. Evaluate the actions of the researchers and determine the possible harm that was suffered by the participants of each event. Describe the possible long- and short-term impacts of these harms. 3. Review the Academy of Criminal Justice Sciences Code of Ethics Part III Ethical Standards – B Member of the Academy as Researchers ((http://www.acjs.org/pubs/167_671_2922.cfm) and summarize the key points. 4. Find the IRB information for your college or university. Review the

directions and forms. Make a checklist of information that you would need to propose your own study. 5. Some colleges and universities have started to require that students complete training related to the ethics in research before participating in research activities, such as surveying fellow students as part of a class project. Complete the four modules for undergraduate students in the Collaborative Institutional Training Initiative also known as CITI (https://www.citiprogram.org/). Once you have completed the training, you will receive a certificate.

CHAPTER 3: The Beginning Basics What You Should Know! For established researchers, knowing what to study and how to study is often simple. However, established researchers can also become stuck in a rut and need to expand beyond their typical areas of research. For most beginning student researchers a consistent issue is having difficulty in choosing what to research, and the number of possible research topics can be daunting. A major question one might ask is, where do I start? Choosing a topic is merely the first step. Other issues include developing the research question, a statement explaining what it is the research is supposed to accomplish. Forming the research question is a critical step. The research question should narrow the topic to a manageable focus and leads to the formation of hypotheses and the identification of variables. After completing this chapter, the reader should be able to: 1. Discuss the issues that should be considered in selecting a research topic.

2. Present and describe the three purposes of research. 3. Describe what a literature review is and what sources are available for such a review. 4. Compare and contrast the various writing styles used by criminologists. 5. Define what an article critique is and discuss its details. 6. Define what is meant by the research question. Give an example of a research question. 7. Define the term hypothesis and describe the types of hypotheses.

Getting Started In the two previous chapters, we discussed why research is necessary and why research ethics are crucial. These are important topics that warrant serious consideration. When charged with the task of writing an empirical research paper, there is a great deal to know about “how” to complete such a project satisfactorily after determining the “why.” This chapter presents and discusses a number of issues that must be considered when starting a research project and provides a guide on how to break down the extensive information available to manageable portions.

One of the most difficult aspects of any endeavor is to begin, and research is no exception. The beginning of the process is where procrastination has the biggest impact because students just don’t know where to start. Of course, some students will argue that they work best under pressure, but just ask any of your professors and they can tell you which term papers were written the night before. By its very nature, the process by which a research problem is selected is a creative process. Keeping in mind that during the creative process there are not right or wrong answers can decrease the amount of stress students feel during this stage of the process. This process may require thinking about various ideas and issues, asking questions, and determining whether the question may actually have answers (Bryman, 2008; Creswell, 2008; Drake & Jonson-Reid, 2008).

Picking a Topic Before beginning the research project, a researcher must first answer this: what should I study? Within the fields of criminology and criminal justice, numerous research topics are available. All one has to do is pick a topic; however, that is not as easy as it may seem.

The beginning of any research project must focus on what is to be studied. Defining the problem is viewed as the most important stage of the research (Fowler, 2009; Gavin, 2008; Hagan, 2006). Crime is a popular topic. It is the basis of television shows and movies and often leads the evening newscasts. However, the topic of crime is too broad to effectively study. Instead of focusing on crime in general, a researcher might narrow the topic to property crime. From this point, one can narrow the topic further by focusing on specific types of property crime, such as theft, vandalism, or destruction of property. To start, the research project should be something of personal interest. If there is no interest in the topic from the start, one will be tired of it before it is completed. It must also comply with any topic restrictions imposed by the individuals or organization for which the research is being conducted. For example, professors frequently restrict the topics that their students may research to avoid emotional diatribes on controversial issues better left to experienced researchers. In addition, before the topic is chosen, one should consider (1) the availability of current research on the topic, (2) any gaps in theory or in the literature available on the topic, (3) the feasibility of conducting the

research, (4) the potential for policy implications, and (5) possible funding availability (Fowler, 2009; Gavin, 2008; Hagan, 2006; Maxfield & Babbie, 2009; McBurney & White, 2007; Vito, Kunselman, & Tewksbury, 2008). There are a number of studies about community policing. Many of them suggest and support the effectiveness of community policing. What seems to be lacking are studies that explain why community policing may be successful in some places and a failure in other places. For example, does success of community policing strategies differ based on urban versus suburban locations? In this example, a gap in theory exists and needs to be filled. Finding such gaps in the literature can assist in choosing a topic. Once an interesting and intriguing topic is found, the feasibility of conducting the research must be considered. Feasibility is primarily linked to logistics (e.g., Is a sample accessible? Does a data collection instrument exist or must one be designed?). Sometimes the topic may be a very good choice for research but is not feasible to attempt. Before finalizing the research topic, the prospective researcher needs to be sure that the study can actually be accomplished. Doctoral

students working on their dissertation are often tripped up at this step. Because of the popularity of some topics, such as job stress, capital punishment, sentencing disparity, and community policing, there is usually a wealth of information available to help build a research base. However, there may be times when the topic is legitimate, but little information exists in the literature to support the research question. This should not stop one from going forward with the research. If the findings can be validated (i.e., shown to meet scientific rigor and supported by ensuing research), it may become a new and significant contribution to the literature, the discipline, and subsequently the practice of criminal justice or criminology. Choosing a research topic that could have policy implications can be very useful. For example, the results of a study on examining job satisfaction or examining the effect education has on promotion could have an effect on policy and procedure. Finally, although funding may not be applicable to many students’ research efforts, it should be taken into consideration. One of the most popular means of funding research is through internal and external

grants. Many universities offer both students and faculty opportunities to apply for internal grants that at least allow the person to start the research project and may help offset personal costs. Ultimately, many researchers seek external grants. However, these are usually not sought by undergraduate students. You may have the greatest idea in the world, but if it requires extensive funding that you do not have, then this situation can create an obstacle to successfully completing the study. Regardless of where the funding may come from, it is important to establish whether any monies are needed and from where they may be sought.

The Purpose of the Research Once consideration is given to the previously mentioned issues, the research itself can begin. To address these issues, one must have an idea of what the research will cover. Perhaps the best place to start is to decide what the research should accomplish. There are three possible expectations or accomplishments: (1) exploring, (2) describing, and (3) explaining (Gavin, 2008; Hagan, 2006; Maxfield & Babbie, 2009).

Exploring Most of us are explorers by nature. Our curiosity about things begins at an early age and persists

throughout our lives. What we explore and how we explore something changes with age and time. If you have spent time around children, you know that at a certain age they explore the world by putting items in their mouths. However, as we mature, so do the ways we explore our world. An older child may ask a parent how electricity works. Exploration occurs either accidentally or intentionally. Intentional exploration may well be considered to be a form of research. Thus, when one wants to know more about something, the tendency is to explore the topic. For example, before buying a new sports utility vehicle, a consumer might read what trade magazines say about them, check out prices through various dealers, and talk with current owners of such vehicles. When finished, one should have enough information to make an informed decision based on the research that has been conducted. With respect to criminal justice and criminology, exploration of one’s interests is often formal, intentional, and accomplished through some form of research.

FROM THE REAL WORLD The attitudes of college students can provide interesting insight as to how young adults

tend to think of various topics. In addition to the school-related interest, such as classes, grades, parking, and graduation, sexuality is an area of particular interest among college students. Because criminal justice students are perceived as a more conservative group, Dantzker and Eisenman (2007) speculated that the attitudes regarding sexuality would differ based on major and that criminal justice students would have conservative attitudes toward sexuality. To study the issue, they conducted a study among criminal justice students examining their attitudes toward homosexuality, pornography, and other sexual matters. The study findings did tend to support the perception that these students possess a conservative attitude.

Exploratory research provides information on a topic on which little is known. Therefore the basis for other types of research such as explanatory research does not yet exist. Exploratory research seeks out information about something that is known to exist, but as to why or how, that has not yet been discovered. The goal of exploratory research is that it will offer additional insights about

something for which there is awareness but limited knowledge.

Describing The purpose of descriptive research is to describe specific aspects or elements of the topic. What is the phenomenon? What are its characteristics? How often is the phenomenon occurring? These are just some of the types of questions answered when conducting research for the purpose of describing. For example, the Bureau of Justice Statistics releases the Correctional Populations in the United States report annually. Information within this descriptive study includes the number of individuals supervised by the adult correctional systems, incarcerated (prison or jail), or supervised in the community (probation or parole), as well as changes in these numbers over the years. A descriptive study specifically on prisons could offer information about the inmates, correctional staff, programs, violent acts, and so forth, which would further enhance what is already known about prisons. This descriptive information often provides the basis for an explanatory study because a researcher first needs to know if there is a phenomenon to study. For this reason many explanatory studies will include a descriptive

section. Descriptive studies are probably among the easier studies to do because the researcher simply needs to explain what he or she sees, hears, or reads with respect to the various elements of the topic.

Explaining Undoubtedly the most in-depth and difficult purpose for conducting research is to provide an explanation. Explanatory research attempts to analyze and fully understand why a phenomenon occurs as it applies to policies, procedures, objects, attitudes, opinions, and so forth. From a criminological perspective, researchers may attempt to understand better why certain males commit sexual assaults and other do not. Another example of explanatory research would be a study examining acts of police corruption to determine why they occur. Sometimes a study may start out exploratory and then be expanded into a descriptive or explanatory study. For example, due to a report of increased criminal behavior, criminologists sought to find out what was going on in a rural high school. They conducted exploratory research to find out about the criminal behaviors. On learning that it was gang related, they could rely on the literature to make recommendations as to how to deal with the

problem. Or they might engage in further descriptive research to learn exactly how the gangs were made up and what they were doing. This process could then lead to explanatory research, which could examine why gangs were able to get a foothold in the high school, what the impacts of such activities are, and how to combat these gang activities. Although all three reasons for conducting research are valid, it is most often the explanatory reason that many criminologists pursue and that is considered the gold standard of research. This type of study relies on strong, clear research questions, hypotheses, and variables. Each of these is discussed in more detail later in this chapter. However, there is still the question of choosing a topic. To demonstrate the myriad possible topics for study, the following is a short list of potential research examples. 1. Examine the role of criminal justice programs in a given state. 2. Survey education and job satisfaction among police, parole, or correctional officers. 3. Compare and contrast criminal justice educators: practitioner-academic versus pure academic and their work products.

4. Evaluate the relationships between job stress and race and ethnicity. 5. Explore the causes and extent of criminality among high school students. 6. Conduct a cost-benefit analysis of private versus state prisons. 7. Research drug usage among adolescents. 8. Evaluate the success of electronic monitoring. 9. Evaluate the effectiveness of community policing. 10. Survey public attitudes toward corporal punishment. The process of choosing a topic can be aided through personal observations, suggestions from academics and other students, and the existing literature in criminal justice and criminology. It is through this third means that a large number of research topics are chosen.

Reviewing the Literature For many researchers, the choice of an idea or concept to study may, at times, be frustrating. What becomes more frustrating is choosing a topic and finding either too little or too much information available in the current literature. Neither of these situations should prevent the individual from conducting the research. There is usually a

substantial amount of literature to support most topics one might wish to research in criminal justice and criminology. With the extensive number of journals currently available, data from government agencies, and Internet access to information, one can usually collect enough information to support a research effort. Ultimately, the best way to begin research is to focus on a particular issue, phenomenon, or problem that most interests the individual. In doing so, one must be sure to determine the problem, issue, or phenomenon and organize what is known about it. This point is where a literature search and review is valuable, and the best place to begin the search is the library.

Become Familiar with the Library A first step in conducting a search of the literature is to familiarize yourself with your university or college library as well as your local library. As a criminal justice or criminology student, you should have access to a good library. If not, find out how far it is to a better library and make arrangements to go there. Too frequently students try to use the excuse that “it was not in our library” after a brief perusal of the texts and journals that are available, which is not an acceptable excuse. As an individual who is

capable of thinking critically, one is expected to solve problems regarding the availability of resources, not to just bemoan them. Keep in mind that your university or college library is normally part of a system that can borrow books from other libraries. Additionally, many resources, such as journal articles, are now available online. Familiarize yourself with the layout of your library and its available search vehicles, and get to know the librarians. There are many more resources available than one may have imagined.

Text and Journal Abstracts Literature or topical searches can start with the use of a source called an abstract. Two popular abstracts are the Sociological abstracts and the Criminal Justice abstracts. With these sources the researcher can look for a particular term (a key word or words, such as “job satisfaction”), concept, topic, or combination of terms to see what has already been published about this subject.

Scholarly Journals From the abstracts one can go directly to identified journals. Scholarly journals are refereed or peer reviewed (meaning that to be published the article appearing in the journal had to pass review by other scholars who were asked by the journal to critique

its contents). How critiques are conducted is discussed in a following section. The findings from a perusal of journal articles may help determine a topic as well as guide research.

Textbooks For research purposes, introductory-level textbooks are generally not of sufficient depth to use as sources in research. It is better to build from them by going to the sources that they cite within a subject area or to rely on texts devoted to the subject in question. However, textbooks are often excellent starting sources for selecting a research topic. If one is assigned a research project for a certain class, review the contents of the text that is being used in that class to determine if something of personal interest may be revealed.

Social Science Indexes Annual indexes for journals are another source of information. Today, journals are available on microfiche, as hard copy, or online. Government documents and the Internet are additional sources for helping choose a topic or gathering information to support the research topic.

Internet Searches

A word of caution is appropriate regarding the use of the Internet. Although it is useful for gaining preliminary information about a topic, it does not replace conducting a solid literature review of the subject area. For all its convenience, the Internet has a great deal of information that is unsubstantiated, if not outright erroneous. If the source is not clearly scholarly (e.g., a reputable online journal), be cautious about using it as a research reference. The Internet augments the library and social science indexes and abstracts as search vehicles; it does not replace them. Nor does it replace journals and textbooks. Although the Internet may provide legitimate sources, a reference page filled with numerous websites instead of text and journal citations is indicative of lazy scholarship on the part of the researcher. Overall, there are a number of sources from which one can choose a topic or find support for a given topic. Box 3-1 provides a sampling of refereed (peer-reviewed) journals that publish articles in criminology or criminal justice. This list is not intended to be viewed as an exhaustive catalogue. The literature search provides the basis for formulating or refining research questions.

Critiquing the Literature

To conduct a sound literature review, whether as a topical search or in developing the scholarly basis for the research being conducted, one must be able to properly interpret the research being read. This section provides guidance on what to look for in evaluating other studies. These guidelines are also helpful in preparing one’s own work for others to review. More detail on preparing work for submission is provided in a later chapter.

Understanding Writing Styles Scholarly journals and textbooks conform to specific writing styles. The various styles are precisely detailed in publication manuals. Several styles are used in criminal justice and criminology. At some point in the research process one may view Turabian (which uses numbers to indicate citations and footnotes at the bottom of each page); Chicago Style (also uses numbers after citations with endnotes instead of footnotes); American Psychological Association (APA; the most commonly used style in criminal justice and criminology, listing the author and year of publication within the text); and American Sociological Association (similar to APA but may use endnotes for specific details and varies in the format of the references). Occasionally, one may also see Modern Languages Association style and

styles unique to specific law or criminal justice journals, which are often variations of the previously mentioned styles. It is important to know what style is used when critiquing an article or text and even more important to know what style is required when submitting a paper, article, or text for review. For example, many criminal justice or criminology professors will require their students to submit all papers in APA format. The best way to become familiar with this format is through the APA stylebook.

Box 3-1 Literature Review: Examples of Sources Journals American Journal of Criminal Justice American Journal of Sociology British Journal of Criminology Canadian Journal of Criminology Crime and Delinquency Criminal Justice Policy Review Criminology

Criminology and Public Policy Justice Quarterly Journal of Contemporary Criminal Justice Journal of Criminal Justice Journal of Quantitative Criminology Police Quarterly Prison Journal Social Forces Social Justice Social Problems Social Science Quarterly Compendiums Abstracts on Criminology and Penology Criminal Justice Abstracts Police Science Abstracts Psychological Abstracts Social Science Index

Sociological Abstracts Government Agencies Bureau of Justice Statistics National Criminal Justice Reference Service (www.ncjrs.org) National Institute of Justice (www.ojp.usdoj.gov/nij)

Knowing What to Look For Most journal articles follow a similar format regardless of the publication journal with sections in this order: abstract, problem statement, literature review, methodology, findings, discussion, and conclusions. When scanning articles to see if they are of value to your topic, it is helpful to first review the abstract, methodology, and conclusions. After determining that the article is useful, you will need to read it in its entirety (Kline, 2009; Kraska & Neuman, 2008; Lavrakas, 2008).

Abstract The abstract is a brief overview of the article. What is the social issue that was studied? How was it investigated: exploration, description, explanation,

or a combination of strategies? Who conducted the research? Who financed it (and is there a conflict of interest)? Did it have clarity of purpose? Were the findings reasonable based on the research design? What were the conclusions and recommendations?

Problem Statement The problem statement identifies the research problem being addressed and as well as laying out possible research questions that the research is trying to answer. What was the problem being investigated? Did the literature review support the need for this research? What was the theoretical orientation? What was or were the research hypotheses? Were concepts properly defined? Were the dependent and independent variables clearly identified and defined?

Literature Review The purpose of the literature review is to provide a foundational understanding for the reader, as well as the researcher, for understanding and evaluating the study. Did the literature review provide a thorough coverage of the prior research? Were previous studies adequately evaluated and discussed? Was the coverage complete? Are classic studies relevant to the research problem included and recent research included? Did the

literature cited provide a justification for the current investigation?

Methodology The methodology section provides details of the procedures used within the study and should provide enough details for the reader to be able to understand, evaluate, and critique the methods used within the study. What was the research design? Was it clearly developed from the theoretical frame of reference alluded to in the problem statement? Who were the subjects, and how were they included in the study? Did the study conform to ethical standards? What was the sampling method, and was it adequate for this research? Was the measurement instrument or strategy satisfactory? How were the data analyzed? Were there any adverse effects or limitations because of the research design and the means of analysis being incompatible? Were measures of association and tests of significance clearly indicated and appropriately discussed? Were the measurement techniques valid and reliable? Methods sections that are incomplete and less than transparent create problems for readers and are ethically problematic if they were deliberately written in this manner by the researchers.

Findings The findings section summarizes the results of the analysis and is typically the location of any tables that display some of the pertinent results. Were the findings displayed in a concise and readily understandable manner? How were the data summarized? Were the tables logical and clear? Were the statistical techniques appropriate? Would other statistical techniques have been more appropriate? Did the findings relate to the problem, the method, or the theoretical framework? How did the findings relate to those of prior research?

Discussion and Conclusions Discussion and conclusion sections are designed to discuss the findings of the research and further analyze through the lens established by the problem statement and literature review. The impacts of study strengths and particularly weaknesses on the results are discussed as well as recommendations for future studies. Are the conclusions reached consistent with the findings that were presented? Are the conclusions compatible with the theoretical orientation presented in the problem statement? Based on the problem statement, the prior research, methodology, and findings, do the conclusions or recommendations make sense? Based on the problem statement and the literature, are the

conclusions or recommendations of this study a significant contribution to the field of study?

The Research Question Once a topic, or research problem, has been chosen, the next step is to create the research questions. A research question is a statement answered through the research process. It generally is formulated from the research problems or purpose and may be synonymous with the research problem. Essentially, what the researcher must do is decide what it is he or she is to study and why. It is the “why” that helps form the research question. For example, at one point, Detroit removed numerous streetlights to save money (Christoff, 2012). A research question could be “did the removal of street lights impact crime in the affected neighborhoods?” The research questions should allow others to gain a clear understanding of why the research was conducted. A well-worded research question should give some indication of the outcomes one might expect at the conclusion of the research. The question should be directly linked to the research problem.

Hypotheses

A hypothesis is a specific statement describing the expected result or the expected relationship between the independent and dependent variables. Variables are factors that can change or influence change. They result from the operationalization of a concept. There are two types of variables. Dependent variables are the factors being influenced to change. The dependent variables are the outcome items or what is being predicted. Independent variables are the factors that influence or predict the outcome of the dependent variable. Using the previous example, the presence of streetlights is the independent variable and the occurrence of crime is the dependent variable. The development of hypotheses and their linkages with theory and research is discussed further in Chapter 4. The three most common types of hypotheses are (1) the research hypothesis, which is a statement of the expected relationship between variables offered; (2) the null hypothesis, which is a statement that the relationship or difference being tested does not exist; and (3) the rival hypothesis, which is a statement offering an alternate explanation for the research findings.

FROM THE REAL WORLD

Dantzker (2010) conducted a study of police psychological pre-employment screening and examined the research problem related to whether there were differences, based on the type of psychologist involved in the process, in the evaluative instruments or protocols used in conducting preemployment screening for potential police officers and whether different types of psychologists select those instruments or protocols for reasons of job-specific validity. The work of psychological screening for potential police candidates is conducted by two types of psychologists: police and general clinical psychologists. Whether there is a difference between the two groups of psychologists in terms of this work has not been examined. One could suggest that if police psychologists do differ in this area from general clinical psychologists, then a question for future research is whether their selections of questionnaires and protocols to conduct such screenings are more applicable than those chosen by general clinical psychologists. This problem produced the following research question:

Are there differences between police psychologists and general clinical psychologists in the evaluative instruments or protocols used in conducting pre-employment screening for potential police officers and do they select those instruments or protocols for reasons of jobspecific validity? After establishing the research problem and then the research questions related to this problem, the researcher must next explain what specifically is going to be studied and the expected results. This requirement is usually accomplished through statements or propositions referred to as hypotheses.

The research hypothesis, which is the focus of the study, is generally a statement that fits the equation, “if X, then Y,” with X representing the independent variable and Y indicating the dependent variable. For example, if there is an increase in the number of patrol units in a given area, then the amount of reported crime increases. Note that the statement suggests a relationship between two variables, patrol units (X) and reported crime (Y), in a manner

indicating the belief that more patrol cars will cause more crime to be reported. Therefore, the research would focus on examining this relationship to disprove the null hypothesis. The null hypothesis fits the equation, “X has no relationship with Y.” For the previous research hypothesis, the null hypothesis would read, “The increase in patrol units will not increase the amount of reported crime (no relationship exists).” If the study results support the research hypothesis, then the null hypothesis can be rejected, indicating that there is a relationship between the variables. The rival hypothesis predicts the opposite of the research hypothesis. Using the previous example, the rival hypothesis would be if there is an increase in the number of patrol units in a given area, then the amount of reported crime decreases. An explanation for the rival hypothesis is helpful to consider in designing the research. Although the general research goal is to support the research hypothesis by disproving the null hypothesis, one should not consider the inability to disprove the null hypothesis as a failure. On the contrary, the failure to support the null hypothesis may actually lead to new information or additional research not previously conducted.

Whether the hypothesis is research, null, or rival, it must be clearly stated and consist of readily identifiable variables. Continuing from the Dantzker study, the independent variable is the individual psychologist and associated demographics. The main element is how the respondent labeled himself or herself as a psychologist. This research uses categories that best represent the potential psychologist involved in police psychological services: Police Psychologist, full time in-house psychologist; Psychologist Consultant, full-time consultant to law enforcement; Clinical Psychologist, occasional service provider to law enforcement; and Other, to allow respondents the opportunity for self-identification into unforeseen categories.

FROM THE REAL WORLD The study by Dantzker from the previous Real World feature offered the following hypotheses: Hypothesis One: Police psychologists will not differ significantly from general clinical psychologists in their choices of instruments or protocols used for preemployment screening of

police officer candidates. Hypothesis Two: Police psychologists will not differ significantly from general clinical psychologists in their reasons for selecting their choices of instruments or protocols used for preemployment screening of police officer candidates.

The dependent variables include the types of protocols and the reasons for their use. The protocol choices were garnered from previous research. The choices or reasons were developed from the literature, existing guidelines, and state legislation. Ultimately, the researcher must identify the variables properly, because misidentification can cause the research to be useless.

Summary Before starting any research effort, a topic must be chosen, keeping in mind that the research effort can explore, describe, or explain the topic. This topic should be of interest, be relevant, and have support in the literature through journals, government documents, and/or online resources.

On choosing the topic, the research question is created, which advises others what is to be studied. From the research question comes the hypothesis or hypotheses, a statement or group of statements that indicate the nature of the relationship to be studied. The three main types of hypotheses are research, null, and rival. The goal of research is to find support for the research hypothesis and disprove the null hypothesis. Finally, for the hypothesis or hypotheses, clearly identifiable variables are required, which include the independent and dependent variable. Failure to clarify the variables could render the research useless.

APPLICATION EXERCISES 1. You recently took a job with the research unit of a police department in a city of 50,000 residents. Within the city there is considerable variation of crime rates between geographic areas. In an effort to combat the problem, the police chief wants to know why some areas have higher rates of crime. You want to impress your new boss by developing a list of three possible research problems related to the topic of crime rates and geographic areas.

2.

3.

4.

5.

6.

Choose the idea that you think would be the most useful and use it as the basis for the proposed research. There are three purposes of research. Explain these purposes and how they may apply to the proposed research related to the topic of crime rates and geographic areas. Because you are new to the department, the police chief wants to know how you will prepare yourself to better understand the concept. Describe what a literature review is and the sources that are available for such a review. Based off your chosen research problem, develop three research questions to help bring the study into focus. Describe the different types of hypotheses. Develop three different research hypotheses and the null and alternative hypothesis associated with each research hypothesis. Based on the three research hypotheses, develop possible variables for the proposed study.

RESEARCH EXERCISES 1. Get in the habit of watching the local and national news as well as reading newspapers. These sources of information are typically easy to access online. In addition to keeping you informed of important issues of the day, these habits will provide a plethora of research ideas (don’t forget to keep adding to the notecards from Chapter 1). 2. Become familiar with the literature review sources that are publicly available as well as sources available to you through your college or university. Make a list of the sources as well as how to access them, including website addresses and necessary passwords. Even after this course is completed, hold on to and add to the list throughout your college career. It will save you time in the long run because you will not need to start from scratch every time you need to write a paper. 3. Complete a brief literature review based on your proposed research

problem, research question, and hypotheses.

CHAPTER 4: The Vocabulary of Research What You Should Know! The process of conducting and understanding research may seem to some like a foreign language. There are words, formulas, and so forth that may have more than one meaning or application. For example, the term research is often used in more than one context. It is quite common for students and teachers alike to use this term to describe a paper assignment that is actually a summary of existing literature. In contrast, the term research as used in this textbook refers to the process of conducting a scientific investigation of a specific phenomenon. Conducting research comes with a vocabulary that one must understand before one can proceed with the process of research. After completing this chapter, the reader should be able to: 1. Define theory and explain to the role that theory plays in research. 2. Describe the conceptualization process.

3. Describe what takes place during operationalization. 4. Define the terms variable, dependent variable, and independent variable. 5. Describe what a hypothesis is and how it differs from an assumption. 6. Present and discuss the types of hypotheses, including research, null, and rival. 7. Identify a population and discuss how it is related to a sample. 8. Define validity and reliability. Explain how these concepts are related. 9. Describe qualitative and quantitative research. 10. Discuss the steps in the research process, including theory, hypotheses, observations, and empirical generalizations.

The Language of Research It is quite common for students and teachers alike to use the term research to describe a paper assignment that is actually a literature review. With respect to criminal justice and criminology, there is more to research than reviewing literature. The multiple uses of the term research is just one example of the need to understand the vocabulary associated with research methods. In Chapter 1, the term research was defined. In this chapter,

terms associated with research methods, such as theory, hypothesis, and variable, are defined and expanded. The process of developing research can appear to be complicated with a lot of moving parts. However, the basics are simple and can be visualized in the form of a circle with four main parts: (1) theory, (2) hypotheses, (3) observations, and (4) empirical generalizations (Healy, 2013). Healy (2013) referred to this circle as the “Wheel of Science.” Explanatory research typically begins with theory and then moves through the remaining four parts.

Theory Theory is a statement or groups of statements that attempt to explain, predict, or understand a phenomena. An interesting debate could arise regarding the role of theory in research, reminiscent of the age-old argument: Which came first, the chicken or the egg? With respect to theory, one side of the debate argues that theories drive the research (theory-then-research), with a basis in deductive logic. Research that is based on deductive logic begins with a theory. From the theory, hypotheses are developed and observations are collected to test the hypotheses. Empirical generalizations are developed from the results of

the hypothesis testing and either support or do not support the theoretical basis. Keep in mind that results that do not support the theory do not necessarily indicate that the theory is wrong. The value of a theory is often determined based on many tests of hypotheses based on the theory over years or even decades and not based on a single research study. The other side of the debate argues that research creates the theory ( research-then-theory), with a basis in inductive logic. Research that is based on inductive logic begins with observations from which empirical generalizations are developed that are used to create theory or theories to explain the observations (Adler & Clark, 2007; Bryman, 2008; Colton & Covert, 2007; Creswell, 2008; Healy, 2013). Deductive logic and inductive logic are both based on the Wheel of Science but with different starting points. Theory is the starting point for deductive logic, and observation is the starting point for inductive logic. A theory is essentially a statement that attempts to make sense of reality. Reality consists of those phenomena that one can identify, recognize, and observe. For example, criminal behavior is observed. Therefore, people breaking the law are a

reality. A question that arises from this reality is what causes people to break the law? It is here that theory becomes useful because it can explain an aspect of reality in general rather than just for individuals. Many criminological theories focus on causes of criminal behavior that include biological, psychological, and sociological factors. For examples of theories in criminology, see Box 4-1.

Box 4-1 Examples of Theories in Criminology Biological A person’s physique is correlated to the type of crime one commits. Criminality is genetic. A chemical imbalance in one’s brain can lead to criminal behavior. Psychological Criminal behavior is the result of an inadequately developed ego. Inadequate moral development during childhood leads to criminal behavior.

Criminals learn their behavior by modeling their behavior after other criminals. Sociological Socializing with criminals produces criminal behavior. Society’s labeling of an individual as deviant or criminal breeds criminality. Failure to reach societal goals through acceptable means leads to criminality.

Whether theories have any merit or are truly applicable is why research is necessary. Proving that a theory is valid is a common goal of criminological and criminal justice researchers. However, to research a theory, the first step is to narrow the focus onto a concept. A single research study will not be able to examine all aspects of all crimes; therefore, researchers must narrow their attention to study a focus that is manageable.

Conceptualization

A concept is best defined as an abstract label that represents an aspect of reality, such as an object, policy, issue, problem, or phenomena. Every discipline has its own concepts. In criminal justice and criminology, some concepts include crime, law, criminals, rehabilitation, and punishment. For example, criminals are individuals who have violated a law and may be punished within the criminal justice system. Punishments are the actions taken against someone who has violated a rule or law. Concepts are viewed as the beginning point for all research endeavors and are often broad in nature. They provide the basis for theories and serve as a means to communicate, introduce, classify, and build thoughts and ideas. To conduct research, the concept must first be taken from its conceptual or theoretical level to an observational level. In other words, a concept must go from the abstract to the concrete before research can occur. This process is often referred to as conceptualization. In most research, it is seldom specified just how the concept is moved from the conceptual level to the observational level. This lack of clarity can cause readers to have problems in understanding what is being researched and why. Therefore, it is often helpful when the researcher can offer readers a

clearer picture of the conceptualization process. Operationalization occurs after the conceptualization process.

Operationalization Operationalizing involves the act of describing of how a concept is measured. This process is best described as the conversion of the abstract idea or notion into a measurable item. Operationalization is the taking of something that is conceptual and making it observable, or going from abstract to concrete. A single concept can be operationalized into multiple measurable items. For example, continuing with the criminal example from the conceptualization section, the concept of criminal could be operationalized into a measurable item that describes the type of crime committed by the offender (e.g., property, violent, other) or the number of crimes committed. The concept of punishment could be operationalized into a measureable item that describes the type of sentence (incarceration, community supervision, or other) or the length of the sentence (e.g., 6 months, 2 years, life without the possibility of parole). Operationalization is one of the more important tasks before conducting any research. There is no one right way to approach this task. How this

process is accomplished is up to the researcher. Yet, it is common for researchers to publish their results without ever explaining how their concepts were operationalized. This shortcoming has made it difficult for many students to comprehend fully the notions of conceptualizing and operationalizing variables. Therefore, when research is conducted that focuses on these two terms, it can be quite useful.

FROM THE REAL WORLD Seldom do researchers report on their conceptualization process. However, when they do, it provides a better understanding of the research. From the following excerpt, can you determine the important concepts within the study? Community empowerment is a concept used to describe individuals living in close proximity who as a group unite to combat a common problem. The focus of the group is the common problem. If a community is to be empowered, the residents must first be aware that a problem exists (community awareness)

to such an extent that it is disturbing or troubling (community concern), resulting in organization of the community (community mobilization) to fight against it (community action). (Moriarty, 1999, p. 17)

The following excerpt shows how a concept is operationalized. Community awareness was conceptualized as the level of knowledge about the use of alcohol and other drugs in the community. Four variables reflected community awareness: (1) drug usage in the neighborhood; (2) drug dealing in the neighborhood; (3) alcohol and drug prevention messages; and (4) availability of certain drugs (eight different drugs in all). The following are the actual questions used to establish each variable (Moriarty, 1999, p. 18): Drug usage in the neighborhood: Respondents were asked, “How many people in this neighborhood use drugs?” Responses included “many, some, not many or no residents use drugs.”

Drug dealing in the neighborhood: Respondents were asked, “How often do you see drug dealing in this neighborhood?” The responses included “very often, sometimes, rarely, never.” Alcohol/drug prevention message: Respondents were asked if they had heard or seen any drug or alcohol prevention messages in the past six months. Availability of certain drugs: Respondents were asked about the difficulty or ease of obtaining specific drugs in the county. The list of drugs included marijuana, crack cocaine, other forms of cocaine, heroin, other narcotics (methadone, opium, codeine, and paregoric), tranquilizers, barbiturates, amphetamines, and LSD. Each drug availability represents one variable.

Variables The primary focus of the operationalization process is the creation of measurable items known as variables and subsequently developing a measurement instrument to assess those variables. Variables are concepts that may be divided into two or more categories or groupings of “attributes” or characteristics. For example, male and female are attributes of the variable gender. The primary purposes of developing these variables are to measure a phenomenon and also to examine the

relationship between two or more variables in explanatory research. When examining the relationship between variables, there are two types of variables involved: dependent and independent.

Dependent Variables A dependent variable is a factor that requires other factors to cause or influence change in it. The dependent variable is the outcome factor or the factor that is being predicted. If a researcher is studying the impact of parental supervision on juvenile delinquency, then juvenile delinquency is the dependent variable. It is the dependent variable because changes in juvenile delinquency are being caused by changes in parental supervision.

Independent Variables The independent variable is the influential or the predictor factor. In other words, these variables are predicted to cause the change or outcome of the dependent variable. Often used independent variables in the study of criminal behavior are demographic characteristics, such as age, gender, race, marital status, and education. When determining independent variables, it is important to keep in mind time order. Independent variables must come before dependent variables.

Identifying and recognizing the difference between the variables is important in research, but sometimes the details may get lost due to the amount of information within any one study. Therefore, when research specifically calls attention to the variables, it can be quite useful. Keep in mind that whether a variable is dependent or independent is defined by the researcher and not by the variable itself. Criminal behavior could be a dependent variable if the researcher is studying the impact of unemployment on theft behavior. On the other hand, criminal behavior could be an independent variable if the researcher is studying the impact of criminal behavior on the stability of marriage. The key to any research is to be able to operationalize the concepts into understandable and measurable variables. Failing to complete this task makes the creation and testing of the hypotheses more difficult.

Hypotheses Once the concept has been operationalized into variables based on the theoretical basis, then most research focuses on testing the stated hypotheses. A hypothesis is a specific statement describing the expected relationship between the independent and dependent variables. There are three common types of hypotheses: (1) research, (2) null, and (3) rival.

Research Hypothesis The foundation of a research project is the research hypothesis. This hypothesis is a statement of the expected relationship between the dependent and independent variables. The statement may be specified as either a positive or negative relationship. In a positive relationship, the dependent and independent variables are moving in the same directions. They are increasing, or they are both decreasing. In a negative relationship, the dependent and independent variables are moving in opposite directions. If the independent variable is increasing, then the dependent variable is decreasing. Or, if the independent variable is decreasing, then the dependent variable is increasing. The example of juvenile delinquency provides an example of a negative relationship, specifically an increase in parental supervision is predicted to decrease juvenile delinquency. This result may be expected because juveniles who are under increased supervision will have fewer opportunities to participate in delinquency.

Null Hypothesis Some researchers argue that the results of the research should support the research hypothesis, including the existence of a relationship as well as the direction of the relationship between the

dependent and independent variables. Others claim that the primary goal is to disprove the null hypothesis, which is a statement indicating that no relationship exists between the dependent and independent variables. Therefore, a null hypothesis may be that parental supervision does not impact juvenile delinquency. By rejecting the null hypotheses, the research goal has been fulfilled.

Rival Hypothesis Before starting the research it is customary to establish the research hypothesis, which are generally the results the researcher expects to find based on the theoretical basis. However, sometimes the results may reject both the null hypothesis and the research hypothesis. This situation allows for the creation of what is called a rival hypothesis. The rival hypothesis is a statement offering an alternate prediction for the research findings. Continuing with the example of parental supervision and juvenile delinquency, the rival hypothesis would be that as parental supervision increases, so does juvenile delinquency. This unexpected result may occur because juveniles who are under increased supervision may rebel and act out due to the stress of being constantly under supervision.

It is usually the goal of the research to be able to reject the null hypothesis. Testing the research hypothesis becomes central to the research, making identifying the hypothesis an important aspect of the research. Yet, although hypotheses often take center stage in research, there is another type of statement that can find its way into the research: assumptions. However, these types of statements should be avoided whenever possible.

Assumptions Hypotheses are educated guesses about the relationship between variables and must be proven by the research results. In contrast, an assumption is a statement accepted as true with little supporting evidence. From a research perspective, assumptions are problematic. It is expected that with statements of inquiry or fact that there be substantiating research. It seems generally inconsistent in scientific research to have assumptions (Banyard & Grayson, 2009; Creswell, 2008).

FROM THE REAL WORLD In examining the effect of participatory management on internal stress, overall job satisfaction, and turnover intention among federal probation officers, Lee, Joo, and

Johnson (2009) offered three hypotheses: (1) organizational variables are more important than individual variables in predicting an officer’s turnover intention; (2) among organizational variables, participatory climate, internal stress, and overall job satisfaction, respectively, have a significant direct effect on an officer’s turnover intention; and (3) participatory climate and internal stress also have a significant indirect effect on an officer’s turnover intention.

Assumptions can be defined as statements one accepts as being true with little or no supporting evidence, a stance inappropriate to scientific research (Gillham, 2009). However, it is difficult to conceptualize a piece of research without having some assumptions about the topic of the study. Some assumptions may be needed in the development of a working hypothesis or hypotheses, to help guide thought about the topic of interest, and to help shape the design of the study. It is good research practice to identify and acknowledge any a priori assumptions of which the researcher is aware while the research study is in the planning stages.

However, assumptions can often lead to research. For example, because of the believed natural caring instincts of women, an assumption might be made that women would make better police officers than men. Because there is little evidence to validate this assumption, and it is not a readily accepted statement, at least among men, there is a need to research this assumption. In this situation, the researcher could move beyond the untestable assumption that women would be better officers because they are more caring by converting the assumption into hypotheses that can be tested. Variables could be created to measure what is meant by caring and what is meant by officer performance. Theory, concept, operationalize, variable, hypothesis, and assumption are all key words in the language of research. Still, they are just the building blocks for other words with which one should be familiar.

Other Necessary Terms There are many other terms a student should be familiar with before undertaking a research effort. Because these remaining terms are covered in greater detail in later chapters, only a brief definition is offered here.

Unit of Analysis A unit of analysis is the level at which the researcher will focus his or her attention. It could be individuals, groups, or social artifacts depending on the nature of the research. If the proposed study is examining the influences of individual criminal behavior, then the level of analysis is the individual. However, if the proposed study is crime rates, then the unit of analysis would be the social artifacts, specifically crime reports at the state, county, or local levels.

Population A population is the complete group or class from which information is to be gathered. For example, if a researcher was studying policing in Atlanta, Georgia, then all Atlanta police officers would be the population. Although it would be great if every member of a population could provide the information sought, it is usually logistically impractical in that it is both inefficient and wasteful of the researcher’s time and resources. Therefore, most researchers choose to obtain a sample from the targeted population.

Sample A sample is a smaller subset chosen from within a target population to provide the information sought.

Choosing this group is referred to as sampling and may take one of several forms, including probability and nonprobability sampling. Sampling is important enough to warrant an entire chapter of its own later in the text. Some examples of sampling are:

Random: A random sample is one in which all members of a given population had the same chances of being selected. Furthermore, the selection of each member must be independent from the selection of any other members. Stratified Random: This sample is one that has been chosen from a population that has been divided into subgroups called strata. The sample is comprised of members representing each stratum. Cluster: Cluster refers to a multistage sample in which groups are randomly selected and then individuals within the groups are randomly selected. Snowball: This sample begins with a person or persons who provide names of other persons for the sample who in turn provide the names of additional persons.

Purposive: Individuals are chosen based on the researcher’s belief that they will provide the necessary information. Once the sample has been identified, the information is collected. The various collection techniques are covered in detail in a later chapter. In collecting this information two concerns for the researcher are the validity and the reliability of the data collection device.

Validity Validity is a term describing whether the measure used accurately represents the concept it is meant to measure. There are four types of validity: (1) face, (2) content, (3) construct, and (4) criterion. Face validity is the simplest form of validity and refers to whether the measuring device appears, on its face, to measure what the researcher wants to measure. Using a ruler to measure time would appear odd and not useful; however, using a ruler to measure distance appears to be valid. Face validity is primarily a judgmental decision. It is the most used form of validity, perhaps due to its simplicity. In content validity, a measuring device is examined to determine how well it covers the concept in question; specifically each element is examined within the measuring device. Only using income to

measure socioeconomic status is missing important elements, such as education and occupation measures; thus it is incomplete and lacks content validity. Construct validity refers to the fit between theoretical and operational definitions of the concept as well as the level of expected agreement between the measure and other variables. For example, a positive relationship might be expected between fear of crime and the likelihood of protective behaviors, such as not walking alone at night. Criterion validity represents the degree to which the measure relates to an external criterion. It can either be concurrent (comparing self-reports of property crime victimization to property crime reports) or predictive (the ability to accurately foretell future events or conditions, such as use of the Scholastic Aptitude Test (SAT) to predict performance in college).

Reliability Reliability refers to how consistent the measuring device would be over time. If the study is replicated, will the measuring device provide consistent results? The two key components of reliability are stability and consistency. Stability means the ability to retain accuracy and resist change. Consistency is the ability to yield similar results when replicated. Keep in mind that a measure can be reliable but not

valid. Each year as daylight savings time begins, many people will forget to reset their clocks. If for the next week, an individual does not reset his clock, it will sound consistently when it reads 7:00 a.m.; however, it will in actuality be 8:00 a.m. (you may be able to use this as an excuse for being late for work once, but beyond that would be pushing your luck).

Data Having established the validity and reliability of the measuring device, the sample can now be approached for information. The information gathered is known as data. Data are simply pieces of information gathered from the sample that describe events, beliefs, characteristics, people, or other phenomena. These data may be qualitative or quantitative.

Qualitative Versus Quantitative Research The debate over qualitative versus quantitative research simply comes down to a question of concepts as ideas or terms versus numerical values. Broadening this distinction in easy terms offers quantitative research that refers to counting and measuring items associated with the phenomena in question, whereas qualitative

research focuses on concepts and verbal descriptions (Given, 2008). In recent years there has been a trend among scholarly journals to demand more quantitative than qualitative research. This preference is because qualitative research is often criticized as not being “scientific” (Bergman, 2009; Creswell, 2008; Given, 2008). Both methods are appropriate and necessary to criminal justice and criminological research. However, either qualitative or quantitative research may be preferred depending on the goal or purpose of the research.

Qualitative Research Defined Qualitative research is defined as a nonnumerical explanation of one’s examination and interpretation of observations with the goal of identify meanings and patterns of relationships (Creswell, 2008; Hagan, 2006; Maxfield & Babbie, 2009). This type of research encompasses interpreting action or meanings through a researcher’s own words (Adler & Clark, 2007) rather than through numerical assignments. For example, saying someone is aggressive and confrontational is a qualitative observation, but saying that someone has been arrested five times for disorderly conduct is a quantitative observation. Such analysis enables

researchers to verbalize insights that quantifying of data does not permit. Continuing with this example, the qualitative observation could be expanded to include specific details related to the aggressive and confrontational behavior, such as alcohol or drug consumption prior to the incidents. It also allows one to avoid the trap of false precision, which frequently occurs when subjective numerical assignments are made. If the disorderly conduct offender were assessed by a counselor using an assessment device that included the item “How aggressive is the arrestee on a scale of one to five with one being the least aggressive and five being the most aggressive,” this example would be quantitative but is subjective based on the judgment of an individual. These quantifications are misleading in that they convey the impression of precision that does not really exist.

Merits and Limitations of Qualitative Research The insights gained from qualitative research and their usefulness in designing specific questions and analyses for individuals and groups make this form of research invaluable in the study of criminal justice and criminology. However, the costs and time involved in such studies may not be logistically feasible (Bergman, 2009; Drake & Jonson-Reid,

2008). One of the major complaints about qualitative research is that it takes too long to complete. Other complaints about qualitative research include that it requires clearer goals and cannot be statistically analyzed (Creswell, 2008; Given, 2008; Maxfield & Babbie, 2009). There may also be problems with reliability in that replication may prove quite difficult. Lastly, validity issues may arise from the inability to quantify the data.

Quantitative Research Unlike qualitative research, the definition of which has been controversial and often inconsistent, definitions of quantitative research are quite consistent: It provides a means of describing and explaining a phenomenon through a numerical system (Berg, 2008; Fowler, 2009; Maxfield & Babbie, 2009). In other words, quantitative research is not based on a possibly subjective interpretation of the observations but is a more objective analysis based on the numerical findings produced from observations. One could look at data often collected by correctional agencies as an example of quantitative research, such as the length of prison sentences (often measured in months or years), the number of prior convictions, or the number of days offenders stay out of trouble upon release. Keep in mind that subjectiveness can be interjected into

quantitative analysis, as in the previous example of the counselor using his or her judgment with an assessment device. Because of the potential for bias and the criticisms of qualitative research as being unscientific, most criminal justice and criminological research tends to be quantitative in nature. A quick review of the leading journals that publish criminal justice and criminology research supports this assertion. There are many issues that are not suitable for numerical assignments; to apply quantitative measurements would be inaccurate and misleading. Some of the more intense theoretical debates, such as the merits of the death penalty, are based on personal beliefs about human nature that are shaped by deepseated religious, political, and moral convictions. As a result, perceptions of these tend frequently to be more influenced by emotion and ideology than by scientific study. This reasoning does not mean that quantitative research cannot be conducted, only that it is difficult to do so.

The Research Process Having been introduced to research and its language, you can now see how the described process fits into the Wheel of Science model. This process begins with a theory, usually identifying

some concept. The concept is then conceptualized and operationalized creating variables, including dependent and independent variables. Completing the identification of both the independent and dependent variables leads to developing the hypothesis. Finally, a sample is chosen, information (data) is gathered from the sample, the information is converted into the proper data for analysis, and the results are reported. This process becomes functionally clearer as the text progresses.

What You Have Not Done Before Developing researchers may believe that they have done qualitative research through the literature reviews and comparative and historical research done for previous courses. If done in a systematic and logical manner consistent with the scientific method, perhaps this is true. But have numerical values been assigned, data collected using that assignment, and the results analyzed? It is doubtful, unless one has a strong background in math and science or has been fortunate enough to have been involved in an empirical research project (for those who are still struggling with methods phobia, fortunate is defined here as having benefited from the knowledge and insights of the experience rather than from the pleasure of the experience). This deficiency is one that the authors hope to aid in

correcting. A college graduate who values and is capable of critical thinking and independent learning needs to be able to conduct quantitative research. It will prove valuable not only in an academic career but in future work tasks or civic duties. It is surprising to find how many things there are in life that warrant “looking into.” In addition, quantitative research is actually easier than qualitative research.

Summary Becoming proficient in research requires knowing the language. Several terms have been introduced that are important to mastering research as a language. The main terms include theory, concept, conceptualization, operationalization, variables, hypothesis, and sample. There are two types of variables: independent and dependent. A sample is a subset of the selected population. Other terms are validity (face, content, construct, and criterion), reliability, and data. With knowledge of these terms, the research process can be taken to another level.

APPLICATION EXERCISES 1. Due to the recession of 2007–2009, many municipalities increased their cost-cutting measures. One city not only turned off a percentage of streetlights but actually removed the

streetlights themselves. You have been approached by concerned neighborhood associations and business associations to study the impact of removing the streetlights on the occurrence of crime, fear of crime, and businesses within the affected communities. The first step is to conceptualize the study terms (crime, fear of crime, and business growth) by determining the abstract concept for each term. Keep in mind that you may conceptualize these concepts differently from your classmates. 2. After completing the conceptualization step, you will need to operationalize the concepts crime, fear of crime, and business growth by developing at least one variable for each concept. An example using the concept of the change in the number of streetlights would be a variable that counts the increase or decrease in the number of streetlights within a defined period during the previous one-year period. Thus, a defined area in which two streetlights were added would be +2

and a defined area where three streetlights were removed would be -3. 3. Based on the variables you created during operationalization, what type of data might you collect in the proposed study for each variable? 4. Assuming that the change in the number of functional streetlights is the independent variable, develop three research hypotheses using each of your previously developed variables for crime, fear of crime, and business growth as the dependent variable within each research hypothesis. Explain why you are expecting this relationship. 5. For each research hypothesis, develop an associated null hypothesis and rival hypothesis.

RESEARCH EXERCISES 1. During class in groups of no more than three students, compare the information you developed from the Application Exercises. As a group, develop a single set of conceptualizations, operationalizations, and hypotheses.

Do not just choose one students’ work but combine and modify information from each group member’s information. 2. After developing a single set of conceptualizations, operationalizations, and hypotheses, develop a research proposal based on these ideas and present your proposal to the class. As a class discuss the similarities and differences between the proposals. 3. Of the notecards that you have developed thus far (see the Research Exercises from Chapter 1), choose the one that you find the most interesting and develop a proposal by developing conceptualizations, operationalizations, and hypotheses related to your chosen topic.

SECTION II: Procedures

CHAPTER 5: Sampling What You Should Know! Gathering data for a research study can be done in a variety of ways, which will be discussed in detail in a later chapter. When using the common method of a questionnaire, from whom the data are collected becomes an important consideration. Often data cannot be collected by every member of a target population due to logistical constraints, such as money and time. Therefore, it is necessary to select data from a subset of the population, also known as sampling. After completing this chapter, the reader should be able to: 1. Discuss the purpose of and need for sampling in research. 2. Define what is meant by the terms population, sampling frame, and sample, and provide examples of each term. Describe how these terms relate to each other. 3. Explain how probability theory enables the researcher to obtain representative samples.

4. Identify and describe the various types of probability samples, including simple random, stratified random, systematic, and cluster sampling. 5. Compare and contrast probability sampling with nonprobability sampling. 6. Identify and discuss the various types of nonprobability samples, including purposive, quota, snowball, and convenience sampling. 7. Explain the importance of sample size. Include confidence intervals, confidence levels, and sampling error in the discussion. 8. Determine how many observations are necessary to obtain a sample with an error tolerance of ±3 at the 95% confidence level. Explain how many more observations you would add to this number and why you would do so.

Sampling Conducting research requires gathering information about a specific concept, phenomenon, event, or group. Gathering information about every element associated with the topic in the social sciences is neither feasible nor necessary. In conducting criminal justice and criminological research, the primary focus is typically on a specific population.

A population is the complete group or class from which information is to be gathered. Examples of populations are criminal justice system representatives (e.g., police, corrections, courts) or offenders (e.g., arrestees, convicts, parolees). Populations are not limited to individuals. Prisons in the United States or counties within a state are other examples of populations for studies at the aggregate rather than individual level. Although it would be great if every member of a population could provide the information sought, it is just not practical. Therefore, sampling the population is a necessity. A sample is a smaller group subset chosen from within a target population. Before one can begin to sample, the first step is to identify the population. Once the population has been identified, then the next step is to acquire a sampling frame. A sampling frame is a list of all units within the population (Bachman & Schutt, 2008; Creswell, 2008; Maxfield & Babbie, 2009). For example, if your population consists of New York City police officers, then the sampling frame would be a list of all officers. Each officer is one unit. Having identified the sampling frame, the next decision is to choose the type of sample to be used. However, it is often the case that a full sampling frame is not available. A researcher may want to

examine the reasons that some individuals are working as prostitutes. However, there is not a list of working prostitutes as there would be of other populations, such as police officers. This scenario is not uncommon when dealing with offender populations in criminal justice and criminological research. The lack of a full sampling frame will influence the selection of the sampling method because some methods require a full (or nearly full) sampling frame and other methods do not. Sampling methods are typically subdivided into probability or nonprobability methods. It is appropriate at this point to provide an overview of probability theory.

Probability Theory A friend of yours frequently buys lottery tickets. On a number of occasions, individuals (usually whom he does not know) have taken it on themselves to inform him, “You are wasting your money. The odds of you being hit by lightning are higher than of you winning the lottery.” To which he honestly replies, “Thank you for your concern. I have been hit by lightning. This is more fun.” The individuals who are warning our friend are statistically correct: His chance of winning the lottery is extremely small. However, because he does not really care if he wins

(he amuses himself by checking his tickets several days later, asserting that he is potentially a winner until he discovers otherwise) and because his investment only averages about $2 per week, statistical probabilities do not mean much to him. Many people who wager far more than they can afford also disregard statistical probability. These individuals tend to believe either that “it is time for their luck to change” or that “God (or fate, depending on their religious orientation) will intervene in their lives.” Although we do not question the benevolence of a Supreme Being, we do believe that if God or even luck pre-ordained such an event, the purchase of one ticket would be adequate. If not preordained, the sincere gambler might want seriously to consider the statistical probability of success. Probability theory is based on the concept that over time there is a statistical order in which things occur. If someone flips an unaltered coin 10 times, it is possible that it will land on heads five times and tails five times. However, it also might land on heads eight times and tails only two. This occurs because each time the coin is flipped it has an equal chance of being heads or tails. What happened previously has no influence on what happens in the future. One

cannot accurately predict what will happen on the next coin toss. Yet, one can accurately assume that over a lengthy period of time the number of heads and the number of tails will be about the same. This occurrence is the basis of statistical probability. Anything can happen, but over the long run there is a statistical order.

FROM THE REAL WORLD Let us continue with Dantzker’s dissertation examining whether there were any differences between two types of psychologists who performed preemployment psychologic examinations of police recruits regarding the instruments they used to conduct the evaluations and the reason for their choice. Because there is no one source identifying all licensed psychologists in the United States, the sampling frame for this study included all clinical psychologists with memberships in the American Psychological Association (APA) and the Psychology Section of the International Association of Chiefs of Police.

The knowledge that over time things tend to adhere to a statistical order allows one to choose samples that are representative of a population in general.

Although a researcher cannot say in advance that a sample is representative, the researcher can follow a procedure that should lead to a representative sample being selected. Because every number (representing people, items, or events) has a known chance of being chosen, then most of the time the sample drawn is representative. However, on occasion it is not representative.

Probability Sampling The general goal when choosing a sample is to obtain one that is representative of the target population. Representation requires that every member in the population or the sampling frame have a known chance of being selected for the sample. If the sample is representative of the population, then the results from the sample can be applied or generalized to the whole population, which is also known as generalizability. Four types of probability samples include: (1) simple random, (2) stratified random, (3) systematic, and (4) cluster.

Simple Random Samples A simple random sample is one in which all members of a given population have an equal chance of being selected. Furthermore, the selection of each member must be independent from the selection of any other member. In other

words, the selection of each member does not impact the likelihood that another member of the sampling frame will be selected. To assist in selecting a random sample, a device known as a random numbers table is often used, which contains numbers that have been randomly generated and can be found in almost every statistics book as well as being readily available on the Internet. To use the random numbers tables, an individual selects a number at random as a starting point and then chooses a pattern, such as moving down or across the table, to select the subsequent random numbers provided within the table until the appropriate number of population members needed has been selected. Today, computers are commonly used to select a random sample. In either case, the researcher must have a complete list of every member of the sampling frame, which is one of the disadvantages of random sampling. Yet, even with this obstacle, random sampling is very popular, partly because it has become easier for researchers to obtain random samples with the use of computers.

Stratified Random Samples A stratified random sample is a random sample that has been chosen from a population subdivided into groups called strata. These strata are homogeneous

groups selected based on specified characteristics (Adler & Clark, 2007; Dunn, 2009; FrankfortNachmias & Leon-Guerrero, 2008). This type of sample requires the researcher to have knowledge of the sampling frame’s characteristics. These characteristics (selected variables) are then used to create the strata from which the sample is chosen. If you wanted to sample students at a high school, you may group them into strata based on their class (freshman, sophomore, junior, or senior) and then select a random sample from within each stratum. Depending on the interests or needs of the researcher, a proportionate stratified random sample may be selected in which the proportions of the strata are similar to the proportions in the population. For example, if freshman represented 20% of the population from the example above, then the sample stratum for freshman would represent 20% of the sample. Disproportionate stratified random sampling incorporates oversampling of specific stratum and may be used if one or more of the stratum is particularly small.

FROM THE REAL WORLD Examining the shadow of sexual assault hypothesis among college and university students across temporal situations and victim-offender relationships, Hilinski (2009)

used a simple random sample. The sample was of 375 undergraduate and graduate students enrolled at a medium-sized public university during the fall 2006 and spring 2007 semesters. She obtained her sample by sending an e-mail invitation to a random selection of 3,500 (approximately 25% of the university’s population) graduate and undergraduate students enrolled during the spring 2007 semester.

Systematic Samples There is some debate over this type of sampling. It has been discussed as both probability and nonprobability sampling. It is offered here as a type of probability sample because it includes random selection and initially allows inclusion of every member of the sampling frame. With a systematic sample, every nth item (e.g., 3rd, 5th, 100th) in the sampling frame is included in the sample. A warning should be given when using systematic sampling. When sampling an organization with a rank or hierarchical structure, one must be sure the selection procedure does not result in a rhythm consistent with the organization’s bureaucratic structure that would cause a particular type of individual to be selected each time. For example, if you are studying employee satisfaction and are

sampling a company with 24 employees in each unit and the 25th individual is a supervisor, then you would need to mix up the sampling frame so that if you choose the 25th person each time, you will not always be selecting a supervisor. This situation negates the purpose of such a sample. To begin systematic sampling, determine the nth unit by dividing the total number of units within the sampling frame by the number of units wanted in the sample, then make sure the sampling frame does not have an existing order, and finally choose a random starting place in the sampling frame.

FROM THE REAL WORLD Dantzker used a combination of sampling techniques for his study. He began with a stratified, systematic random sample drawn from the membership of the American Psychological Association Division 12 (Clinical Psychologists). After removing students, non–United States members, and retirees, a population of 3,609 members was left. A random sample of 1,000 was selected through stratification by state. Because 1,000 individuals was the sample goal, which was 28% of the population, approximately 28% of each state’s members were randomly chosen using a random numbers table.

Cluster Samples The last of the probability sampling methods is the cluster sample. Cluster sampling is a multistage random sample (sampling occurs two or more times) in which groups (clusters) are sampled initially and then individuals are sampled within each cluster (Adler & Clark, 2007). Clusters are groups within the population (e.g., police precincts, prisons). For example, a sampling frame of correctional institutions nationwide may be created, and then a random sample of correctional institutions is selected from this frame. Employee information from each of the sampled institutions might then be obtained. From these lists, a random sample of correctional employees from each institution may be drawn. This method is popular for national victimization studies or other national interest topics (e.g., politics), which may be begin with clusters of counties or census tracts. Usually, researchers want to make use of probability samples primarily because they are often more statistically stable and the results can be generalized from the sample to the population. However, random sampling can be expensive and logistically difficult to complete. Therefore, it is common to find nonprobability sampling in criminal justice and criminological research.

Nonprobability Sampling The major difference between probability and nonprobability sampling is that one provides the opportunity for all members of the sampling frame to be selected, whereas the other does not. This shortcoming of nonprobability sampling often leads to questions and concern over the representativeness of the sample. However, when the sample produces the requisite information, representativeness is often not as much of a concern, although its limitations must still be noted. Furthermore, a nonprobability sample could be perceived as representative if enough characteristics of the target population exist in the sample. There are four types of nonprobability samples: (1) purposive, (2) quota, (3) snowball, and (4) convenience.

Purposive Samples Among the nonprobability samples, the purposive sample seems to be the most popular. The sampling process is based on the researcher’s skill, judgment, and needs to select an appropriate sample (Dunn, 2009). When the subjects are selected based on the researcher’s view that they reflect normal or average scores, this process is sometimes referred to as typical-case sampling. For example, a researcher may select an “average”

student for a sample of high school students. If subgroups are sampled to permit comparisons among them, this technique is known as stratified purposeful sampling (Adler & Clark, 2007). A major factor of purposive sampling is accessibility to units or individuals that are part of the target population. In all of the studies from the preceding From the Real World, the researchers chose their samples because they believed they best fit the needs of the study. The selection was based on their knowledge of the topic, the target populations, and accessibility. Although the samples may not have been representative, they did provide the requisite data to complete the studies. In most cases, because the purposive sample is not representative, findings cannot be generalized to complete populations. Despite statistical concerns about the sample, purposive sampling can offer researchers a legitimate and acceptable means of collecting data.

FROM THE REAL WORLD Monto and Julka (2009), in framing prostitution as an economic exchange, evaluated some of the consequences of conceiving of sex as a commodity rather than as an aspect of an intimate interpersonal relationship among the

customers of prostitutes. Their data were collected from a purposive sample of 700 men arrested while trying to hire street prostitutes. DeMatteo, Marlowe, Festinger, and Arabia (2009) used a purposive sample of 28 individuals assigned to a drug treatment program to study what effect the program may have had on the participants’ continued drug use. To gather opinions of police officers with many years of experience responding to emergency calls and pursuing fleeing suspects, Schultz, Hudak, and Alpert (2009) used data collected from participants of In-Service Training in Emergency Vehicle Operations and Police Pursuits.

Quota Samples Sometimes a study simply needs data collected from a set amount of participants fitting the sampling needs. These types of research efforts often rely on quota sampling. For this type of sample, the proportions (or percentages) may be based on the researcher’s judgment for inclusion. Does the individual or unit fit the needs of the survey? Selection continues until enough individuals have been chosen to fill out the sample. For example, assume one is required to conduct a study of high

school drug use with a sample size of 100 firstsemester college freshman at a given university. Because there is no means of identifying these individuals, a booth is set up in the student union where students are stopped as they come by and inquiry is made as to their status at the university. Only those who advise that they are first-semester freshman students are surveyed, and this approach is continued until the desired sample size or quota is reached. To ensure some level of representativeness by gender and race, the quota is set at a percentage equivalent to what the university claims to have in its population. For example, by gender, the university is 45% female and 55% male, whereas racially it is 43% white, 37% African American, and 20% other. Reaching some minimal level of representativeness requires one to continue selecting students until the quota sample seems to be comprised in similar fashion to the university. Obviously, this can be a painstakingly long method and does not guarantee representation. One way to visualize this process is to use a table. Continuing with the example of high school drug use reported by first-semester college students, a table could be created using more than one factor based on the characteristics of the population. Based on the following table, 24% of the population consists

of white males; thus, once 24% of the total sample is white males, then no more white males would be sampled.

Race

Sex

White

African-American

Other

Male

24%

20%

11%

Female

19%

17%

9%

Snowball Samples Snowball sampling is commonly used as a qualitative technique. The snowball sample begins with a person or persons who provide names of other persons for the sample. This sample type is most often seen used in exploratory studies where an appropriate target population is not readily identifiable, making a sampling frame more difficult to select (Senese, 1997). Additionally, despite the issue of representation, snowball sampling requires the researcher to rely on the expertise of others to identify prospective units for the sample. Snowball sampling is frequently used in field research when the researcher must rely on introductions from group members to access other group members. Using the previous prostitution sample, you may interview one person working as a prostitute who in turn introduces you to others. The sample may

represent a small geographic area or the acquaintances of an individual and not be representative of the population of prostitutes in the metropolitan area. Thus snowball sampling may lead to a sample that has no representative or generalizable attributes. However, if the data address the research question, then these shortcomings are acceptable.

Convenience Sample The last choice for a sample is the convenience sample or a sample of available subjects. Here there is no attempt to ensure any type of representativeness. Usually this sample is an abstract representation of the population. Units or individuals are chosen simply because they were in the right place at the right time. Analysis of data from a convenience sample is extremely limited. The sample selected may or may not represent the population that is being studied. Therefore, generalizations that are made about the population cannot be considered to be valid (Adler & Clark, 2007). For example, if you wanted to research the study habits of college students, you may stand outside of the college library to collect the convenience sample. But keep in mind that students who are visiting the library may differ in

their study habits compared to the population of college students as a whole. Because of this limitation, convenience samples are not useful for explanation or even for description beyond the sample surveyed. They are often useful as explorations on which future research may be based. The quality and quantity of the data are dependent on the sampling technique. Statistical support is stronger for probability samples. However, there are times when nonprobability samples are fruitful. Regardless of which type is used, an important element of each is the sample size.

FROM THE REAL WORLD Flanyak (1999), conducting a qualitative, exploratory research study, used both snowball and convenience sampling methods. Through this method she obtained 22 subjects for the study. Although her sampling frame consisted of lists of sociology and criminal justice department faculty members in several universities in Illinois, it was the result of initial interviews during a pilot study with respondents at her academic institution that other potential subjects were identified. Furthermore, other subjects were

chosen from the faculty of her undergraduate institution to ensure a high response rate.

FROM THE REAL WORLD To study officer opinions on police misconduct, Hunter (1999) developed a survey instrument based on the findings of prior research on police ethics and misconduct. Before administering this survey to a sample of several hundred officers in the southeastern United States, he sought feedback from a convenience sampling of currently serving officers. One group consisted of officers on a patrol shift from a mid-sized metropolitan police agency. A second group was composed of officers taking college courses at a regional college. A third group was made up of officers in their final phase of training at a regional police academy. Findings from this convenience sample not only permitted Hunter to refine the questionnaire, but he was also able to gain insights as to how officers felt about different forms of police misconduct and a variety of proposed solutions aimed at curbing such behaviors. The results were only indicative of the opinions of the subjects within this sample and could not be

construed to be representative of the opinions of officers outside the sample. However, as an exploratory study it was quite useful.

Sample Size The quality of a sample is largely dependent on its size (Banyard & Grayson, 2009; Creswell, 2008; Frankfort-Nachmias & Leon-Guerrero, 2008). The belief is that the larger the sample, the more likely the data will more truly reflect the population. An interesting aspect about sample size seems to be that there is no ideal set size. Usually, the sample size is the result of several elements: (1) requirements for sample accuracy; (2) economic feasibility (how much does one have to spend); (3) the availability of requisite variables (including any subcategories); and (4) accessibility to the target population. Furthermore, confidence level is important.

Confidence Levels Deciding how large the sample should be requires an understanding of confidence intervals, which indicates a range of numbers (e.g., ±15). Because a sample is merely an estimated reflection of the target population, a confidence interval suggests the accuracy of the estimate; the smaller the confidence

interval, the more accurate the estimated sample. The estimated probability that a population parameter will fall within a given confidence interval is known as the confidence level (Adler & Clark, 2007). Levels of 0.05 (95%) and 0.01 (99%) are most readily acceptable in most research (Gillham, 2009; Shaughnessy, Zechmeister, & Zechmeister, 2008). To reduce sampling error, the researcher desires a smaller confidence interval. To do so, he or she selects a smaller confidence level. For example, Dantzker (2010) sought to meet an alpha or confidence level of 0.05, which required a sample size of 385 (the process of arriving at this number is explained below). In social science research, confidence levels are important, although they are rarely explained or identified. In many cases readers are expected to accept that samples are statistically acceptable as an accurate estimate of the target population. A confidence level outside the expected range does not mean that the findings should be ignored. It simply means that their application must be more conservative and judicious.

Sampling Formulas The key aspect to selecting appropriate confidence levels is the sample size. As previously suggested,

the larger the sample, the more accurate the estimate. Therefore, it is beneficial to know just how large a sample is needed to attain the best confidence level. There are several mathematical formulas to assist in determining sample size. Unless one is a good mathematician, it is suggested that preexisting tables or computer statistical packages be used to make that determination. Should those means not be available, the following paragraphs demonstrate how such a formula may be used.

A Commonly Used Sampling Formula In selecting a sample size, one is seeking to draw a large enough number of observations from the target population to ensure that the sample accurately represents that population. As was discussed previously, the larger the sample size, the more likely that it is representative. However, because the costs in time and money of sampling large numbers are prohibitive, probability theory is relied on to estimate the proper sample size. One is guided in the selection by knowledge of acceptable sample sizes. Generally, in social science research, one seeks a sample size that 95 times out of 100 varies by 5% or less from the population. In some cases one may use less stringent requirements and

in others one may wish to have (and be able to afford) a higher level of accuracy. Assuming that one wishes to have a sample of a large population that is at the 95% confidence level and has a sample error within 5% of the population, one could use the following formula: n = (1.96)2 [p (1 − p)]/se where n = sample size needed, p = assumed population variance, se = standard error, and 1.96 represents a normal curve z score value at a confidence level of 95%. For an error tolerance of 5% at the 95% confidence level, one would use 0.05 as the sample error. The formula then becomes: n = (1.96)2 [.5 (1 − 0.5)]/0.05 n = 384.16 which rounds up to n = 385 Thus, a sample size of 385 provides a sample that had an error tolerance of +5% at the 95% confidence level. If one wished to change

confidence levels or error tolerance, one would then adapt the formula. For example, if one wanted to ensure that the sample was representative and had the money and time to do so, an error tolerance of +1% at the 99% confidence level would result in the following: n = (4.2930175)2[0.1 (1 − 0.1)] / 01 n = 16586.99 or = 16,587

A Sampling Size Selection Chart Having read how the previous formula was used to obtain the desired sample size, you may decide that you have no desire to ever perform these calculations. To save you from such an exercise, a simple chart is included that makes sample size selection much easier. Simply determine the confidence level and error tolerance desired in the survey sample and look it up in Table 5-1. The number indicated is how many observations from the study population are needed to select randomly to have a representative sample. Table 5-1 provides the sample size needed based on the error tolerance and confidence level desired. However, those are the numbers of observations needed in the sample. To ensure that those numbers are obtained, it is recommended that one

always oversample by 20% because not everyone selected will agree to participate in the research study. If this is not enough, one can always add more observations as long as they are randomly selected from the same population and any time differences do not affect responses. TABLE 5-1 Sample Size Selection Chart

Error tolerance

Confidence levels

(percent)

95%

99%

1

9604

16,587

2

2401

4147

3

1068

1843

4

601

1037

5

385

664

Source: Modified from Cole, R. L. Introduction to Political Science and Policy Research. St. Martin’s Press, 1996.

Summary Gaining data from a complete population is usually impossible. In most cases, research data are best obtainable through a sample. Before a sample is

chosen, identifying the sampling frame is necessary. An identifiable sampling frame leads to a decision as to what type of sample to select: probability or nonprobability. Probability samples include simple random, stratified random, strata, and cluster sampling. Purposive, quota, snowball, and convenience are forms of nonprobability sampling. Regardless of the type of sampling, there is a question of sample size. Although no magical number exists, confidence levels provide a statistical means for establishing legitimacy of the sample. The smaller the confidence level, the more representative is the sample. As can be seen, careful thought should be given before choosing a sample type and a sample size.

APPLICATION EXERCISES 1. You are working for a large, private research firm that has been hired by the comprehensive University and College System (UCS) of a large state to study the issue of violence on college campuses. How would you explain to the UCS officials the difference between a sample and a population and why sampling would be appropriate for this study?

2. What is a sampling frame? What would be an appropriate sampling frame of individual students for this study? Describe how you would collect the data for sampling frame. 3. Describe probability sampling to the USC officials, including the advantages and disadvantages of probability sampling. 3a. Describe how you would collect a simple random sample. 3b. Describe how you would collect a stratified random sample. Describe at least two factors that you could use to identify strata. 3c. Describe how you would collect a systematic sample. 3d. The system includes 130 four-year universities, four-year colleges, twoyear community colleges, and two-year technical colleges. Describe how you would collect a cluster sample using two levels (institution and individual) and three levels (institution, major, and individual).

4. Describe nonprobability sampling to the USC officials, including the advantages and disadvantages of nonprobability sampling. 4a. Describe how you would collect a purpose sample, including the specific factors that would be used to select the “normal” or “average” student. 4b. Describe how you would collect a quota sample using three different factors. Set up a quota table (see text) using two of the factors. 4c. Describe how you would collect a snowball sample, including how you select your initial contacts. 4d. Describe how you would collect a convenience sample. Choose three locations on a college campus that you think would be best for completing a convenience sample and explain why these locations are appropriate. 5. Choose the sampling method that you think is most appropriate to complete this study and explain the reasons for your selection to UCS officials.

6. The USC officials want to know how the results might apply if less than half of those surveyed return completed questions. Explain the importance of sample size. Include confidence intervals, confidence levels, and sampling error in your discussion. 7. Determine how many questionnaires would be necessary to obtain a sample with an error tolerance of +3 at the 95 percent confidence level at a single college of 10,000 students.

RESEARCH EXERCISES 1. Using the search procedures you developed in Chapter 3, locate an article or government report that uses a probability sample. Locate another article or government report that uses a nonprobability sample. 2. Assess the effectiveness of each type of sample within the article and government report. Is the sampling method appropriate based on the goals and circumstances of the study? Explain why. 3. In the Chapter 3 Research Exercises, you chose your research idea that you

found the most interesting. If you were to develop a survey to acquire opinions based on your research idea, what would be the appropriate population for your study? For example, if we were to develop a survey to collect opinions on violence on college campuses, then college students would be one population and college administrators would be another population. Choose the sampling method that you think is most appropriate to complete your study and explain the reasons for your sampling method selection.

CHAPTER 6: Introduction to Research Design What You Should Know! A research design is essentially a plan that maps out the course of action for a research project. Before choosing a research design, it is important to recognize the goals of the study. For example, is the study attempting to determine the causes of homicide? All of the choices regarding research design are driven by the research question— specifically, what is the researcher trying to find or determine? After completing this chapter, the reader should be able to: 1. Define empiricism and discuss how it relates to criminal justice research. 2. Identify and discuss the three criteria for causality. 3. Contrast idiographic and nomothetic causal explanations. 4. Contrast necessary and sufficient cause. 5. Compare experimental and quasiexperimental research designs.

6. Discuss the three elements of experimental research design. 7. Describe the classical experiment research design. 8. Compare the classical experiment research design to the pretest-posttest, protest only, and factorial research designs. 9. Describe the four levels of measurement: nominal, ordinal, interval, and ratio. 10. Identify the questions that can be used to help correctly identify a variable’s level of measurement.

Empirical Observation Thinking back to an introductory course in criminology and criminal justice or a course in criminological theory, students may realize they have already received an introduction to empiricism. Cesare Lombroso, the founder of the Positive School of Criminology, used empiricism in his study of criminals. Lombroso’s study of criminals focused on observing and measuring physical characteristics, such as facial features. Earlier scholars also used empirical techniques. For example, Quetelet and Guerry compared the crime rates of geographic areas in France.

Empiricism is defined as seeking answers to questions through direct observation, specifically sensations and experiences, to arrive at conclusions about the world in which we live (Jeffery, 1990). The use of the scientific method, with its focus on structured inquiry rather than casual observation, is what makes empiricism important. This emphasis on trying to see things as they are rather than idealistically is the basis on which positive criminology is founded. Rather than just quoting an “eye for an eye,” empirical techniques can be used to gather and evaluate data as to the effectiveness of correctional programs, to decide which patrol strategy is more cost effective, or to determine whether extralegal factors influence conviction rates.

Causality In applying empirical observation to criminal justice research, the focus is on causal relationships. Simply stated, what behaviors or events lead to other behaviors or events? When trying to answer that question, one seeks to determine causality. One of the main questions in criminal justice and criminological research is what “causes” people to commit crime? By determining the causes of criminal behavior, the criminal justice system could develop interventions to decrease or prevent crime.

Idiographic and Nomothetic Causal Explanations The examination of numerous explanations to describe why a single event occurred is known as idiographic explanation. Historians tend to use this method to explain the occurrence of events, such as explaining why the terrorist attacks on September 11, 2001 occurred. Police officers investigating a homicide are searching for an idiographic explanation. They are trying to determine why a single homicide occurred and not trying to determine why all homicides occur. When researchers in criminal justice and criminology aim to determine why behaviors (such as homicide) or events occur in general, rather than to explain a single event, they are using nomothetic explanations (Berg, 2008; Bergman, 2009; Gavin, 2008). Rather than trying to provide a total picture of every influence in a causal relationship, nomothetic explanation focuses on one or a few factors that could provide a general understanding of the phenomena being studied. Nomothetic explanations of causality are based on probabilities. In other words, what factors may increase or decrease the likelihood of behaviors or events? For example, city officials who want to determine how to reduce

vandalism in areas of tourism determine that increasing surveillance, such as increasing the presence of police, decreases occurrences of new vandalism.

The Criteria for Causality In investigating to determine if there is a causal relationship between the events or issues that being studied, three criteria must be observed (Adler & Clark, 2007; Dunn, 2009): (1) time order—the independent variable (the variable that is providing the influence) must occur before the dependent variable (the variable that is being acted on); (2) association—a relationship between the independent and dependent variable must be observed; and (3) elimination—the apparent relationship is not explained by a third or external variable. For example, one sees individual A struck by person B. Person A then falls down. Using the criteria for causality, one would conclude that the striking (independent variable) led to the falling (dependent variable). It occurred before the falling and was clearly related to that which was witnessed. If no other event (a third variable, such as a second blow struck by person C) occurred, then it is reasonable

to assume that the criteria for causality have been met.

Necessary and Sufficient Cause In investigating causality, one must meet the previously mentioned criteria (time order, association, and elimination of the impact of an external variable), but there is no requirement to demonstrate a perfect association. In other words, every time changes occur in the independent variable, changes may not occur in the dependent variable because the relationship is based on probabilities or likelihoods. In probabilistic models, such as those used in most inferential research, one often finds exceptions to the rule. Necessary cause refers to a condition or event that must occur for another event to take place. For example, to collect a paycheck, one must be employed. The cause must be present for the effect to occur. When the presence of a condition ordinarily causes the effect to occur, this is known as a sufficient cause. The cause usually, but not always, creates the effect. Playing golf in a thunderstorm may not result in being struck by lightning every time (the authors still do not recommend it), but the conditions are sufficient for the lightning strike to occur. In social science research, it is preferred to identify a

necessary cause for an event but to do so is often impossible. Instead, one is more likely to identify causes that are sufficient.

Experimental Research Designs Although it is most often used in the natural sciences, research requiring a true experimental design occasionally may be attempted by social scientists. Though it is more likely for criminal justice and criminological researchers to use quasiexperimental research designs, it is important to understand experimental research designs. Typically, experimental research designs are used in assessing cause-and-effect relationships in which a specific group experiences something and this group is compared to another group that did not experience the something. The classical experiment is considered the gold standard of experimental research designs and includes all three major components of experiments: (1) random assignment, (2) experimental and control groups, and (3) pretesting and posttesting (Campbell & Stanley, 1963; Maxfield & Babbie, 2009). Experimental research designs that are modifications of the classical experiment include the pretest-posttest, posttest-only, and factorial experiment research designs.

In classical experiment research designs, the first step is to determine which subjects will be assigned to the experimental or control groups using random assignment. Random assignment is essentially flipping a coin to determine to which group each subject will be placed. This process should result in experimental and control groups that are roughly equivalent. Subjects in the experimental group are exposed to the experimental stimulus, which is based on the hypothesis’s independent variable (Campbell & Stanley, 1963; Maxfield & Babbie, 2009). For example, if the purpose of a study is to investigate the impact of an intervention designed to increase the empathy of violent offenders toward crime victims, then the hypothesis would be that exposure to the intervention would increase empathy of violent offenders toward crime victims. After violent offenders who were serving time in prison were selected to participate in the study, they would be randomly assigned to the experimental or control groups. The subjects in the experimental group participate in the empathy intervention, but the subjects in the control group do not. The inclusion of the control group allows for the elimination of the impact of external variables, which is one of the criteria for determining causality. The structure of the classical

experiment research design also addresses the time order element of the causal criteria. The experimental research designs can be used to address the issue of time order because one has the ability to control when the experimental group is exposed to the experimental stimulus. Additionally, one can control how much exposure occurs by determining how often the duration occurs and how long the subject is exposed to the experimental stimulus. Borrowing a term from medicine, this is known as “determining the dosage.” Both the experimental group and the control group complete pretesting and posttesting. The pretest and posttest are typically the same measure designed to assess changes in the dependent variable. Continuing with the example of the empathy intervention, the subjects (violent offenders) in the experimental group would complete the same survey designed to assess the level of empathetic attitudes before (pretest) and after (posttest) the intervention. The control group would also undergo pretesting and posttesting even though these subjects were not exposed to the experimental stimulus (empathy intervention). The inclusion of pretesting and posttesting in the experimental research design allows for assessment of the changes in the level of empathy

before and after the intervention within the experimental group and control group as well as comparing the impact of the intervention between the experimental and control groups.

Figure 6-1 Classical Experiment Research Design

In the classical experiment research design, one can see that all elements of the experimental research designs are present, including random assignment, control and experimental groups, and pretesting and posttesting. In Figure 6-1, O represents observations, and X represents the experimental stimulus. The experimental group is displayed on the first line with O X O, indicating an observation (pretest), experimental stimulus, and then the second observation (posttest). The control group is displayed on the second line with O O, indicating an observation (pretest) and then a second (posttest) but no experimental stimulus. The third line shows the time order, with t1 representing the first time period, t2 representing the second, and t3 representing the third.

The other experimental research designs—pretestposttest, posttest-only, and factorial experimental research designs—are variations of the classical experiment with elements either subtracted or added. The pretest-posttest experimental research design (Figure 6-2) differs from the classical experiment research design in that it does not contain a control group, which also negates the need for random assignment. This experimental design allows for the comparison between pretest (before exposure to the experimental stimulus) and posttest (after exposure) (Campbell & Stanley, 1963). Using the empathy intervention example, the pretest-posttest design would allow for comparisons between before and after the intervention. Therefore, one could not determine if the same changes would have occurred without the intervention since the results of a control group’s pretest and posttest are not available to compare to the results of the experimental group. Furthermore, the exclusion of the control group from the pretestposttest research design means that the presence of external variables cannot be eliminated in the analysis. For example, a victim of a violent crime may have been a guest speaker at the prison between the pretest and posttest, which could impact the results since the effect of this event may have had an impact on empathy in place of or in

addition to intervention itself.

Figure 6-2 Pretest-Posttest Research Design

Figure 6-3 Posttest-Only Research Design

The posttest-only experiment research design (Figure 6-3) differs from the classical experiment research design in that there is no pretest. Subjects are still randomly assigned to the experimental and control groups but only receive the posttest. In the posttest-only research design, comparisons can be made between the experimental and control groups, which allows for the elimination of the impact of external variables. However, the exclusion of the pretest does not allow assessment of the impact of the experimental stimulus, since a baseline observation is not available (Campbell & Stanley, 1963; Maxfield & Babbie, 2009). Researchers may argue for the use of the posttest-only research design if they perceive the pretest as having an impact on the behavior of the subjects and

convoluting the results. For example, subjects in the experimental and control group may consciously or subconsciously change their reported behavior on the posttest due to perceiving the desired results of the study from taking the pretest. However, the value of the pretest in being able to compare the results to the experimental and control group posttest results typically outweighs the concern of the unintended impacts of the pretest. Unlike the pretest-posttest and posttest-only experiment research designs, the factorial experiment research design (Figure 6-4) adds to the classical experiment research design. Rather than having one experimental stimulus, the factorial design has two or more experimental stimuli. The experimental stimuli may differ based on dosage (Campbell & Stanley, 1963; Maxfield & Babbie, 2009). For example, the subjects in the first experimental group (X1) for the empathy study may receive a single, hourlong exposure to the experimental stimulus (the intervention), and the subjects in the second experimental group (X2) may receive one two-hour session per week for 11 weeks. The type of experimental stimuli can also differ, such as one being a therapy session and another as simple as watching a video. Even though the factorial experiment research design can contain

an unlimited number of experimental stimuli, one should keep in mind what the primary goal of the research is and focus the experimental stimuli on the primary goal.

Figure 6-4 Factorial Experiment Research Design

The primary advantage of the experimental design is its ability to isolate the experimental variation and assess its impact over time. Individual experiments can be limited in scope and require little time, money, and number of subjects. In addition, it is often possible to replicate the results: Because an experiment is a controlled environment, the procedures can be replicated and the results compared. The major disadvantage is artificiality. Processes that occur in a controlled setting may not actually occur in the natural setting. Violent offenders randomly assigned to participate in the empathy intervention may not decide to participate on their own so the usefulness of the results are called into question.

With respect to criminal justice and criminology, this type of research is often expensive and logistically difficult to perform. One of the most difficult issues is obtaining consent when the research involves human test subjects. However, sometimes consent is easier to obtain if the experiment could prove useful to the subjects. For example, assume a new drug has been created that could suppress sexual desires. A group of convicted pedophiles are asked if they will participate in a study in which part of the group will receive the new drug, whereas the other half are given a placebo. After a certain number of weeks, both groups are tested for sexual response to certain stimuli, and the results are compared. Another major problem in conducting true experimental research is the difficulty in being able to maintain and control the environment in which the experiment is conducted. The environment in which criminal justice and criminology research is conducted is often far from stable and is filled with possible interfering variables. As a result of the control and consent issues, along with costs and other logistical problems, it is rare to see a criminologist conduct true experimental research. These difficulties have not stopped some efforts to conduct this form of research. Some examples found in criminal justice include the Kansas City

Preventive Patrol Experiment, the Minneapolis Domestic Violence study, and San Diego’s oneversus two-person patrol units.

FROM THE REAL WORLD For one year starting in 1972, the Kansas City Police Department implemented the Kansas City Preventive Patrol Experiment to determine the impact of police patrols on the occurrence of crime. In the study, three areas were selected, and different levels of routine patrols were implemented. In one area there were no routine patrols, and officers only answered service calls. In a second area, the levels of routine patrols were not changed, and this area acted as the control group. In a third area, the levels of routine patrols were more than doubled or even tripled. It was hypothesized that the higher levels of routine patrols would decrease the occurrence of crime. However, the results indicated little impact of the routine patrols on the occurrence of crime (Kelling, Pate, Dieckman, & Brown, 1974). Reality dictates that few experimental research projects are possible in criminal justice and criminology. The limitations, however, can be addressed to some degree

with a quasi-experimental design, which retains some, but not all, elements of the experimental research design.

Quasi-Experimental Research Design Unlike the true experimental design in which the researcher has almost complete control over relevant factors, the quasi-experimental design offers only some control. The quasi-experimental research design allows for the approximation of conditions similar to an experiment. However, this research design does not allow for the random assignment of subjects to experimental and control groups, nor does it include the ability to control or manipulate the experimental stimulus. Random assignment in particular is difficult to implement in criminal justice or criminological research due to ethical and legal concerns. Although easier to implement than the true experimental design, the quasi-experimental design has its difficulties, making it less appealing to most social scientists. The main difficulty lies in the interpretation of the results, specifically being able to separate the effects of a treatment from the effects caused by the initial inability to make comparisons between the average units in each

group. Since random assignment is not performed within quasi-experiments, the groups are not truly equivalent. Quasi-experimental research design is most commonly used in evaluation research to assess new approaches in the criminal justice system and to solve problems with direct applications to the criminal justice system. The empathy intervention was used as an example of an experimental research design. However, the assessment of the empathy intervention would be more likely to be completed using a quasiexperimental research design. Subjects choose whether to participate and then changes in the level of empathy before and after the empathy intervention are compared between subjects who participate and subjects who do not participate. One issue with this example is that the subjects who choose to participate may be different from those who do not participate. Specifically, they may be more likely to display empathy in the first place. Overall, there are a number of research designs from which to choose. The design chosen depends largely on what the researcher is seeking to discover, explain, or describe. Other considerations when choosing a research design include economics, logistics, and time. Ultimately, the

researcher must decide which design allows for the best results.

Quantitative Levels of Measurement Experimental and quasi-experimental research is typically quantitative. Quantitative data have variables with various levels of measurement, including nominal, ordinal, interval, and ratio. Based on the previously mentioned definition of quantitative research, measurement is viewed as the assignment of numerical values or categorical labels to a phenomenon for the purpose of analysis. Determining what level of measurement to use can often be a confusing part of conducting quantitative research. The four levels of measurement range from low (nominal is the simplest variable) to high (ratio is the most complex variable).

Nominal Level Variables The simplest level of measurement is the nominal level. At this level, measurement is categorical with no specified order within the categories of the variable. An example of a nominal level variable is month of birthday, with the categories of January through December. The categories are mutually exclusive. In other words, the categories do not overlap. A single item cannot fit into more than one

of the variable categories. For example, a variable describing vehicles with the categories two-door, four-door, truck, sedan, minivan, and SUV would have overlap between the categories, such as a two-door truck. Since there is overlap between the categories, they are not mutually exclusive. The way to fix this problem would be to specify separate categories; for example, the single vehicle variable would become two variables, the first being the number of doors and the second being the vehicle type.

Ordinal Level Variables The next level of measurement is the ordinal level. This level moves beyond being merely categorical by assigning a rank or a placement of order to the categories of a variable. As in the nominal level variables, the categories in an ordinal level variable are mutually exclusive. Although in using this level of measurement, numbers may be assigned for ranking purposes (e.g., on a scale of 1–10), these numbers are not scores, but are simply viewed as a demonstration of where the respondent believes the item to fall. For example, in looking at the difference between the seriousness of criminal offenses in which murder is labeled as a 9, robbery a 5, and theft a 1, there is a four-unit difference between each type, but one cannot explain what that

difference truly represents. Another example is that the types of prisons may be broken down into the categories of minimum, medium, and maximum. Most often ordinal measures are found in attitudinal surveys, perceptual surveys, quality of life studies, or service studies. For example, individuals could be given a list of occupations that respondents could be asked to rank in order of how stressful they perceive each to be, from most stressful (10) to least stressful (1). Although they would write any number between 1 and 10, the listed order would only tell what the individuals perceive. There is no way to determine how much difference there is in the perceived stressfulness between each occupation. A common type of ordinal level variable is a Likert scale, which consist of five categories. For example, students may be asked to respond to the statement “Research methods will play a vital role in my career in the field of criminal justice” by choosing one of five provided responses, including “strongly disagree,” “disagree,” “neutral,” “agree,” and “strongly agree.”

Interval Level Variables The third highest level of measure is interval, which has scores instead of categories. With scores, there is an expected equality in the distance between

choices on the continuum. The use of scores rather than categories allows the use of more sophisticated techniques during data analysis. There is no absolute zero or starting point for interval data. Because numbers assigned have an arbitrary beginning, the usefulness of the information may be limited. For example, the difference between an IQ of 135 and 150 is the same 15-unit difference as between 150 and 165. However, there is no distinction as to what the difference means. One can comment on the differences but cannot explain what that difference means. To be able to do so requires ratio level data. For practical purposes, interval level variables typically are treated as ratio level variables for the purpose of analysis.

Ratio Level Variables The highest level of measurement is ratio. This level is primarily characterized by an absolute beginning point of zero, and the differences between scores are equal and can be explained. Two of the more common ratio level variables are age and income. For an age variable, the respondent would provide their exact age, typically in years. For an income variable, the respondent would provide their exact income, usually for a single year. An example of the absolute zero relevance for the ratio-level variable

would be a subject that provides a 0 as a response to an income question because they were unemployed during the time period. With respect to research, ratio measures can be collapsed into ordinal measures, such as the individual scores of a ratio level income variable being collapsed into an ordinal level variable with the following categories: under $10,000, $10,000–$24,999, $25,000– $49,999, over $50,000. Many people, including students, find it frustrating to accurately determine a variable’s level of measurement. Sometimes, this situation is due to confusion on the part of the individual, but at other times the variable information is poorly worded. Asking the following question can help determine a variable’s level of measurement. First, are the variable’s possible outcome scores (numeric) or categories (named units or ranges of values)? If the variable’s possible outcomes are scores, then the variable is ratio. If the variable’s possible outcomes are categories, then the next question is: do the categories have an order? One way to check for order is to move a middle category to the top or bottom. If the categories look out of order or off, then the categories do have an order. If the categories have an order, then the variable is ordinal. If the categories do not have an order, then

the variable is nominal. These questions should help guide the determination of a variable’s level of measurement. The level that is chosen has an important impact on how the data are collected and analyzed. Researchers can move down from a higher level of data to a lower level during data analysis, but they cannot move up to a higher level from a lower level.

Summary This chapter provides an introduction to research design. Research design is based on empiricism with the goal of establishing causality (or more likely, association). The basis for research design is experimental research, including classical, pretestposttest, posttest-only, and factorial experiment research designs. Even though experiments are difficult to implement in criminal justice and criminological research, understanding the basics are important since it provides the basis of quasiexperiments as well as the cross-sectional and longitudinal research designs (discussed in the next chapter). Experimental research is largely quantitative; thus, identifying the variable level of measurement, including nominal, ordinal, interval, and ratio, is important.

APPLICATION EXERCISES 1. In your position within the research staff of a large federal prison, you have been assigned to study the impact of educational and vocational programs on the recidivism of inmates. Provide the definitions for the key components of the study, including educational programs, vocational programs, and recidivism. 2. The warden of the prison would like to know if educational programs and/or vocational programs “cause” a decrease in recidivism. Describe to the warden the criteria for establishing causality. 3. Describe experimental research design and quasi-experimental research design to the warden. 4. Identify the elements of the classical experiment research design and describe how you would implement this research design. a. Identify the difference between the classical experiment research design and the pretestposttest research design, and describe how you implement the

pretest-posttest research design. b. Identify the difference between the classical experiment research design and the posttest-only research design, and describe how you implement the posttest-only research design. c. Identify the difference between the classical experiment research design and the factorial research design, and describe how you implement the factorial research design with educational programs as one of the experimental stimuli and vocational programs as the other. 5. Describe how you would differentiate between the following levels of measurement: nominal, ordinal, and ratio. 6. Identify and define the independent variable(s) and the dependent variable for the study.

7. For each of the variables created in step 6, identify the level of measurement.

RESEARCH EXERCISES 1. Using the search procedures you developed in Chapter 3, locate an article or government report that uses an experimental research design. 2. Determine the type of experimental research design within the article and government report. Assess the effectiveness of experimental research design. 3. In the Chapter 3 Research Exercises, you chose the research idea that you found the most interesting. Develop a hypothesis that could be tested using an experimental research design. Identify the type of experimental research design you have chosen and explain how you would implement the research design. Create the independent and dependent variables and identify each variable’s level of measurement.

CHAPTER 7: Research Designs What You Should Know! A conscious effort is required to conduct research and to use the most appropriate design. A multitude of methods are available to conduct research, and the method chosen by any researcher may be reflective of a style he or she is most comfortable with, believes will provide the best results, or selects based on the research question. This chapter examines the various designs from which researchers can choose. After completing this chapter, the reader should be able to: 1. Discuss the issues to consider in selecting a research design. 2. Describe how a historical research design is conducted. 3. Explain how a descriptive research design is conducted. 4. Compare and contrast cross-sectional and longitudinal research designs.

5. Discuss the various types of longitudinal research design, including trend, cohort, and panel. 6. Describe how a case study research design is conducted. 7. Describe correlational and causal– comparative studies. 8. Discuss evaluation research.

Research Designs To complete any type of research successfully, it is important to establish a feasible plan or blueprint, known as a research design. This plan primarily responds to the common five Ws (who, what, where, when, and why) and H (how) of investigation. Because criminal justice and criminological researchers have a choice of research designs, it is important to be able to match the design properly to the desired outcomes. In selecting a research design, a number of issues should be considered. Creating an outline is recommended to ensure that all relevant issues have been considered. Box 7-1 is an example of such an outline.

Box 7-1 Issues to Consider in Selecting a Research Design Purpose of research

Identify the purpose of the research project, which should be clearly indicative of what will be studied. Prior research

Review similar or relevant research. This review promotes knowledge of the literature. Theoretical orientation

Describe the theoretical framework on which the research is based. Concept definition

List the various concepts that have been developed and clarify their meanings. Research hypotheses

Develop the various hypotheses that will be evaluated in the research. Unit of analysis

Describe the particular objects, individuals, or entities that are being studied as elements of the population. Data collection techniques

Determine how the data are to be collected, who will collect it, who will be studied, and how will it be done? Sampling procedures

Specify sample type, sample size, and procedures to be used. Instruments used

Describe the nature of the measurement instrument or data collection device to be used. Analytic techniques

Determine how the data will be processed and examined. What specific statistical procedures will be used? Time frame

Pinpoint the period of time covered by the study. This will include the time period examined by research questions and the amount of time spent in preparation, data collection, data analysis, and presentation. Ethical issues

Address any concerns as to the potential harm that might occur to participants. Also deal with any potential biases or conflicts of interest that could affect the study.

FROM THE REAL WORLD The Bureau of Justice Statistics produced the report Homicide Trends in the United States, 1980–2008. This report describes homicide patterns and trends based on the Federal Bureau of Investigation’s Supplementary Homicide Report. Information is provided on victim and offender demographic characteristics (age, race, gender), victim/offender relationships (intimate, family, infants, elders), circumstances (felony, sex related, drug related, gang related, argument, workplace), weapon involvement (gun, arson, poison), and homicides with multiple victims and/or offenders.

Descriptive Research Design A descriptive research design focuses on the description of facts and characteristics of a given

population, issue, policy, or any given area of interest in a systematic and accurate manner. The descriptive research design focuses on answering the “what” question in research design, but it cannot provide an answer to the question of “why” a phenomenon is occurring. The information obtained in descriptive studies can provide insights not recognized in prior research and provide the basis for additional studies focused on explanation. For example, studies reporting police patrol trends and crime rates may provide the basis for a study to examine the impact of police patrol patterns on crime rates.

Historical Research Design The historical design is the study of actions, events, and phenomena that have already occurred (Berg, 2008; Creswell, 2008; Givens, 2008). This design allows the researcher to systematically and objectively reconstruct the past through the collection, evaluation, verification, and synthesis of existing documenting information to test a hypothesis. It can assist in determining why or how an event occurred and whether such an event could happen again. A historical research design is also a means by which researchers may compare and contrast events or phenomenon that have occurred.

There are many sources of existing information, including text-based information, visual and auditory information, and secondary data. Text-based existing information may include news reports, public records, or even private diaries. Often one thinks of text-based information as the basis of historical research designs; however, pictures or maps as well as audio and visual recordings can provide useful information. A researcher may use crime rate maps to examine the impact of the opening or closing of businesses on crime over time. Or, photographs and audiovisual clips could be used to examine the impact of propaganda on the treatment of disadvantaged groups during times of conflict. Another source of existing information is secondary data, which is the data that has already been collected for another purpose. Secondary data is used extensively in criminal justice and criminological research. Researchers may collect data for their own studies and then make that data available through repositories, such as the Interuniversity Consortium for Political and Social Research (ICPSR). Government agencies also collect a considerable amount of information for management and evaluation purposes. Recidivism, the recurrence of criminal behavior, is an important

issue in criminal justice research and can be measured using records from correctional and court agencies. One of the most debated topics in criminal justice and criminology is the deterrent effect of capital punishment. A common hypothesis for capital punishment supporters is that the death penalty serves as a better deterrent against committing homicides than life imprisonment. To study this hypothesis a historical research design is appropriate and, for example, could include a study of homicide rates in the United States between 1950 and 1997. Although just studying the numbers might provide an interesting conclusion, the true historical research design study requires inclusion of possible influencing factors, such as U.S. Supreme Court decisions, sentencing patterns, social episodes (e.g., a war), and population growth. A successful historical study considers all relevant information to provide proper conclusions. The historical design is an economically efficient means for conducting research. Considering the vast array of records available related to criminal justice, there is no shortage of possible research topics. A shortcoming of this design is the difficulty of expanding beyond what is documented, therefore

limiting the scope of the research. Researchers are limited to the information in the files and seldom have means to follow up or get clarification of the available information. In addition, there is the old computer maxim of “garbage in garbage out.” The research is only as good as the data that is contained in the records. If there is inaccurate information or missing data in the original source material, the research suffers.

FROM THE REAL WORLD To determine the impact on recidivism of matching youth with services consistent with the clinical recommendations that they had received, Vieira, Skilling, and PetersonBadali (2009) conducted a study in which data were collected and compiled from a variety of secondary data sources. These sources included clinical charts; participants’ court reports; each youth’s scores on a risk and need measure (Youth Level of Service/Case Management Inventory); probation notes; parent and mental health service reports; and school records, probation files, and court records. The researchers found that having only a few of the recommended services was associated with an increase in the number of new

convictions and a decrease in the amount of time until recidivating.

Both descriptive and historical research designs can be cost-effective and logistically easier to conduct than other designs. However, they present researchers with limitations as to what variables can be examined and the extent of the information available. They may also be more ‘time sensitive’, which means that data may only be available for a certain time frame and the information obtained may be limited in its usefulness.

Cross-Sectional Research Design The primary concept of the cross-sectional research design is that it allows for the study of a phenomenon at one point in time. These studies are best suited for exploratory or descriptive research but can be used in explanatory research by attempting to study the association between variables. Cross-sectional studies provide a snapshot of the phenomenon and do so by examining the relationship between variables at one point in time. For example, Violanti et al. (2009) examined whether suicide ideation, planning suicide, and suicide attempts were more likely to occur among police officers because of their exposure to suicides. For their study they did a

cross-sectional study involving 115 randomly selected police officers from a mid-sized urban police department of 930 officers. One drawback to the cross-sectional study design is that it is difficult (or even impossible) to determine time order of the independent and dependent variables, which is an issue addressed in longitudinal research design.

Longitudinal (or Time Series) Research Design Perhaps a police agency is interested in following the activities of members from a rookie police class from graduation through their first 5 years of service. They may be primarily interested in the turnover rate, promotions, occurrence of injuries, accommodations and complaints, and levels of job satisfaction. The best research design for this type of study is longitudinal, or time series. This type of research design allows for the investigation of specifically identified patterns and events, growth, or change by collecting data at two or more distinctive time periods. It also allows for determining the time order of the independent and dependent variable, which the cross-sectional research design struggles to accomplish, and is therefore particularly suited for explanatory research as well as exploratory and descriptive research. This type of research can be costly and time-consuming; thus cross-sectional

research design is used more often. Examples of longitudinal research designs include trend studies, cohort studies, and panel studies.

FROM THE REAL WORLD In a study of binge drinking among college students, Wechsler et al. (1994) surveyed a total of 17,592 students at 140 colleges in the United States regarding binge drinking and the consequences of the reported drinking behavior at one point in time, such as being hungover, doing something that you regretted, missing classes, forgetting what happened, getting behind in school work, arguing with friends, engaging in unplanned sexual activity, not using protection during sex, getting injured, damaging property, getting into legal trouble, and needing medical treatment for an alcohol overdose. Binge drinking was defined as having four or more drinks in a row for women and five or more drinks in a row for men. According to the study, almost half of the students (44 percent) reported at least one instance of binge drinking within the last two weeks.

Trend Research Design

Trend research design examines changes in a general population over time. For example, data is collected for the census every 10 years from all individuals residing in the United States. One might compare results from several census studies to determine what demographic changes have occurred in that population, such as employment or income changes. Another example is the General Social Survey, which annually surveys a sample representative of the adult population in the United States. Surveys might indicate that opinions on the death penalty fluctuate over time depending on social conditions not related to crime or changes in the incidences of murder or the occasional occurrences of sensational murders.

Cohort Research Design Cohort research design involves studies that focus on the changes occurring in specific subpopulations over time, such as age groupings. This research design is highly flexible and could be used to examine social, political, or economic changes or have a more specific focus. For example, a cohort research design may be used to examine the oneyear recidivism rates of inmates released from a state facility during 2010.

In a famous criminological study, Wolfgang, Figlio, and Sellin (1972) examined delinquency among a cohort of juveniles, specifically males born in 1945, who were living in Philadelphia, Pennsylvania, on their 10th birthday. The researchers were particularly interested in the age that delinquent behavior began as well as when delinquent behavior stopped for this cohort. The findings from this study significantly impacted future research and practice in juvenile justice. In addition to primary data, many cohort studies use secondary data, making it cost effective.

Panel Research Design Panel research design studies the same set of people at two or more distinct points in time. Unlike the cohort studies, these subjects may not have a single unifying characteristic, such as an age cohort. By using the same individuals, couples, groups, and so forth, researchers are able to more precisely examine the extent of changes and the events that influenced them. However, due to attrition because of the effects of deaths, movements from the area, refusal to continue as subjects, and other factors that cause the sample to lose members, these studies are logistically difficult to continue over an extended period of time. For example, the National

Crime Victimization Survey interviews a member of a selected household every six months.

FROM THE REAL WORLD Hunter and Wood (1994) were interested in the relationship between severity of sanction and unarmed assaults on police officers. They obtained assault data on officers for all 50 states. They compared these data with the sanctions applied for unarmed assaults on police officers within each respective state during 1991. They then compared the rates in states that had felony sanctions for weaponless assault on police officers to their neighboring states from 1977 through 1991. This strategy resulted in the longitudinal analysis of four groups of states. Analysis of results in these groupings did not reveal support for the hypothesis that more sanctions would decrease the incidence of weaponless assaults on police officers.

Case Study Research Design The case study research design allows for the intensive study of a given issue, policy, or community in its social context at one point in time even though that period may span months or even years (Adler & Clark, 2007). It also may be used to

study specific individuals or groups. The case study research design includes close scrutiny of the background, current status, and relationships or interactions of the topic under study. Case study research design often focuses on a specific phenomenon, such as community policing.

FROM THE REAL WORLD The National Longitudinal Survey of Youth 1997 (NLSY97) is a panel study of a nationally representative sample of youth who were 12 to 16 years old as of December 31, 1996. The first NLSY97 survey of the youth and the youth’s parent took place in 1997, and the panel members continue to be surveyed on an annual basis, with the most recent survey occurring in 2013. The study is focused on entry into the workplace but contains a considerable number of criminal justice variables.

Case studies may be longitudinal in that they sometimes observe a case or cases over a certain length of time. These observations are closely linked with the observing of potential independent variables that may be associated with changes in the dependent variables. There are three basic features to this design (Dunn, 2009; Hagan, 2006;

Maxfield & Babbie, 2009): (1) qualitative or quantitative descriptions of a variable over the extended period of time, (2) a context wherein the researcher can observe changes in the variables, and (3) the potential for developing measurement instruments and the testing of their reliability over time. Case study research designs are not limited as to what can be studied and are particularly useful in exploratory research. However, they can be costly and time prohibitive and may not provide an explanation for the results. If one wants to know why something is or has occurred and possible correlating factors, then a correlational design may be more appropriate.

Determining Correlations and Causations Research often has the goal of determining that there is a causal relationship between two variables (causal–comparative research) or at the very least an association (correlational research). The selected research design will determine the ability to determine causation or correlation. Most studies are trying to determine correlation between two variables.

FROM THE REAL WORLD Chappell’s (2009) abstract states the following: Community policing is the operating philosophy of the majority of American police departments in the new millennium. Though most departments claim to engage in community policing, research has shown that implementation of the strategy is uneven. One way to investigate the implementation of community policing is to study patrol officer attitudes toward community policing because prior research has shown that attitudes are related to behavior. The present study used qualitative data in a case study research design to explore the extent to which patrol officers have endorsed and implemented community policing in one medium-sized agency in Florida. Furthermore, the research sought to gain insight into the organizational barriers that prevented officers from adopting community policing in their daily work. Results indicated that although most officers agreed with the philosophy of community policing, significant barriers, such as lack of resources, prevented its full implementation in this agency. Implications of the findings

and directions for future research are discussed.(p. 5)

Correlational Research A popular research design is one that allows researchers to investigate how one factor may affect or influence another factor, or how the one factor correlates with another—the correlational design. In particular, this type of design focuses on how variations of one variable correspond with variations of other variables. An example is a study of the level of education of police officers and promotion rates, arrest rates, or job satisfaction. The goal of this research design is to obtain correlational coefficients that are statistically significant (discussed in a later chapter). For example, a crosssectional study examining the association between drug use and academic performance is a correlational research design, but its design does not allow for the determination of a causal relationship.

Causal–Comparative Research Why do men rape? Why do teenagers turn to gangs? Why do individuals become serial killers? To answer these types of questions, a causal– comparative (or ex post facto) design is useful. This design allows the researcher to examine

relationships from a cause-and-effect perspective, which is done through the observation of an existing outcome or consequence and searching back through the data for plausible causal factors. The criteria to determine causality is often not met outside of the experimental research designs. Of the previously discussed causal criteria, being able to account for the impact of an external variable is particularly problematic. Thus, causal–comparative research is less common than correlational research.

FROM THE REAL WORLD Adding to the literature on whether gang membership is uniquely related to victimization experiences for females compared to males, Gover, Jennings, and Tewksbury (2009) produced a correlational study in which they examined the relationship between gender, gang membership, and three types of victimization. They used data from the 1999 South Carolina Youth Risk Behavior Survey, an ongoing state and national survey conducted by state contracts for the Centers for Disease Control and Prevention, Division on Adolescent and School Health. Their results indicate gang membership is

significantly related to the risk of victimization regardless of gender.

FROM THE REAL WORLD Stevens’s (1999) study of the relationship between drug addiction and criminal activity among incarcerated women who were in a prison drug rehabilitation program could be viewed as a causal–comparative design. This study addressed the belief that drug addiction gave rise to criminality among females. Yet the results of the data did not support this belief. One implication of this finding was that drug addiction in itself is not necessarily a causal factor for producing crimes of violence, especially among females.

Evaluation Research To this point, almost all the research designs discussed are used after an event, situation, or other unexplained phenomenon occurs and one wants to understand it better. These designs are quite useful to academic researchers as well as practitioners. A practitioner may want to know how something will work or what might occur when something not previously done is attempted.

Evaluation research is particularly important if managers or administrators want to make sweeping policy implications. To gain support for these types of changes, the managers or administrators would have to demonstrate that the changes would have the desired impact. Evaluation research can assist in the development of new skills or approaches. It also aids in the solving of problems with direct implications for the “real world.” This type of research typically has a quasi-experimental research design. For example, a police agency is debating whether to add a nonlethal weapon, a stun gun, to what is available to its officers. To see how stun guns might be used and what outcomes might result from their use, an evaluation research design would use a select group of officers who are issued stun guns for a set period of time. Each time the weapon is used, a report explaining the reason and results must be filed. At the end of the data collection period, the reports are analyzed; depending on the results, stun guns would be issued to all officers or certain officers, or the recommendation might be not to issue these weapons at all. Evaluation research is important for the criminal justice practitioner– researcher because it assesses

the merits of programs or policies being used (or under consideration for use) in the field. Whereas more basic research seeks to develop theoretical insights and much applied research seeks to determine if a theory can actually be applied in the field, evaluation research allows practitioners to determine the costs and effectiveness of the program or project that is being or has been implemented. For this reason, evaluation research that studies existing programs is frequently referred to as program evaluation.

FROM THE REAL WORLD To investigate the effectiveness of Thinking for a Change, a widely used cognitive behavioral curriculum for offenders, Lowenkamp, Hubbard, Makarios, and Latessa (2009) evaluated the impact of the program using a group of felony offenders placed on probation. The experimental group consisted of probationers who were referred directly to the program from court, and the comparison group included probationers who were not referred to the program. The results of the study found that the offenders who participated in the program had a significantly lower recidivism rate than similar offenders who had not been in the program.

Summary Selecting a topic and creating the research question are just the beginning of conducting research. One of the most important steps is choosing an appropriate research design. Although one of the most popular designs is survey research, which is also a means of collecting data in the other designs, other possible methods include: Descriptive—describing the facts and characteristics of a given population, issue, policy, or any given area of interest in a systematic and accurate manner. Historical—studying actions, events, and phenomena that have already occurred to systematically and objectively reconstruct the past to test a hypothesis. Cross-sectional—studying the issue of phenomenon at one point in time. Longitudinal—investigating patterns and sequences of growth or change as a function of time, including trend, cohort, and panel studies.

Case study—studying intensively the background, current status, and environmental interactions of a given social unit, such as individual, group, institution, or community. The researcher must clearly understand what compromises exist in the internal and external validity of his or her design and proceed within these limitations. The chosen design should best meet the needs of the research goals.

APPLICATION EXERCISES In a city that bases a large portion of its annual operating budget on tourism, there has been an increase in pickpocketing, purse snatching, and armed robberies. The local police chief has hired your company to determine the extent of these problems, examine their progression, and identify possible solutions. 1. Describe the descriptive research design to the police chief, and explain how this research design could contribute to the study. 2. Describe the historical research design and explain its uses.

3. Compare and contrast cross-sectional and longitudinal research designs. 4. Describe the cross-sectional and different types of longitudinal research designs, including trend, cohort, and panel, and explain how these research designs could contribute to the study. 5. Describe the purposes of correlational and causal–comparative research. 6. Propose a possible program or strategy that could be used to address the problem, and specify how evaluation research would be used to determine the effectiveness of the program or strategy in addressing the problem.

RESEARCH EXERCISES 1. Using the search procedures you developed in Chapter 3, locate an article or government report that uses a cross-sectional or longitudinal research design. 2. Determine the type of cross-sectional or longitudinal research design within the article and government report. Assess the effectiveness of research

design in answering the research question. 3. In the Chapter 3 Research Exercises, you chose the research idea that you found the most interesting. Develop a hypothesis that could be tested using longitudinal research design. Identify the type of longitudinal research design you have chosen and explain how you would implement the research design. Specify the independent and dependent variables.

CHAPTER 8: Qualitative and Quantitative Research Designs What You Should Know! A long-standing debate among academic researchers is over what type of research is more acceptable, qualitative or quantitative. The debate about which form of research is more appropriate in criminal justice tends to pit theorists against applied practitioners. Interestingly, as the debate continues, one could argue that both have a place in criminal justice. Therefore, students should have enough knowledge to differentiate between the two types of research. After completing this chapter, the reader should be able to: 1. Compare and contrast quantitative and qualitative research. 2. Describe the types of field interviews, including structured, semi-structured, and unstructured interviews, and provide examples.

3. Describe focus groups and their uses. 4. Describe the different roles that may be used in field observation, including full participant, participant researcher, researcher participant, and complete researcher. 5. Explain the differences between qualitative and quantitative field observation. 6. Describe ethnographic research. 7. Describe sociometry. 8. Describe the types of survey research, including personal interviews, mail questionnaires, and telephone surveys. 9. Describe mixed methods research. Because of its long-standing image as an applied social science, some criminal justice research conducted and published has had detractors. This is primarily the result of what is perceived as a lack of statistical sophistication. This perception is central to the continuing debate over what is more “academic,” qualitative or quantitative research. This chapter examines research designs that are closely tied specifically to qualitative and quantitative research designs. The selection of a research design for a study should be based on the research questions and goals of the study.

Qualitative Research Design

“Qualitative researchers … are most interested in how humans arrange themselves and their settings and how inhabitants of these settings make sense of their surroundings through symbols, rituals, social structures, social roles, and so forth” (Berg, 2008, p. 7). Conducting qualitative research may be time consuming, but because it better reflects the actual being of something, the time factor is accepted as a given for the qualitative researcher. Therefore, the bigger issue becomes what method to use. This decision depends on the goals of the research and the source(s) of the information to be sought. When information is to be gathered directly from individuals, the most common method to use is field interviews.

Field Interviews Interviewing involves one individual asking questions of another to obtain information. If the interview consists of specific questions for which designated responses may be chosen and numerical values assigned to the responses, then these qualify as quantitative research questions. For example, “Are you employed (Yes or No)?” is an example of a quantitative research question. Field interviews consist of open-ended questions. Openended questions do not have designated responses, and interviewees provide answers in their own

words. During the interview, the response given by the interviewee is recorded exactly as stated rather than assigned to a predetermined category. There is no one “right” way to conduct a field interview. However, asking the right questions is vitally important and more complicated than it may appear at face value. The questions included in the interview would be based on the type of interview being conducted and the information sought. Interviews may be (1) structured, (2) semistructured, or (3) unstructured (Dunn, 2009). The type of interview is based on the level of control exerted by the interviewer, with structured being the most controlled and unstructured being the least controlled.

Structured Interviews A structured interview entails asking every respondent preestablished open-ended questions. Responses are recorded as given, and the interview pace is such that all the questions can be asked and answered in a timely fashion, but neither the interviewer nor the respondent feel rushed. Even though the structured interview has open-ended questions, it is not similar to a conversation because the interviewer is strictly asking questions and does

not deviate from the script or respond to the interviewee’s answers. We recommend the following list of “do nots” as guidelines for conducting a structured interview: 1. Do not get involved in long explanations of the research. 2. Do not deviate from the study introduction, sequence of questions, or question wording. 3. Do not let another person interrupt the interview. 4. Do not respond for the person being questioned, suggest, agree to, or disagree to, an answer, interpret the meaning of a question, or improvise. It is natural human behavior to want to engage beyond a predetermined script, but by doing so, the benefit of the structured interview is lost. The structured interview is geared toward limiting errors and ensuring a consistency of order in the responses even though the responses themselves may vary. This type of field interview also allows the use of multiple interviewers who are collecting the same information in the same manner, which the semi-structured and unstructured interviews are not able to guarantee based on their designs. On the

other hand, even though the structured interview is designed to elicit rational, legitimate responses, it does not consider the emotional aspect. In conducting research in which emotional aspects play a role, a semi-structured and unstructured interview format might be more useful.

Semi-Structured Interviews Semi-structured interviews are similar to structured interviews except that the interviewer has more freedom to go beyond the responses for a broader understanding of the answers. This follow-up to questions is known as “probing for more detail.” Probing may consist of asking for more explanation of an answer than has been given or following up with an additional question or questions depending on the answers given. For example, a study of fear of crime may include the question, “What increases your fear of crime in the neighborhood?” If the respondent answers “gang activity,” then in a semistructured interview, the interviewer could ask a probing (or follow-up) question, such as “What is it about the gang activity in your neighborhood that makes you afraid?” or “How does the gang activity in your neighborhood change your behavior?” In a structured interview, the interviewer cannot ask these questions unless they are already in the script. There will be times when a follow-up question

is not readily apparent until a response is received. Semi-structured interviews allow the interviewer to follow up and collect additional information that would not have been collected with the structured interview.

FROM THE REAL WORLD Dantzker and McCoy (2006) used a structured interview to obtain information regarding psychological screening of police candidates from the 17 largest municipal police agencies and the Department of Public Safety in Texas. Once contact was made with the appropriate person, two basic questions of the structured interview were asked and the responses were recorded: (1) What psychologic protocol(s) are used to screen police applicants, and (2) what is the general content of the psychologic interview?

FROM THE REAL WORLD Charles (2009) examined spirituality and law enforcement using semi-structured interviewing, designed to explore the expression of spirituality as revealed by officers in their work. This particular method of narrative inquiry allowed the participants to “tell their stories.” The author asked the

officers eight standardized, open-ended questions with the ability for the authors to ask additional probing questions. 1. When did you become a law enforcement officer and why? 2. Tell me your spiritual history starting with your parents. 3. Describe your spiritual practice. 4. Tell me about your work as an officer. 5. Has your spirituality influenced your work as an officer? 6. What has been most challenging to you while working in a profession where you are constantly exposed to human destructiveness and suffering? 7. How have you changed as an officer? How has your spirituality helped? 8. How do you cope with the human destructiveness and suffering encountered in police work? What is your support system when you are overwhelmed?

Unstructured Interviews The unstructured interview is far less rigid than either the structured or semi-structured interview methods. Often the interviewer may have topics that

will be addressed and maybe some questions to start the interview process, but generally openended questions are created as the interaction proceeds (Adler & Clark, 2007). This style of interview is particularly useful when used in conjunction with participant observation or within an ethnographic study (both ideas will be discussed in subsequent sections). For example, to gain a better understanding of what police detectives find stressful, a researcher chose to spend several days with detectives from a particular police agency. Although the primary role was to observe how detectives responded to certain situations or events during “down times” (e.g., between calls, during meals, after the shift), the researcher was able to ask open-ended questions pertaining to what the detectives found stressful. One such question might be “What do you find stressful about being a detective?” Because of the nature of an unstructured interview, the researcher must be able to complete several tasks for the study to be successful (Adler & Clark, 2007; Creswell, 2008; Dunn, 2009; Given, 2008; Maxfield & Babbie, 2009): 1. Gain access to the setting.

2. Understand the language and culture of respondents. 3. Decide how to present oneself. 4. Locate a contact or informant. 5. Gain the respondent’s trust. 6. Establish rapport. By meeting these requirements, the researcher should be successful in his or her efforts. Often times, more than one interviewer is involved in a project, and personality characteristics can dictate if an individual is appropriate as an interviewer. When acting as unstructured interviewers, researchers must be able to think quickly on their feet and assess the information provided to develop the next question. Several problems may arise that could affect the outcomes of structured, semi-structured, and unstructured interviews. One possible error is the respondent’s behavior; for example, is the respondent being truthful or only saying what he or she believes the interviewer wants to hear? Another factor is the setting of the interview. For example, the interaction dynamic is different between face-toface and telephone interviews in that the telephone interviews are less personal. A third factor is question wording. If the questions are confusing or

uncommon terms are used, then this factor may decrease the amount as well as the accuracy of information provided. For most people, performing interviews is not intuitive and requires training, which can include directions to clearly enunciate or not change the wording of questions. Finally, if an interviewer is not familiar with the respondent’s background, culture, education, or other factors, this unfamiliarity can be detrimental to the interview.

Focus Groups Perhaps the best way to define a focus group is the interviewing of several individuals in one setting. Although not meant to replace individual interviews, focus groups have long been used in marketing and politics (Given, 2008). Movie studios may use focus groups to determine the final editing of a movie or the marketing strategy. Focus groups are convened during presidential debates to assess reactions to candidates’ statements. The focus group is an information-gathering method in which the researcher-interviewer directs the interaction and inquiry. This process can occur in either a structured (e.g., pretesting a questionnaire) or unstructured (e.g., brainstorming) manner. In either case the researcher–i nterviewer must meet the same guidelines offered for conducting any interview.

The use of focus groups has advantages and disadvantages. The advantages include limited expense and flexibility. The disadvantages include group culture, dominant responder, and topic sensitivity (Adler & Clark, 2007; Creswell, 2008; Dunn, 2009; Given, 2008; Maxfield & Babbie, 2009). The focus group can be a useful qualitative method for gathering interesting information.

Field Observation Field research typically entails interviewing or observation or a combination of both. The method of observation allows researchers to actually perceive the phenomenon under study using their own senses, such as sight, sound, and smell, rather than relying on subjects to relate their perceptions. It is more likely to be the method for a qualitative study but may also be used within a quantitative study (discussed in a subsequent section). For example, if the purpose of a study is to determine social interactions on the street, then a researcher could actually observe these interactions firsthand. Although this method of qualitative research has not received the same attention as interviewing (Adler & Clark, 2007; Dunn, 2009), it is still a viable tool, especially in criminal justice and criminology. Qualitative field observation aims to describe concepts and ideas rather than developing a

numeric assessment, which would be quantitative field observation (this topic will be discussed in a subsequent section). The role of the observer in field observation differs based on the amount of interaction the observer has with the environment and subjects as well as the level of deception employed regarding the observer’s identity and purpose. There are various methods for categorizing observing methods, but one that takes into account the level of both interaction and deception includes four types: (1) full participant, (2) participant researcher, (3) researcher participant, and (4) the complete researcher (Senese, 1997). Keep in mind that these broad categories actually include a range of behaviors within these constraints.

The Full Participant Sometimes it may be fitting or perhaps even essential that a researcher become part of the study. The full participant method allows the researcher to carry out observational research, but does so in a “covert” or “undercover” manner. This particular method is both high on the level of interaction as well as deception. For example, to study the workings of a gang, the researcher takes an active role as a gang member, but the other

gang members are unaware that the individual is actually observing for the sake of research. The full participant method increases the likelihood of acquiring accurate information because subjects are not changing in response to knowing they are being watched. However, the full participant method is impacted by ethical and moral dilemmas that could adversely affect the research if the wrong decision is made, such as having to compromise one’s beliefs or even place a researcher in legal jeopardy (e.g., being present during the commission of a crime). To avoid ethical and moral dilemmas (discussed in Chapter 2), a researcher may decide to adjust the method to eliminate the use of deception; thus, his or her identity is known, and the researcher become a participant researcher.

The Participant Researcher Participating in the activities of research subjects can offer insights and information not attainable through other forms of research and may also avoid ethical dilemmas facing full participation. As a result, the participant research method may be a preferable option. This method involves the researcher participating in the activities of the research environment with the full knowledge of the research subjects. For example, to study the behavior of males in a reformatory, the researcher

goes in and participates in all activities, but everyone knows he is there collecting information. The biggest negative of this research method is the possibility that subjects’ behavior may be influenced by their knowledge of the researcher’s role. The subjects may not act as they would under unobserved conditions, either over-exaggerating or under-exaggerating their actions. This modified subject behavior is known as “subject reactivity.” If you are filming children playing (with the permission of their parents of course), you will observe firsthand subject reactivity. Once they realize they are being filmed, some children hide from the camera and others show off and mug for the camera. On a serious note, by being part of the activities, the researcher can influence or even create outcomes and behaviors that may not have existed without his or her presence. Using the example of observing within a male reformatory, the inmates may be better behaved because they know they are being observed.

Researcher Participant (The Researcher Who Participates) Rather than taking part in activities, the “researcher who participates” method requires nothing more than observation by the researcher whose status as

a researcher is known to the research subjects. For example, to study whether a particular treatment offered to incarcerated juveniles is working, the researcher enters the environment where his or her status as a researcher is known, but does nothing more than observe. Although this method eliminates many of the problems of the previous two methods, the mere presence of the researcher who participates can still influence behavior and activities, but may do so to a lesser degree. The participant researcher and researcher participant can be viewed as being on a continuum with the identity being known for both, but with the participant researcher having more interaction with the environment and subjects than the researcher participant. The first word (participant or researcher) keys into the top priority of the observer.

The Complete Researcher The way for a researcher to minimize the problems generally associated with observation research is to avoid all possible interaction with research subjects. Data collection may involve covert methods of observation, such as a disguised vantage point. Within a prison, a guard tower in a maximum level prison could provide this disguised vantage point. Cameras provide another disguised vantage point and are fairly typical in our public life, and we are

often unaware or unconcerned that we are being filmed. The benefits that are gained by covert observation or assumption of a noninteracting role are countered by possible resentment and denial of access by those being observed, who may feel that they are being spied on and seek to protect their privacy. One drawback for the complete observer is the inability to interact with the environment or subjects to learn more about what the observer is seeing. Regardless of which form is used, field observation can be time consuming and may not provide the results ultimately sought. Still, it can be a useful method for gathering data that may not be available through another method. Field observation can also provide context that other types of research designs are not able to provide.

Ethnographic Research Ethnographic research is the study of people and cultures for the purpose of providing a detailed and in-depth description. It typically involves a combination of field interviewing and field observation methods in varying degrees. Attributes of ethnographic research include exploration of the nature of particular social phenomena, a tendency to work primarily with unstructured data, the

investigation of a small number of cases, and data analysis involving explicit interpretation of the meanings and functions of human actions (Berg, 2008; Creswell, 2008). The product of such a study is in-depth descriptions (Given, 2008).

FROM THE REAL WORLD To study the implementation of community policing, Memory (1999) conducted research as the “researcher as participant.” His methodology took the form of ride-alongs, where he spent 180 hours riding with police officers from selected cities or counties. Departmental contact persons arranged the ride-alongs ahead of time. Because six law enforcement agencies had agreed to be involved in a larger overall research project, the researcher did not need to “gain access” but still had to comply fully with agency and officer directions to maintain access. Memory rode during both daylight and darkness in each jurisdiction but rode very little between 1:00 a.m. and 1:00 p.m. As a researcher as participant, he dressed in slacks, a dress shirt, and a sport coat. He observed everything that occurred concerning the officers with whom he rode. The data collection involved taking notes on the front

and back of data-collection forms. The researcher recorded observations, words of officers and citizens, and his own impressions and ideas.

Ethnography is labor intensive and not a widely used method of conducting research; however, it can provide insights not found through quantitative research. For example, the study of gang graffiti from a quantitative perspective might simply identify the differing types of symbols drawn and the number of each. However, an ethnographic study could provide what the symbols mean, why they have been drawn in specific places, and who is placing them. This information provides additional detail to the information collected in a quantitative approach. An interesting aspect of ethnographic study is the possibility for the researcher to examine social interactions, which leads to another form of qualitative research: sociometry.

Sociometry Sociometry is a technique by which the researcher measures social dynamics or relational structures within a specific environment (Berg, 2008; Creswell, 2008; Givens, 2008). Information can be gathered through interviews or by observation and indicates the subjects chosen for participation and

characteristics of those who do the choosing. The lunchroom of a middle school would provide an ideal setting for a study using sociometry. The study may examine the interactions between students, such as who sits together and how seats are chosen. With respect to criminal justice, a sociometric study might involve prosecutor–defense attorney relationships, where the researcher observes the interaction of prosecutors with defense attorneys noting how each treats and is treated by the other and how these interactions affect outcomes, such as plea bargains. This study might show whether there is a hierarchy among lawyers or among potential defendants (the accused).

FROM THE REAL WORLD Harpster, Adams, and Jarvis (2009) examined verbal indicators to analyze 911 homicide statements critically for predictive value in determining the caller’s innocence or guilt regarding the offense. They listened to 100 audio recordings and transcripts of 911 homicide telephone calls obtained from law enforcement departments throughout the United States formulating the linguistic attributes of these communications and identifying variables that they then analyzed for association with the likelihood of the

caller’s guilt or innocence regarding the offense of homicide.

As with observation and ethnography, the researcher’s presence, attitudes, and biases can influence the outcomes of sociometric research. Because the mere presence of the researcher can be problematic, a less obtrusive means of conducting qualitative research would be to use records, such as interviews or 911 calls.

Quantitative Research Design Quantitative research provides a means of describing and explaining a phenomenon through numerical findings (Berg, 2008; Fowler, 2009; Maxfield & Babbie, 2009). A variety of designs are available to criminologists for conducting quantitative research. Yet, despite all the possibilities, one form of research is both a research method and a research tool: the survey. Because of its dual nature, it is discussed here as a quantitative research design and later in Chapter 10 as a data collection device.

Survey Research The survey is one of the most popular research methods in criminal justice. The survey design is used when researchers are interested in the

experiences, attitudes, perceptions, or beliefs of individuals or when trying to determine the extent of a policy, procedure, or action among a specific group. Most often a researcher contacts a sample of individuals who are presumed to have participated in a particular event, who belong to a certain group, or who are part of a specific audience having experienced similar events. Of the identified group, certain questions pertaining to the topic under study are asked, either verbally or in written form. The solicited responses to the questions comprise the data used to test the research hypothesis. The three primary survey methods are (1) personal interview, (2) mail questionnaire, and (3) telephone survey.

Personal Interviews Personal interviews are surveys administered by face-to-face discussions between the researcher and the survey respondent. The researcher reads from a previously developed questionnaire to which responses are numerically assigned. Such interviews may in fact be nothing more than the reading of a questionnaire that could have been mailed out to respondents. Personal interviews permit the researcher to obtain not only responses to the questions asked but observations as to the respondents’ demeanor and nonverbal reactions. For example, in a study of criminal behavior, the

interviewer could assess the subjects’ level of discomfort using a scale of 1 (comfortable) to 10 (uncomfortable) to questions regarding recent theft behaviors. They are, however, more costly and time consuming and less safe than other strategies that do not bring researchers into physical contact with respondents.

Mail Questionnaires Mail questionnaires are survey instruments mailed to selected respondents to complete on their own rather than being directly interviewed by the researcher. They are much cheaper and safer to administer. In addition, they enable researchers to survey large numbers of people quickly and easily. During the last census, someone in your household received the questionnaire, completed it, and mailed it back. If a response is not received, then an inperson, residence visit is completed by a censustaker. Mail questionnaires may be administered in three ways. Typically, the questionnaire may be mailed with a request for it to be completed and returned in a self-addressed stamped (or metered) envelope to the researcher. Additionally, the researcher may drop off the questionnaire in a face-to-face contact with the respondent with the request that it be

mailed back or mail the questionnaire and advise that he or she will come by at a later time to retrieve it. Either of the latter two strategies greatly increases the time and cost involved in conducting the survey; however, the third method can be viewed as an attempt at intimidation and receive a negative response. Technological advances have impacted the mail questionnaire. If you have been shopping recently, then you may have received instructions on your receipt for filling out a survey by logging into a website in return for the possibility of winning a prize for participating, such as a gift certificate. A particularly attractive benefit to having subjects respond electronically is that the data are automatically entered into a computer.

Telephone Surveys Telephone surveys are popular because they are quick and easy to do, enabling the researcher to contact large numbers of people in an efficient manner. They are safe in that verbal abuse or disconnects are about the worst that researchers can incur from their dealings with respondents. Telephone surveys that use computer-generated random digit dialing to sample the population may be more efficient. If you have received a phone call and heard silence before someone came on the line, then the company is probably using random

digit dialing. A computer dials the numbers, and a human only joins a call when someone answers. This technology offers the added advantage of inputting data into a computer as the surveyor asks the questions. The biggest disadvantage is that people who are sick of telephone solicitors often refuse to participate or do not answer the phone.

Pros and Cons of Survey Research Two reasons why the survey design is popular are (1) survey research makes it feasible to use large samples and (2) this design uses standardized questionnaires, which from a measurement perspective offers strength to the data because the same question is being asked of all respondents, thus giving credence to the same response from a large number of respondents. Despite these positives, the survey design also has its share of negatives: (1) a standardized questionnaire is limited with respect to whether the questions are appropriate because it is designed for all respondents and not for a select group; (2) it can seldom account for the contexts in which respondents are thinking or acting; (3) survey research is inflexible in that it typically requires that no change occur throughout the research, so that a preliminary study may be needed to guide research

design; and (4) the survey tool is often subject to artificiality. Even with these negatives, the survey design remains one of the most popular methods for conducting research. Furthermore, it is an extremely popular tool for collecting data. The issues involved in conducting survey research are discussed in more detail in a subsequent chapter.

Quantitative Field Observation Previously, field research was discussed as a qualitative research method. That discussion indicated that such research may also be quantitative in nature. When field observations are made allowing for numerical assignments, this method involves quantified field research. If the local police department is observing intersections that have a high rate of accidents by counting the number of times drivers run the stop sign within the observation period, then this study is an example of quantified field research.

FROM THE REAL WORLD Standardized field sobriety tests (SFST) were developed during the 1980s to enable traffic officers to estimate accurately drug and alcohol impairment. These SFSTs

consist of Walk-and-Turn, One-Leg Stand, and Horizontal Gaze Nystagmus. Traffic officers in all 50 states have been trained to administer these tests to individuals suspected of impaired driving and to score numerically their performance on them. It has been claimed that the SFST test battery is valid for detection of low blood alcohol concentrations and that no other measures of observation offer greater validity for blood alcohol concentrations of 0.08% and higher (National Highway Traffic Safety Administration, 1999). To study the legitimacy of these claims, Burns and Dioquino (1997) set up a field research in Pinellas County, Florida. Eight sheriff’s deputies with years of experience in traffic control, each having made hundreds of driving-under-the-influence arrests and all having extensive training in driving-underthe-influence enforcement including certification in SFST, were selected to participate. A rigorous observation procedure was established in which both officers and observers assigned to each officer carefully recorded the SFST scoring and the actual outcome of measured blood alcohol concentrations. A total of 379 traffic stops

were evaluated. In 313 cases SFSTs were administered. Drivers refused to take breath tests in 57 of these cases. In 256 cases, a breath test was administered to verify the accuracy of the SFST. The traffic officers were found to have an accuracy level of 97% in their arrest decisions based on the use of SFSTs.

Even though the different types of research are presented individually in this text, they can be combined for a more effective and comprehensive study. The combining of qualitative and quantitative research methods is referred to as mixed methods. Sometimes, the combining of different types of qualitative research methods or different types of quantitative research methods is also referred to as mixed (Johnson, Onwuegbuzie, & Turner, 2007). These combinations can expand the focus and results of the study as well as providing context to the results.

Summary There is a continuing debate between criminal justice and criminology researchers as to what type of research is best, qualitative or quantitative. This chapter argues that they complement each other and have appropriate roles in related research. To

that end, this chapter explores both qualitative and quantitative research. Although there are many forms of qualitative research, the more popular include the following: field interviewing (structured, semi-structured, and unstructured), focus groups, field observation (full participant, participant researcher, researcher participant, and complete researcher), ethnographic research, and sociometry. Survey research, including personal interviews, mail questionnaires, and telephone surveys, are popular quantitative research designs.

APPLICATION EXERCISES The chief operating officer for a nationwide chain of superstores is interested in determining the amount of theft that is occurring and how it is occurring. This information will be used to train loss prevention staff. 1. Explain the differences between qualitative and quantitative research design to the chief operating officer, and describe the type of information that could be collected through qualitative and quantitative research designs. 2. Describe the different types of field interviewing, including structured,

3.

4. 5. 6.

semi-structured, and unstructured, and explain how the different types of interviews could be implemented to study loss prevention. Describe the different types of field observation, including full participant, participant researcher, researcher participant, and complete research, and explain how the different types of field observation could be implemented to study loss prevention. Describe ethnographic research. Describe sociometry. Describe the different types of surveys, including personal interviews, mail questionnaires, and telephone surveys, and explain how the different types of interviews could be implemented to study loss prevention.

RESEARCH EXERCISES 1. Using the search procedures you developed in Chapter 3, locate an article or government report that uses a qualitative or quantitative research design. 2. Determine the type of qualitative or quantitative research design within the

article or government report. Assess the effectiveness of research design in answering the research question. 3. In the Chapter 7 Research Exercises, you developed a hypothesis based on your research idea. Determine the type of information that could be collected with one qualitative research design and one quantitative research design. Based on the idea of mixed methods, explain how the information from both research designs could be related.

CHAPTER 9: Questionnaire Development What You Should Know! Data provides the basis of research. Chapters 6 through 8 describe different types of research designs, which provides the structure of the research study and is directly related to data collection. One of the most popular means of gathering data is through a questionnaire. This chapter specifies steps in developing questionnaires for collecting data. After completing this chapter, the reader should be able to: 1. Define the term questionnaire and explain its association with the term survey. 2. Explain how a questionnaire could be used in both quantitative and qualitative research. 3. Describe what is involved in listing the items one is interested in knowing about the group, concept, or phenomenon.

4. Explain how to establish validity and reliability. 5. Discuss why the wording in the questionnaire must be appropriate for the target audience. 6. Explain why it is necessary to clearly identify who should answer the questions. 7. Discuss why one should avoid asking any questions that are biased, leading, or doublebarreled in nature. 8. Explain why before construction, a decision must be made whether to use open- or closeended questions or a combination of both. 9. Describe the possible impact of the respondents not having the general information needed to complete the questionnaire. 10. Describe the benefits of pretesting a questionnaire before it is used. 11. Discuss why one should set up questions so that the responses are easily recognizable whether the questionnaire is selfadministered or conducted in an interview. 12. Explain why the questionnaire should be organized in a concise manner that keeps the interest of the respondent, encouraging him or her to complete the entire questionnaire. 13. Define the term scales and explain their purpose.

14. Compare and contrast the Thurstone Scales and Likert Scales.

Surveys and Questionnaires Surveys are one of the common methods for collecting data in criminal justice and criminology. The word survey often creates confusion in research methods courses because it refers to both a research design and data collection device. The term questionnaire tends to be used interchangeably with the term survey. For the sake of clarity, questionnaire will be used within this textbook to refer to the survey data collection device. Questionnaires as a form of quantitative research design rely on the numerical assessment of data. This assessment consists of a series of questions with specific responses provided from which the subjects choose the one that is the most appropriate. In a survey study regarding the occupational stress of correctional officers, a question may be “How stressful is your job?” and the possible responses could be “not stressful,” “neutral,” or “stressful.” Each of the possible responses would be assigned a number: 1 for “not stressful,” 2 for “neutral,” and 3 for “stressful.” Typically, the lowest assigned number indicates the least (e.g., least amount of stress) and the highest assigned number indicates the most (e.g., most

amount of stress). Setting up the questions in this logical manner will simplify the analysis and interpretation of the information collected. Questionnaires consisting of open-ended questions are also used within qualitative research, including structured and semi-structured interviews. As discussed in Chapter 8, types of surveys include personal interviews, mail questionnaires, and telephone surveys. The data collection for each type of survey varies in form, but each is based on the use of a questionnaire. Within the personal interview, the interviewer and interviewee meet face-to-face, and the interviewer reads the questions aloud and records the responses. The interviewer may also record the observations about the behavior or demeanor of the subject, such as ranking the level of the subject’s nervousness on a scale of 1 (least nervous) to 10 (most nervous). In the mail questionnaire, a hardcopy or electronic copy of the questionnaire is provided to the subject, who completes and returns it. The telephone survey is similar to the personal interview in that the interviewer reads the questions aloud and records the responses; however, the interviewer and interviewee are not face-to-face.

FROM THE REAL WORLD

Dantzker (2010) created a study-specific questionnaire to examine whether there were any differences regarding what they used to conduct the evaluations and the reason for their choice between police and general psychologists who performed preemployment psychological examinations of police recruits. Because of the information sought, no existing questionnaire could accomplish this goal and a new instrument was necessary. The questionnaire began with closed-ended questions answered through telephone interviews and was then developed into a self-administered questionnaire (Dantzker & McCoy, 2006), which eventually was revised into the final questionnaire used for the final study (Dantzer, 2010).

When conducting survey research, examine questionnaires that have been developed and tested previously to see if one of the questionnaires would be appropriate for your research. The primary reason for including a previously used questionnaire is that it reduces the likelihood of validity and reliability issues, two major concerns of questionnaire development. However, an instrument may not exist for a particular research question, or if

it does exist, it may not meet the researcher’s specific needs. Therefore, the researcher must resort to creating a research-specific questionnaire. Researchers may also use an existing instrument for the basis of their questionnaire by making adjustments where needed. In a study of the impact of violent victimization on post-traumatic stress, researchers may use an existing victimization survey as the basis for questions related to violent victimization. In creating a new survey instrument there are several things to consider, including reliability, validity, and the levels of measurement to use. Every textbook offers a different way to approach this task, but in the end, the basic elements of questionnaire development are similar. To simplify this task, a set of guidelines are offered in the manner of rules.

Rules for Questionnaire Construction Many people prefer not to follow rules, or at least to bend them to meet their satisfaction. The general rules presented here do not need to be followed in a strict manner but do provide a useful path to avoid a failed questionnaire. For many students, as well as researchers, getting started is the hardest part.

Figure 9-1 Sample Questionnaire. Source: Reproduced from Dantzker, M. L. and Waters, J. E. Examining students’ perceptions of policing—A pre- and postcomparison between students in criminal justice and non-criminal justice courses. In Dantzker, M. L., ed. Readings for Research Methods in Criminology and Criminal Justice. ButterworthHeinemann. Copyright Elsevier 1999.

FROM THE REAL WORLD In Dantzker and Waters’s (1999) study of students’ perceptions of policing, the data for their sample was gathered using a

questionnaire and included the information shown in Figure 9-1. As you can see in this figure, basic demographic information is collected, including gender, age, and race, and then questions about the student’s year and major. In this case the major is used as a qualifying question. If subjects answered that their major is criminal justice, then they answered the questions regarding perceptions of the police. Subjects with other majors did not answer the questions. Ford and Williams (1999) used a shortanswer question approach in their survey instrument to collection information regarding the impact of taking a human diversity course (Figure 9-2). This short-answer survey can be compared to the previous From the Real World example. Both questionnaires provide an example of the mixed methods approach, discussed in Chapter 8, with the inclusion of open-ended questions.

Figure 9-2 Short-Answer Question Approach. © Jones & Bartlett Learning.

Regardless of which approach is used, the key is to ask all the questions or create all the statements believed necessary to obtain the information desired. Listing the information that you want to collect from a sample seems easy, but it can be more difficult than it first appears. Often researchers run into problems because they fail to list everything they know or want to know about the subject. This failure could cause a shortage of important data because once the information has been collected, researchers can rarely follow up with the subjects. The key is to decide what is of interest and what is needed and then base the questions on this information. One can always eliminate data, but it is difficult to go back and get what was missed. In sum, Rule One suggests making a list of information

desired before creating any questions or statements.

Rule One: Start With a List of All the Items One Is Interested in Knowing About the Group, Concept, or Phenomenon How many times have you gone grocery shopping without a list? When that happens, it is common to end up with many things not wanted and missing things that are really needed. Questionnaire development should be approached with a grocery list mentality. List all the things you want to know, things you would like to know, and things that would be interesting to know about each item about which information is being sought. Keep in mind that you are not committing to the items on this list, and some of the items will be excluded from the final version of the questionnaire. Think of this list as the “everything” (including the kitchen sink) list. Once you have the everything list, it will be much easier to remove and modify items rather than having to add items. For example, the primary interest of Dantzker’s (2010) study was to examine whether there were any differences between psychologists as to the measures they use to conduct pre-employment

psychological screening as well as the reasons for their choices. To collect the relevant information, a questionnaire had to be developed that asked the respondents to provide certain identifying characteristics (the independent variables) along with their responses to questions about preemployment screening (dependent variables). What was asked depended on the type of comparisons or analysis that was to be conducted. Because two groups were being compared, minimally required questions included collected information about the respondent and about the evaluation. Regardless of the nature of the sample, similar questions in various formats should be asked.

Rule Two: Be Prepared to Establish Validity and Reliability If an existing questionnaire is not available for use within the study, then establishing validity and reliability for the instrument is critical. Validity and reliability were briefly discussed in Chapter 4. If the questionnaire is not considered to be both valid and reliable, then other researchers may not place much value in the results. Validity refers to whether the questionnaire is in fact measuring what it claims to measure. It is imperative that the questions or statements be a

true measure of the topic under study, which requires that the researcher be able to determine and establish the questionnaire’s validity. There are different types of validity, including: face, content, construct, and criterion validity. Each has a specific purpose in establishing validity.

Face The simplest means of establishing validity is face validity. Essentially, it is a judgment call by the researcher that the questionnaire is measuring what is being attempted to be measured. However, this type of validity lacks empirical support and requires the researcher to demonstrate why he or she believes it measures what is expected. Face validity decisions are often based on the researcher’s knowledge, background, and observational experiences. Although this process is an acceptable means of validation, it is the least acceptable because it is empirically weak.

Content The second form of validity, content validity, also suffers from being judgmental and usually nonempirical. Unlike face validity, where the belief in validity focuses on the questionnaire as a whole, content validity emphasizes each individual item’s ability to measure the concept in question. Instead

of simply supporting the complete questionnaire’s ability to measure what is expected, the researcher must be able to explain why each item measures what is expected. Again, the responsibility falls to the researcher to support why it is believed that each item measures what is expected.

Construct Although one is seeking to measure a particular phenomenon, there may be related concepts that are equally important to understanding the phenomenon in question. Construct validity seeks to demonstrate that the questions actually measure what they have been designated to measure by examining the relationship the measure has to other variables. The interest here is in establishing the fit between the theoretical and operational aspects of the item. With respect to the job satisfaction study, it was expected that individuals with a college degree would indicate a higher level of satisfaction with some items than individuals without a college degree. If the responses to those particular questions support this measure, then one has established construct validity. However, if the responses from both groups indicate equal satisfaction, then the construct validity may be challenged. Construct validity can be reinforced

through empirical measures, unlike face or content validity.

Criterion Criterion validity is concerned with the relationship between the questionnaire’s results and an external criterion. The assumption is that if the questionnaire is valid, a specified empirical relationship should exist between the data collected and other existing measures of the phenomenon. There are two types of criterion validity: concurrent and predictive. The key requirement for concurrent criterion validity, however, is that a reliable and valid measure must already exist to make the comparison. To apply criterion validity to a job satisfaction measure, one group of officers is given the job satisfaction measure, which is then compared to the results of a reliable and validated job stress measure (because it has previously been established that job stress is linked to job satisfaction). With the two sets of results, a correlation coefficient can be computed to provide what is called the validity coefficient. The more common form of criterion validity is predictive validity, which rests on the questionnaire’s ability accurately to predict future conditions or responses. Upon exiting prison, offenders may have to complete an assessment designed to determine the likelihood they will recidivate. After three years, the

prison may compare the results of the assessment and the actual occurrence of recidivism. If offenders who were assessed as more likely to recidivate actually had a higher occurrence of recidivism, then predictive criterion validity has been established. There may be debates over the best type of validity. Failing to establish any type of validity devalues the data. The more ways one can establish validity, the better. Still, validity alone is not enough. It must be accompanied by reliability. Everyone who lives in a climate where there are several days of extremely cold weather wants a car battery that starts the vehicle day in and day out. Perhaps the battery was purchased because of its reliability to do just that. When it does not start as expected, it is no longer considered reliable. A questionnaire has the same expectation—that it reliably does what it is designed to do every time it is used. If the questionnaire is consistent over time and yields similar results each time it is used, it is reliable. To establish reliability further, one must demonstrate stability and consistency.

FROM THE REAL WORLD As previously noted, Dantzker (2010) created a questionnaire specifically for his

research. For this questionnaire, face validity was addressed through the research-er’s knowledge of the subject, comparison to previous research efforts, and support from neutral observers with questionnaire construction experience and knowledge of the subject. The questionnaire also met the criteria for content validity. The content of the questionnaire was primarily the identified protocols and reasons for choosing them, both of which have been established by relevant literature. The questionnaire was reviewed by three licensed psychologists who all indicated they found that the questionnaire does reflect the intended objectives. Finally, for criterion and construct validity, the questionnaire had to cover the full range of possibilities within the concept and required a preexisting questionnaire with which to compare or a concept that cannot be directly observed or isolated. Because the concepts measured were based on actual use and established reasons for its use, the questionnaire was found to measure the primary concept.

Stability occurs when, under similar conditions, a respondent provides the same answers to the same

questions on a second testing. It is reasonable to expect that individuals would have similar results on IQ (Intelligence Quotient) tests two weeks apart. However, not all measures are expected to be stable. The expectation is that questionnaires designed to measure fear would be noticeably different before and after a terrorist attack. Consistency is determined when the set of questions is strongly related and provides predictable results. There are three standard ways to test reliability: (1) test– retest (pretesting); (2) split-half technique; and (3) using multiple forms. Pretesting is the best method for testing reliability, yet perhaps the most inconvenient in terms of time and money. The test–retest method requires distributing the questionnaire to the same population twice. If the results are the same, then reliability is accepted. Another method is to distribute the questionnaire to similar samples and look for consistent results between the samples. A popular and widely used method is the split-half technique. Here the questionnaire is divided into sections or halves. Both sections are given to the same group or among similar groups. A similarity of scores between both halves supports reliability. Using several variations or formats of the same

questionnaire, the multiple forms method can support reliability. As with the previous methods, if scores on each format are similar, one can assume reliability. All these methods are acceptable; however, when possible, one should use the test–retest method. In addition, many statistical packages offer methods for statistical comparison by item-to-item and itemto-scale analyses and the use of Cronbach alpha, a commonly used reliability coefficient. Overall, there are a number of ways to establish validity and reliability, and it is important to do so.

Rule Three: Word the Questionnaire Appropriately for the Target Audience In Western democracies it sometimes feels as if we are constantly being required to complete questionnaires of one type or another. Sometimes the questions are quite clear. However, at other times the questions may befuddle respondents. When developing a questionnaire, the first guideline is to be sure to use language geared toward the target population. One should not use words or phrases with which the respondent is not familiar because it can cause confusion and misunderstanding, which may lead to tainted data.

For example, on a questionnaire for college students on criminal behavior, a question about drug use should use the common usage (or slang) rather than scientific: Have you ever smoked cannabis? (wrong); Have you ever smoked marijuana? (right). Also, keep in mind the reading level of your audience. If your audience is the general public, then writing questions at a graduate school level could create problems because some members of the audience would not understand the questions. Pretesting the questionnaire within a similar audience to your target audience can help identify issues. It is important that the questions or statements be written in a manner that the target audience can understand.

Rule Four: Clearly Identify Who Should Answer the Questions You have probably received a questionnaire in the mail that has been addressed to Dear Occupant or Dear Current Resident. On opening it, you may have discovered that it is not clear who should be completing this questionnaire. For his questionnaire, Dantzker (2010) included this question and response instructions to clarify who should complete the questionnaire:

Do you personally provide psychological screening services to

any type of police/law enforcement agency? Yes

□

IF YES, please respond to the remainder of survey.

No

□

IF NO, you may stop here. Thank you.

All it takes is a simple statement advising who should complete the questionnaire. If the questionnaire requires that multiple subjects complete different questions within the same survey, then the directions should clearly identify who should be completing which questions. For example, the first data collection from the panel study National Longitudinal Survey of Youth (1997) used a questionnaire to acquire information from both youth and their parents. It is important to clearly specify who fills out or responds to which questions.

Rule Five: Avoid Asking Questions That Are Biased, Leading, or Double-Barreled in Nature Questions must be well worded. Questionnaires should not contain questions designed to elicit a specific response. “Does it not feel great to get high?” and “Do you cheat on exams and if you do, do you know you are only cheating yourself?” are examples of biased and leading questions because they create a push toward a specific response and

convey the researcher’s bias. Questionnaires should also not include double-barreled questions, which are questions that actually contain more than one questions. For example, “How often do you get high and do you enjoy it?” is a double-barreled question. The simple fix to a double-barreled question is to split it into two different questions. Questions or statements that confuse respondents can cause ambiguous responses, and questions that seek to guide the respondent can create blatantly false responses. In addition, the structure of the questionnaire may be such that preceding questions influence the responses to later questions. If a question regarding recent incidents of terrorism precedes questions regarding the respondent’s fear of terrorism within a questionnaire, then it is possible (or even probable) to assume that presence of the previous terrorism question would increase fear of terrorism. If any of these issues with question wording, structure, or instructions occurs, then the validity of the findings is decreased. Ultimately, the wording of questions and the questionnaire format can influence responses. Be particularly aware of these issues when responding to surveys conducted by ideological groups or

organizations that hold particular positions on controversial issues, such as the opposing views of gun control advocates and the National Rifle Association. Potential biases may also be observed among the questionnaires of scholars and students who are trying to support a favored hypothesis. If those administering a survey have an interest in or might benefit from the outcome of the survey, they may word the questions or structure the format so as to enhance the likelihood of desired responses. Though this issue may be unintentional, be skeptical of study findings by such individuals or groups.

Rule Six: Before Constructing a Questionnaire, Decide Whether to Use Open- or Closed-Ended Questions or a Combination of Both Because the goal of the questionnaire is to acquire specific information related to the topic for ready analysis, deciding what type of questions should be asked is crucial. Open-ended questions can make data analyses somewhat more difficult but can provide more in-depth responses. On the other hand, well-constructed closed-ended questions can provide sufficient data that are easier to analyze. An increasingly popular method is to combine the types of questions to collect the most pertinent information

about the topic, which is an example of mixed methods research discussed on Chapter 8. The questionnaire in Figure 9-3 was created as a telephone survey to investigate community satisfaction with a police department. Although this questionnaire had been created as part of a proposal to evaluate a police agency, it was never tested or published elsewhere. Observe how the questionnaire is composed of both closed- and open-ended questions. Notice how the open-ended questions are specifically worded so that responses could be more readily coded for statistical analysis. For example, the inclusion of “the main problem”, “the best thing”, “the worst thing”, or “one change” within the question are designed to elicit very specific opinions and do not ask for an explanation for the opinion, which would be much more difficult to code.

Rule Seven: Keep in Mind That Respondents May Not Have All the General Information Needed to Complete the Questionnaire Under any circumstances, making assumptions could be problematic, but this is a special risk in questionnaire development. It is a common error to believe that possible respondents have all the

information needed to respond to the questionnaire. For example, several questions about drug and alcohol programs on campus might be asked in a Student Criminality study simply because it is assumed that all students are familiar with these programs. This assumption might result in few responses because respondents do not have the necessary information and choose not to respond. The responses that are received could be biased because only students who are familiar with these programs may have responded. Or students may have responded based on their own assumptions about the programs rather than firsthand knowledge. To avoid this dilemma, always provide an “escape” response, such as “unknown,” “no opinion,” or “unable to respond” (refer to Figure 93).

Figure 9-3 Questionnaire: Open- and Closed-Ended Questions. © Jones & Bartlett Learning.

Rule Eight: Whenever Possible, Pretest the Questionnaire Before It Is Officially Used Although this is a time-consuming step, pretesting is undoubtedly important and should be used whenever possible. For example, with the Student Criminality questionnaire, after completing the first draft of the questionnaire, researchers may have students in their classes complete the questionnaire. Pretesting can uncover errors in construction, language, or other problems that may cause the data to be useless if not corrected. Keep in mind that this process may require just a few

individuals from the target population willing to complete the questionnaire and provide feedback. Although this process might take a little more time and effort, it is well worth it in the long term and can also be used to assess reliability. Returning to Dantzker’s (2010) questionnaire, pretesting could make sure that the instructions are clear and that there are no issues with the questions (Figure 9-4).

Figure 9-4 Sample Questionnaire with Specific Instructions.

© Jones & Bartlett Learning.

Rule Nine: Set Up Questions So That the Responses Are Easily Recognizable Whether the Questionnaire Is SelfAdministered or an Completed in an Interview The fastest way to jeopardize research is through a questionnaire in which the directions are not clear on how to respond. Be sure to provide adequate, clear instructions and establish recognizable means for responding. Also, try not to make the format too busy, such as the overuse of decorative designs. Avoid using small print because it is hard to read and can be confusing if too many questions are squeezed on a single page.

Rule Ten: Organize the Questionnaire to Keep the Respondents’ Interest, Encouraging Them to Complete the Entire Questionnaire How often have you started a novel only to give up after the first few chapters because you found it boring? If you had continued to read, it might have become more pleasurable, but you lost interest. This same concept is applicable to questionnaire

development. If the beginning questions are not interesting and do not hold the respondents’ attention, chances are that they may not complete the rest of the questions, which results in missing data. Therefore, it is beneficial to create questions that may pique respondents’ interest at the beginning and the end. In addition, if respondents see several pages of questions, they are less likely to begin the survey. Although it is tempting to try to cover everything, a clear and concise survey that consists of a few easy-to-read questions typically receives more responses than a lengthy questionnaire. Although specialized questionnaires that target a specific population (who may have a vested interest in the subject matter) may be longer, it is recommend that most surveys be kept to two pages of questions using normal-size type. Rules and guidelines are fallible, but the 10 rules offered for questionnaire construction, if followed, improve the chances of obtaining quality, analyzable data. Still, keeping all the rules in mind might not prevent the creation of a poor questionnaire. Beyond these questions, a key aspect in questionnaire construction within a quantitative research design is development of questions that use scales.

Scales A common element of quantitative survey research is the use of scales. A scale can either be a measurement device for responding to a question or statement or a compilation of statements or questions used to represent the concept studied. For example, in the questionnaire in Figure 9-3, responses to several statements regarding police performance are on a scale that ranges from 1 (wholly unsatisfactory) to 4 (completely satisfactory). Each statement represents a separate variable of the respondent’s perception. Putting the statements from each perceptual question together (attaining a numerical result for responses to all statements) gives the researcher what might be referred to as the “Community Members’ Perceptions of Their Local Police Agency Scale.” Scales as compilations are particularly important and a relevant part of research for three primary reasons (Creswell, 2008; Dunn, 2009; Frankfort-Nachmias & LeonGuerrero, 2008): (1) to allow the collapsing of several variables into a single variable producing a representative value and reducing data complexity; (2) to offer measures that are quantifiable and more open to precision and statistical manipulation; and (3) to increase the measurement’s reliability.

To accomplish these ends, a scale must fit the Principle of Unidimensionality (Frankfort-Nachmias & Leon-Guerrero, 2008). According to this principle, the items making up the scale need to represent one dimension befitting a continuum that is supposed to be reflective of only that specific concept. For example, if one is measuring job satisfaction, the scale should not be capable of also measuring job stress.

Scaling Procedures Ultimately, to conduct research one looks to complete a measurement. But what is the actual purpose of this measurement? Measurement is used in research as a means of connecting phenomena with numbers for analytical purposes. Scaling is identified as a means of assisting in making these necessary and proper connections. Despite the existence of numerous scales, there are times when researchers must create their own. The key is to keep in mind that the goal is to explain a phenomenon. Thus, the scale must meet this goal. There are two primary types: arbitrary scales and attitudinal scales.

Arbitrary Scales An arbitrary scale is designed to measure what the researcher believes it is designed to measure and is

based on face validity and the professional judgment of the researcher. Although this process allows for the creation of many different scales, it is easily criticized for its lack of substantive support. Still, this type of scale does provide a viable starting point for exploratory research, even though it is the less recommended method of scaling.

Attitudinal Scales More often employed in criminal justice and criminological research than arbitrary scales are two common types of attitudinal scales: Thurstone and Likert.

Thurstone Scales The construction of a Thurstone scale relies on the use of other individuals (sometimes referred to as “judges”) to indicate what items they think best fit the concept. There are two methods for completing this task. The first method is paired comparisons. Here, the judges are provided several pairs of questions or statements and asked to choose which most favorably fit the concept under study. The questions or statements picked most often by the judges become part of or comprise the complete questionnaire. For example, the judges are asked which of the following questions might best fit the concept of job satisfaction: “I enjoy going to work

every day.” “Sometimes I am very tired when I get home from work.” The second and more frequently used procedure is referred to as equal-appearing intervals. For this method, the researcher submits a list of questions or statements to the judges who are then asked to give each a number, indicating the strength of the question or statement to the concept. The researcher then keeps those items on which judges were in the strongest agreement and eliminates those with the weakest indicator scores. For example, the researcher decides to design a questionnaire to examine criminality among college students and calls it the “College Students’ Criminality Questionnaire.” A decision is made to have the questions form a 15-point item criminality scale. Fifty questions are submitted to judges, who are instructed to score each question from 1 (strongest indicator) to 15 (weakest indicator). The top 15 questions become the scale. Thurstone scaling is not very popular because of the time it takes for the judges to complete their tasks. Furthermore, because the judges must be experts in the area of study, finding an adequate pool of qualified judges could be problematic for the researcher.

Likert Scales Probably the most commonly used method in attitudinal research is the Likert scale. This method generally makes use of a bipolar, five-point response range (i.e., strongly agree to strongly disagree). Questions where all respondents provide similar responses are usually eliminated during the data analysis portion of study, which will be discussed in Chapter 11. The remaining questions are used to comprise the questionnaire. Figure 9-5 provides an example of Likert-type questions. One of the benefits of the Likert scale is that it provides a neutral response, such as the “Doesn’t Really Matter” option located in the middle of the scale. Another benefit is the balance it provides between collecting detailed information and not being so intensive that respondents do not finish answering the questions. There are various other types of scaling procedures. However, because so few are used in criminal justice and criminological research, they are not discussed here. Furthermore, advanced statistical techniques, such as factor analysis and Chronbach alpha, are much faster and simpler to use to determine a scale’s composition. For example, the Student’s Perceptions of Police scale (Dantzker & Waters, 1999) was originally 20 items before a

factor analysis and Chronbach alpha eliminated six items, bringing it to its final 14-point scale. The question at this point is, why use scales at all?

Figure 9-5 Example of Likert Scale Statements. Source: Reproduced from Eisenman, R., & Dantzker, M. L. (2006). Gender and ethnic differences in sexual attitude at a Hispanic-serving university. Journal of General Psychology, 133(2): 153–162. Reprinted by permission of the publisher (Taylor & Francis Ltd, http://www.tandf.co.uk/journals).

There are two primary reasons or advantages to using scales. First, a scale allows for a clearer and more precise measure of the concept than individual items. Second, scales can be replicated and used

as longitudinal measures. The disadvantages to scales are twofold: There is concern as to whether true attitudes can be measured on a scale, and there is the question of validity and reliability. Despite the shortcomings, overall, scales can be quite useful in measuring data and should be used where and when it is appropriate and necessary.

Summary To conduct research, data must be collected, which generally means that some type of tool, such as a questionnaire, must be available to assist in collecting the data. If a questionnaire is available that has previously been tested and used, this option may be preferable to designing a new questionnaire. However, when there is not an existing questionnaire, then a new questionnaire must be constructed. Following the suggested rules provided in this chapter can help create an acceptable tool. The rules are as follows: 1. Start with a list of all the items one is interested in knowing about the group, concept, or phenomenon. 2. Be prepared to establish validity and reliability. 3. Word the questionnaire appropriately for the target audience.

4. Identify clearly who should answer the questions. 5. Avoid asking questions that are biased, leading, or double-barreled in nature. 6. Before construction, decide whether to use open- or closed-ended questions or a combination. 7. Consider that the respondents may not have the general information needed to complete the questionnaire. 8. Whenever possible, pretest the questionnaire before it is officially used. 9. Set up questions so that the responses are easily recognizable whether the questionnaire is self-administered or completed in an interview. 10. Organize the questionnaire to keep respondents’ interest, encouraging them to complete the entire questionnaire. In addition to the rules, questionnaire development requires familiarity with such issues as reliability, validity, measurement level, and scales. Scales can be either arbitrary or attitudinal in nature. Two popular attitudinal scales are Thurstone and Likert, with the most popular scale being the five-point Likert scale.

APPLICATION EXERCISES 1. As state legislatures consider various bills related to reducing gun violence, you have been instructed by the legislator you work for to develop a questionnaire to collect information from her constituents. Of particular interest are the “campus carry” laws, which may allow individuals to carry firearms on college campuses. Now she wants to know what you think should be included in the questionnaire. Describe what you are interested in knowing about the topic of how to reduce gun violence and how campus carry laws might affect that goal. 2. Since an appropriate questionnaire doesn’t exist for this topic, you explain to the legislator how you will have to create a new questionnaire. She wants to know what that process will entail. Answer the following questions within your explanation: a. How will a list of possible items related to the topic be created? b. How will you establish validity and reliability, and why is this a

c.

d.

e.

f.

necessary step? Why must the wording in the questionnaire be appropriate for the target audience? Why must the instructions clearly specify who should answer the questions? How and why should you avoid asking any questions that are biased, leading, or doublebarreled in nature? Why should the questionnaire be pretested before it is officially used?

3. Why you have selected either a Thurstone or Likert scale?

RESEARCH EXERCISES 1. Using the search procedures you developed in Chapter 3, locate an article or government report that uses a questionnaire. 2. Describe the type of information collected through the questionnaire and how the information was used within the study.

3. In the Chapter 7 Research Exercises, you developed a hypothesis based on your research idea. Create a list of the information that could be collected with a questionnaire. For one item from the list, create either a Thurstone scale or a Likert scale.

CHAPTER 10: Data Collection What You Should Know! To this point, this textbook has described how to structure a study, from choosing a topic to selecting a sample to determining a research design. The selected research design will determine the appropriate data collection technique. As with sampling strategies and research designs, each data collection strategy has strengths and weaknesses. This chapter examines a variety of methods for collecting data. Keep in mind that some of the data collection strategies can be used with multiple research designs whereas other strategies are more specific. After completing this chapter, the reader should be able to: 1. Identify and describe the four primary data collection techniques. 2. Explain the advantages and disadvantages of mail surveys. 3. Describe the advantages and disadvantages of Internet surveys.

4. Discuss the advantages and disadvantages of structured, semi-structured, and unstructured interviews. 5. Compare and contrast face-to-face and telephone interviews. 6. Discuss the advantages and disadvantages of field observation. 7. Explain the advantages and disadvantages of analyzing secondary data. 8. Describe the advantages and disadvantages of content analysis. One of the most crucial aspects of the research effort is data collection. Improperly collected or incorrect data can delay or even cause the cancellation of the research effort. Therefore, before researchers begin any type of data collection, they must be sure to choose the right data collection technique based on the selected sampling strategy and research design. Although one of the best methods of collecting data is through the experimental design, this method is not conducive to social science research. However, there are effective and efficient alternatives. In criminal justice and criminological research, four types of data collection techniques are commonly used: (1) survey, (2) interview, (3) observation, and (4) secondary data methods. Interviews have been

broken out of survey research for discussion purposes in this chapter, but keep in mind that interviews are usually considered to be a component of or related to survey research. These research methods have been discussed in detail in previous chapters. The focus in this chapter is on the issues involved in using these data collection strategies.

Survey Research The most frequently used method for data collection is the survey. A survey is an excellent tool for gathering primary data and is quite useful in both descriptive and correlational studies. In criminal justice and criminology, some of the uses for surveys include measuring attitudes, fears, perceptions, and victimizations. The data collection device for a survey is a questionnaire. One reason that a questionnaire is so useful in data collection is that it can be self-administered, thus an interviewer or researcher does not have to be present while an individual answers the questions. Perhaps the most commonly used method of distributing a selfadministered questionnaire is through the mail.

Mail Distribution of Surveys Mail distribution of surveys allows for the use of larger samples, broader area coverage, and

minimized cost in terms of time and money as compared to other data collection methods. Additional advantages include that no field staff is required, the bias effect possible in interviews due to the presence of an interviewer is eliminated, and respondents are allowed greater privacy to answer the questions. With fewer time constraints on respondents, they can give more consideration to their answers. The mail survey method also poses several disadvantages. One of the most frustrating is the lack of responses. The percentage of subjects who complete and return the questionnaire is known as the response rate; this rate is determined by dividing the number of responses by the total number of possible responses. It is common to send reminders to participants to increase the likelihood of subject participation. For example, a researcher may plan to send follow-up postcard reminders to subjects who had not responded at three-week intervals or follow up with a phone call. Although following up with subjects helps increase the number of respondents, it also can add to the costs; experienced researchers know to include the follow-up as part of the original research plan. Low response rates can lead to the delay or cancellation of a study.

Another disadvantage is the possible differences that might exist between the respondents and nonrespondents. Individuals may not respond for any number of reasons. The survey may have gotten lost, or respondents might not have time to complete it. The likelihood of bias is increased if the reason an individual did not respond to the survey has to do with the survey itself. If only individuals who had an interest in the topic responded, then the findings might be biased toward a particular response, which would not provide valid and reliable information. For example, if a survey was mailed to individuals in a specific neighborhood about building a halfway house for newly released prisoners and only those who were opposed answered, the results would be biased. The key to data collection for a mail survey is to attain the highest response rate possible. In addition to following up with subjects, other ways to increase response rates include (Adler & Clark, 2007; Bachman & Schutt, 2008; Dunn, 2009; FrankfortNachmias & Leon-Guerrero, 2008; Shaughnessy, Zechmeister, & Zechmeister, 2008): 1. Using an attractive and shortened format. It is surprising how much more likely an individual is to respond to a questionnaire that is

2.

3.

4.

5.

visually interesting. Furthermore, keeping the format short and simple is more likely to generate a better response than a lengthy questionnaire, which respondents may start but not finish. Or they may be dissuaded from starting in the first place due to the length of the survey. Offering some type of remuneration or “reward” for completing the survey. One popular reward is to offer a cash incentive, such as a gift certificate or the chance to win free groceries, if the survey is completed. Appealing to respondents’ altruistic side or sense of civic duty with the message that their response will be extremely helpful in learning more about the subject of the research. This strategy is more likely to work if respondents are already interested in the subject. Indicating that the survey is sponsored or endorsed by a recognizable entity, such as a college or university. Respondents may be more encouraged to complete a survey when they recognize and respect an entity supporting the research. Personalizing the survey. Addressing the questionnaire to a specific person often adds more legitimacy to the research as opposed

to addressing the survey to “Dear occupant or resident.” 6. Timing the survey carefully. When a survey is sent could be extremely important. For example, many individuals spent hours and hours watching the O.J. Simpson trial. Research geared toward public perceptions of the court process as observed in the trial probably would have received much better response immediately following the completion of the trial than if such a study had been conducted in 2015. However, if a survey were to be conducted today, we may find the results to be significantly influenced by 2016’s The People v. O.J. Simpson: American Crime Story television series.

FROM THE REAL WORLD To examine social determinants that explain correctional officer exposure to blood and bodily fluids, Alarid (2009) conducted a study of experienced correctional officers from five prisons. The correctional officers were selected at random and sent an anonymous mail survey regarding situations that may have placed them at risk for exposure to HIV while at work. Although a total of 500 officers were selected, only 192

officers returned the surveys for a response rate of 38.4%. Sixteen of those surveys were not included because of a disproportionate number of incomplete responses, making the final sample size 176.

Acceptable Mail Survey Response Rates The previous section discussed the frustration of dealing with low response rates and provided recommendations for enhancing return rates. Perhaps the best way to cope with low return rates is (in addition to following the recommendations that were provided) to allow for them in the research design. In Chapter 5, one recommendation for dealing with a low response rate is to oversample the population to meet sample size requirements. In mail surveys a 20% oversampling rate may not be enough. Researchers (Adler & Clark, 2007; Dunn, 2009; Frankfort-Nachmias & Leon-Guerrero, 2008) suggest these expectations for response rates for mail surveys: (1) 40% within 2 weeks, (2) 20% within 2 weeks of a follow-up letter, and (3) 10% within 2 weeks of the final contact. Response rates of 50% are considered adequate for analysis and reporting; however, 60% is good, and 70% is very good (Adler & Clark, 2007; Dunn, 2009). Note that based on the return rate indicated previously, it

may require several follow-ups to obtain a good or even adequate response rate. The response rate may vary depending on the type of survey and the targeted respondents. Ordinary citizens are far less likely to respond to a general survey than are members of a constituent group being polled by an organization in which they hold membership. Overall, a mail survey can be very effective and efficient.

Surveys and the Internet Although survey data collection through the mail has long been a popular method of obtaining data, technology has provided other methods of delivery. The technological changes over the last few decades have touched many aspects of our lives, and the way data is collected for research is no exception. Through the Internet, researchers can now reach target populations that are literally located throughout the world or simply within a single country or state. A popular Internet site for conducting surveys is Survey Monkey. SurveyMonkey. com allows the researcher to create a survey study by providing choices related to design, question type, and analysis. Surveys provided through Internet sites provide quick access for respondents, allowing them to complete a survey

at their convenience, and follow-up with nonrespondents can be completed through email. Dantzker (2010) used an Internet survey delivery method to conduct his study among clinical psychologists in the United States by setting up the questionnaire on SurveyMonkey.com, sending the sample members a link to the questionnaire, and providing a date by which the questionnaire had to be completed. This strategy allowed respondents to complete the survey at their leisure. Surveys through the Internet have the same advantages of mail surveys, including targeting large samples, covering wide geographic areas, cost efficiency, and the ability to address a wide variety of topics. Additionally, collecting survey data through the Internet makes data processing considerably easier than traditional mail surveys. The data collected through Internet sites automatically populates a database; thus, it does not require an individual to type in the information. Internet surveys also have the same disadvantages as mail surveys, including a low return rate, nonresponse to some questions, and misinterpretation or misunderstanding of the questions. The return rate may be even lower than that for mail surveys because individuals may not

receive the email announcement about the survey due to spam filters. Spam refers to unwanted or unsolicited emails. Overall, mail and Internet surveys offer an economical and efficient method for collecting data.

FROM THE REAL WORLD Hunter’s (1988) study of convenience store robberies used mail surveys that were sent to approximately 130 law enforcement agencies seeking robbery data on the 200 stores in the sample. The letter requesting assistance was from the Florida Attorney General to the Sheriff or Chief of Police of the agency stating their mutual interest in combating the crime of robbery and requesting assistance. The letter from a high-ranking state official, the brevity of the survey, and interest in the survey topic resulted in an initial response rate of approximately 67%. A follow-up packet containing the original letter from the attorney general and a new survey was sent after one month. A one-month response expectation is not unreasonable when requesting information from agencies that must be looked up or calculated since these actions must be done in addition to the day-

to-day activities of the agency. The follow-up resulted in an overall response rate of 92% by the end of the second month. A follow-up telephone call was then made to the 8% of agencies that had not responded. In the telephone calls respondents were asked either to respond by mail or by telephone in the next week. All but two agencies did so. Another follow-up telephone call to those agencies finally resulted in the information. A survey from a student or a professor without a connection with the attorney general’s office would have been fortunate to have achieved a 60% response rate.

FROM THE REAL WORLD Lurigio, Greenleaf, and Flexon (2009) explored whether the police-related views of African American and Latino students differ with respect to attitudes, feelings, and behavioral intentions toward the police. The survey data were obtained from students who were enrolled in 18 Chicago public schools. Student surveys were anonymous; due to confidentiality concerns, information was not collected on the characteristics of

the individual schools. The questionnaire consisting of 131 items in open- and closedended response formats was distributed during advising periods that the school had reserved for standardized test administration.

As useful as mail and Internet surveys are, some populations will not be accessible through these methods. For example, the population may consist of subjects to whom access is restricted, such as prison inmates; others may wish to remain anonymous or do not want to allow personal contact with researchers (e.g., police officers). However, these types of populations can offer vital data regarding numerous research topics. In these cases, a researcher may need to provide a written questionnaire that is distributed to the selected sample in a structured environment. Respondents are allowed to complete the survey within a given time period, and then they return it to the researcher or a representative of the researcher. For example, Gershon, Barocas, Canton, Li, and Vlahov (2009) used a hand-delivered survey to examine the impact of a wide range of police stressors on potential health outcomes while controlling for various coping strategies in a large sample of urban police officers. The sample was recruited during roll calls at each of the department’s nine districts and at three other

major divisions, including headquarters. Officers completed the questionnaires before going out on their shift. These types of survey do not typically allow for indepth responses or for the researcher to follow up on why a particular response was given. If this information is vital, then interviewing may be a more appropriate data collection strategy.

Interviews For the purpose of this text, interviewing is viewed as the interaction between two individuals where one of the individual’s goals is to obtain recognizable responses to specific questions. Interviewing is not an easy task and to be reasonably good at it often requires years of training. However, this does not preclude many researchers from trying this method. Interviews may be (1) structured, (2) semi-structured, or (3) unstructured. Typically, interviews are completed face-to-face, but can also be completed by telephone, Skype, and other teleconferencing activities.

Face-to-Face Interviews Face-to-face interviews provide contact between the researcher and the respondent. This contact can be

positive reinforcement for participating in the research. Often simply receiving a questionnaire in the mail can be sterile; because of that lack of personal touch, would-be respondents may simply ignore the survey. The interview also can usually guarantee a higher response rate. Another advantage to this type of interview is that any misunderstandings or confusion can be cleared up. This procedure helps ensure that the responses are accurate. It also allows for the researcher to act as an observer, giving the interviewer the opportunity to focus on nonverbal cues. The other advantages to the face-to-face interview include being able to use audiovisual aids, schedule additional interviews, make use of language the respondent can relate to, and show discretion. Although the advantages clearly outweigh the disadvantages, the latter must be considered in planning and implementing face-to-face interviews. One disadvantage is that face-to-face interviews are time consuming and costly because the interviewer has to travel to the interview. Another disadvantage is that the presence of the interviewer may impact the interviewee’s response. For example, if the interview is focused on illegal behavior, the subject may be less likely to be honest. There is also the

possibility of interviewer bias or error, which threatens the validity of the study. Finally, the interviewer’s skill or lack thereof could be detrimental to data collection (Bachman & Schutt, 2008; Dunn, 2009; Shaughnessy et al., 2008). It is common for researchers to use graduate students as interviewers. However, few graduate students come equipped with the skills required to conduct a good interview. Failing to provide or hone the necessary skills negatively affects the data. To avoid interviewer errors it is important to properly train interviewers who will be conducting the interview, which can be enhanced through the use of audio and video recording.

Structured, Semi-structured, and Unstructured Interviews The most used type of interview in criminal justice research is the structured interview. One type of structured interview, associated with quantitative research, uses closed-ended questions that every individual interviewed must be asked in the same order. Responses are set and can be checked off by the interviewer. The advantage to this type of interview is that it can be easily administered, has high response rates, and makes data processing much easier. In effect, this is an orally administered

questionnaire with the interviewer completing the form for the respondent. Two disadvantages are that it does not allow for further exploration of the responses and can limit the types of responses given. Another type of structured interview uses a standardized list of open-ended questions, which allows the interviewee to provide detailed responses. One disadvantage to this strategy is that since the list is standardized, the interviewer cannot deviate and ask follow-up questions. Another disadvantage, which is shared by the interview data collection strategies that use open-ended questions, is that it is much more difficult to analyze the data and interpret the results as compared to data collected using close-ended questions. Additionally, studies that use interviews based on open-ended questions typically have smaller sample sizes, but the trade-off is that more detailed information is acquired from subjects. The semi-structured interview consists of an original set of open-ended questions. It differs from the structured interview in that the interviewer can further explore why the response was given by asking additional questions. The advantage to this method of interviewing is that a substantial amount

of information can be obtained and clarification and follow-up can be completed during the interview if needed. In comparison to the structured and semi-structured options, the unstructured interview is more like a conversation in which appropriate open-ended questions are developed during the actual interview. The advantage to the unstructured interview is that it can lead to unexpected, but highly important topics that would not have been reached through a more structured approach. A disadvantage is that it is more susceptible to intervening or biasing elements. Unstructured interviews require an experienced and disciplined interviewer to be successful. The face-to-face interview can be a useful data collection technique. However, there are times when it is not possible to conduct face-to-face interviews, but the interview method is still necessary. This is when the telephone survey is useful.

FROM THE REAL WORLD To study how academic researchers gain access to data, Flanyak (1999) used semistructured interviewing. This method provided a high response rate and allowed

the researcher to have personal contact with the subjects and to observe them during the interview. To assist in conformity, Flanyak made use of an interview guide instrument, which provided her with the opportunity and discretion to explore or probe respondents for detail in specific areas. Most interviews lasted from 45 to 75 minutes, with the longest being two hours in length. Follow-up interviews were completed as needed and lasted approximately 15 minutes.

TABLE 10-1 Comparison of Survey and Interview Methods

Criteria

Mail

Internet

Face-to-

Telephone

Face Cost

Low

Low

High

Moderate

Response

Moderate

Moderate

High

High

Control

None

None

High

Moderate

Diverse

High

Moderate

Moderate

Moderate

Low

Low

High

Moderate

rates

population In-depth information

Timelines

Slow

Moderate

Fast

Fast

© Jones & Bartlett Learning.

Telephone Interviews There are several advantages to using telephone interviews. One is the ability to eliminate field staff, which can be replaced with a smaller in-house staff. It is also easier to monitor interviewer bias in telephone interviews, and the impact of nonverbal cues is eliminated. Although not common, unstructured interviews could be completed by telephone. Finally, the telephone interview is less expensive and can be completed more quickly than face-to-face interviews. Disadvantages to the telephone interview include: (1) limitations in the scope of research, (2) difficulty in obtaining in-depth responses, (3) elimination from the sample of anyone without a telephone, and (4) possible high refusal rates (Adler & Clark, 2007; Bachman & Schutt, 2008; Dunn, 2009; Shaughnessy et al., 2008). The increased use of cell phones as an individual’s primary number has resulted in the inability to create a full sampling frame, and the ability to identify and block callers allows possible subjects to avoid the calls of researchers.

There are advantages and disadvantages to the different types of survey and interview methods, and a brief comparison of the methods is useful (Table 10-1).

Field Observation Assume criminal justice researchers want to understand the inner workings of a gang or the impact of particular patrol techniques on citizen satisfaction. They could conduct interviews or create a self-administered survey, but neither of these methods may be able to give a complete picture because the researchers may not know everything that needs to be asked in advance. In such cases, observation may be the best method for collecting data.

FROM THE REAL WORLD Cavacuiti et al. (2013) conducted a qualitative study using semi-structured phone interviews to examine road rage incidents from the victims’ perspectives in Toronto. Newspaper advertisements were used to recruit participants who identified themselves as road rage victims and were at least 18 years old. The researchers completed 30- to 60-minute interviews with 29 subjects. Of the 29 subjects, 20 were in a

motor vehicle and 9 were either pedestrians or cyclists at the time of the road rage incident. The interviewer asked the participants to describe the events before, during, and after the road rage incident with a specific focus on environment (e.g., weather) and situational (e.g., being in a hurry) factors.

Field observation typically falls into one of four approaches: (1) full participant, (2) participant researcher, (3) researcher participant, and (4) complete researcher (Senese, 1997). Regardless of which approach is used, observation allows the researcher to see firsthand how or why something works. It provides an opportunity to become aware of aspects unfamiliar to those who do not have firsthand experience. To conduct observational data collection, the researcher needs to make several decisions: (1) where the observations are to be done, (2) what the focus of the observations will be, and (3) when the observations will be conducted (Senese, 1997; Shaughnessy et al., 2008). A choice of where observations are to be done is determined by the goal of the research study. For

example, to observe how incarcerated juvenile offenders respond to a particular treatment program may only be possible in the institutional setting. However, if the program is expanded to include nonincarcerated juvenile offenders, then the observations may also take place in a public setting (e.g., school). Selecting the wrong setting generally means an unsuccessful research venture, just as asking the wrong questions during an interview would lead to a failed study. First, it must be asked: What is it that one wants to know? Deciding what aspect to observe is another important element. If researchers know absolutely nothing about a phenomenon or entity, they may end up trying to observe everything about it. Researchers need to have an idea of what aspects are to be studied and how to do so. This process is similar to the first rule of creating a questionnaire (Chapter 9). The structure of field observation differs between qualitative and qualitative research. For qualitative research, it may be enough to be in the setting and record observations. However, quantitative research requires more preparation. If the research is quantitative, then the researcher needs to create a checklist in advance. Hunter (1988) created a checklist for use in his study of the impact of environmental factors on convenience

store robberies (Figure 10-1). Examples of items on the checklist include the location of the cashier (center or side), the number of clerks (one, two, or three), and security devices in use (yes or no). Finally, the time frame or period when the observations are to be conducted should be determined to provide the best possible opportunity of collecting the desired data.

Figure 10-1 On-Site Evaluation Form. Source: Reproduced from Hunter, R. D. (1988). The effects of environmental factors upon convenience store robberies in Florida. Tallahassee, FL: Florida Department of Legal Affairs. Courtesy of the Florida Department of Legal Affairs.

Like survey and interview research, observational research has its advantages and disadvantages. One of the greatest advantages is the direct collection of the data. Rather than having to rely on what others have seen, the researcher relies only on his or her own observations. However, this advantage could also serve as a disadvantage because of researcher misinterpretation or misunderstanding of what is seen. An extremely important part of field observation is recording the observations either in field notes, with an audio recorder, or by video as quickly and as accurately as possible, which also helps reduce inaccuracy and inconsistency. It is further recommended that one transcribe field notes as soon as possible afterward to ensure an accurate interpretation of hurried handwriting and any abbreviations that were used while they are still fresh in one’s mind. The fact that the research is being conducted in the phenomenon’s natural environment is a bonus. Recall that a shortcoming of experiments is that the environment is controlled, perhaps biasing results.

FROM THE REAL WORLD Rydberg and Terrill (2010) examined the impact of officer education on arrest, search, and use of force decisions using researcher participant field observation. The field

observations were collected as part of the Project on Policing Neighborhoods (POPN) in two medium-sized cities (Indianapolis, Indiana, and St. Petersburg, Florida). During 1996–1997, two beats were selected from each city based on the socioeconomic index (percentage of female-headed families, percentage of adults unemployed, and percentage of the population living below 50% of the poverty level). In the researcher participant role, the observers’ identities as researchers were known, but they made no effort to participate in the events being observed. The observers (university students) were trained for the study and took notes on the specifics of officer–citizen encounters, including the people involved and the events that occurred. In all, 11,985 officer–citizen encounters were observed. In addition to the field observations, researchers also interviewed the officers. The results of the study indicated that higher education reduced the likelihood of force occurring but did not impact arrest or search decisions.

Observational research takes place in the real world, legitimizing the observations—“This is what

actually happened” rather than “This is what should happen” or “This is what was thought to have happened.” However, observational, survey, and interview data collection may not be used in all circumstances. What happens when a topic deals with a phenomenon that has already occurred and can no longer be observed or surveyed? This type of data collection may need to rely on using secondary data.

Secondary Data Analysis of secondary data is an efficient way to conduct criminal justice or criminological research. Data are often available through a variety of sources, including government agencies, research groups, or other researchers (Box 10-1). Rather than gathering new data, researchers obtain and reanalyze the existing data that have been collected by the sources for their own purposes. For example, a corrections agency may have individual data on the length of individual sentences and recidivism based on reincarceration following release. However, it is the responsibility of researchers to verify that the data is appropriate for their purposes. The majority of data available for secondary analysis, particularly through government agencies, are quantitative; however, some qualitative data is also available. Data can also be collected from

social artifacts, such as written documents or recordings, for the purpose of completing a content analysis. The use of secondary data is a form of unobtrusive research. Unobtrusive research does not require researchers to be directly involved with the subjects of their studies. It also does not involve observation of or interaction with the individuals or groups because the data have already been gathered by someone else or the data are available in a format that does not require such interaction. Compared to other fields of study, criminal justice has a considerable amount of data available for secondary analyses.

Sources of Secondary Data Criminal justice and criminology researchers have a wide variety of official statistics and records available for the purposes of research. Government agencies collect a considerable amount of information, and some of the information is then in turn released for public use. One of the best known sources for criminal justice information is the Uniform Crime Reports (UCR) published annually by the Federal Bureau of Investigation (FBI). Police agencies from all over the country voluntarily provide data for the UCR on a monthly basis. The

UCR include data on arrests, clearances, and law enforcement employees. In addition to analyzing national data, researchers can also access data for each state or even the individual law enforcement agency. By using the UCR crime data, researchers could do an in-depth analysis of crime patterns or arrest characteristics without the arrestees or victims having any knowledge of the research in progress. In addition to the FBI, several other federal agencies provide data to the public for secondary analysis, such as the U.S. Bureau of Labor Statistics and Bureau of Justice Statistics. The National Longitudinal Survey of Youth 1997 (NLSY97) is a panel study of a nationally representative sample of youth who were 12 to 16 years old as of December 31, 1996, that is collected through the U.S. Bureau of Labor. The first NLSY97 survey of the selected youth and their parents took place in 1997, and the youth continue to be surveyed periodically. The purpose of the study is to collect information regarding entry into the workplace; the data include a substantial number of criminal justice-related variables, such as alcohol and drug use, illegal behavior, criminal justice contact, peer behavior, and perceptions of the future. Illegal behavior questions range from the destruction of property

and stealing (less than $50) to robbery and assault; questions regarding criminal justice system contact focus on charges and convictions. Even though the study itself is not focused on criminal justice, these factors can influence entry into and success in the workforce. According to the mission statement for the Bureau of Justice Statistics (BJS), its mission is: “To collect, analyze, publish, and disseminate information on crime, criminal offenders, victims of crime, and the operation of justice systems at all levels of government. These data are critical to federal, state, and local policymakers in combating crime and ensuring that justice is both efficient and evenhanded” (BJS, 2016). BJS provides data collections related to crime, courts, corrections, victims, and law enforcement in addition to other topics. The National Crime Victimization Survey (NCVS) is a widely used source of information associated with the Bureau of Justice Statistics. Data for the NCVS is collected through a self-report survey in which subjects answer questions related to victimization incidents within the last six months, including the number of incidents, crime characteristics, and reports of crime to police. For certain types of crime, such as sexual assault, there

are differences between the crimes reported to the police and crimes not reported to police. Criminal justice data can also be available through state and local agencies for secondary analysis. These data may be collected for the purpose of providing information to the public, but they will largely include information for the management of day-to-day operations. For example, the Bureau of Research and Data Integrity at the Florida Department of Juvenile Justice is tasked with gathering and analyzing data to provide information on juvenile delinquency trends, the root causes of juvenile delinquency, and assessments of juvenile delinquency prevention, intervention, and treatment strategies. This information is also valuable to researchers who are interested in juvenile justice issues. State and local agencies tend to be more topic-specific in the data that are collected and made available compared to the federal agencies that may collect data on a variety of topics.

FROM THE REAL WORLD Janku and Yan (2009) conducted a study to determine whether objective assessment of the risks and needs of court-involved youth leading to judicial processing is less vulnerable to actual or perceived racial

discrimination. Data for the study was acquired from the automated Missouri Justice Information System (JIS), which is maintained by the Office of the State Courts Administrator. “Because of the wide differences across circuits in court location, court culture, and caseload size, we chose one midsized circuit in Missouri for this study to avoid potential noise created by those differences. The rationale for selecting this particular circuit is threefold: First, the caseload contains equal proportions of Caucasian and African American juveniles; second, the circuit contains a mix of metropolitan and rural areas; and third, the circuit has been on the JIS long enough to provide relatively reliable data” (p. 405). The researchers found that African Americans were overrepresented in the juvenile justice system.

Data for secondary analysis may also be available through research groups that are not tied to a specific government agency. NORC (formerly the National Opinion Research Center) at the University of Chicago conducts the General Social Survey (GSS). Data is collected through face-to-face interviews. Between 1972 and 1994, data were

collected annually, but now data are collected every other year. The purpose of the GSS is to collect and analyze data related to the opinions, attitudes, and behaviors of adults living in the United States. Information is available on a variety of topics, such as attitudes towards capital punishment. Individual researchers may collect data for their own purposes. They may be willing to share the data (upon request) with other researchers. The original researchers may decide to make the data available through repositories, such as the Inter-university Consortium for Political and Social Research (ICPSR) associated with the University of Michigan. The ICPSR collects and manages data for the purpose of making it available for secondary research. The amount of data available in depositories can be overwhelming to new researchers; it is extremely useful to have a specific idea or topic in mind when reviewing the data available. As valuable as criminal justice data available for secondary analysis can be, a disadvantage is that it can be more difficult to establish validity and reliability, which is based on the summary information provided by the original researchers. Another disadvantage is that data may not be

available for a specific topic of interest. It is important to keep in mind that you cannot manipulate the available data or variables to make them fit your study. Not only will the results be questionable if not outright useless, but this type of manipulation is also unethical (Chapter 2).

Content Analysis Content analysis is the study of social artifacts to gain insights about an event or phenomenon. Even though it uses unobtrusive methods, it differs from the previously discussed use of secondary data. Secondary data analysis is focused on data that was collected for management and research purposes on a specific topic. Content analysis focuses on the coverage of an event or phenomenon by the particular medium being evaluated (books, magazines, television programs, news coverage, and so forth) rather than the event or phenomenon itself. It is a favorite strategy among researchers who wish to compare social events from different eras. Data for content analysis is often found in two types of records: public and private. Public records include actuarial, political, judicial, and other governmental documents and publications and broadcasts from mass media. Private records include diaries, letters,

and autobiographies. One of the biggest problems with private records is proving the authenticity of the data. Data collection for content analysis allows researchers to be less intrusive than other methods and usually offers sufficient data to assess the phenomenon even if it can no longer be observed. One of the benefits of content analysis is that there are many possible sources of information; however, access or even knowing the data source exists can create problems. This obstacle can be overcome by making the right inquiries and making the time to wade through the bureaucratic processes. Depending on how the research is conducted, content analysis may be either qualitative or quantitative in nature. Qualitative content analysis emphasizes verbal rather than statistical analysis of various forms of communication. For example, Klinger and Brunson (2009) conducted a content analysis of detailed accounts by police officers of how they perceived what transpired during incidents where they shot citizens. When engaging in quantitative content analysis, it is recommended that researchers try to create a quantifiable instrument to use. This strategy requires serious thought about the issue being researched and the creation of a code sheet or checklist to record observations. A summary sheet or form that permits

the code sheets to be easily totaled is also recommended. These summary forms can then be used to input data for analysis. For example, terrorism is considered a priority topic. However, the media coverage tends to wax and wane based on recent terrorist events. There is a surge of media coverage soon after events, but then the media coverage focusing on terrorism usually drops substantially within a few days or weeks. If a researcher is interested in quantitative analysis of the media coverage of terrorism, then a code sheet could be created that keeps track of specific information, such as the number of mentions or the emotional intensity of news coverage. Emotional intensity could be evaluated based on a 10-point scale with 1 being the least emotionally intense and 10 being the most emotionally intense. In contrast, a qualitative content analysis would focus on describing the characteristics of the emotional intensity and the possible impacts on viewers. There are a variety of methods for collecting data. Ultimately, it is the task of the researcher to be able to properly choose one or more methods that provide the best access to the required data (Table 10-2). When choosing a research design, it is essential to keep the data collection process in

mind. If you are not able to actually collect the data you need or will not have access to the data, then you will not have a successful research study. TABLE 10-2

Survey

Interview

Observation

Secondary

Wide variety of

Wide variety of

Actual

Wide variety of

topics

topics

observation

topics

Advantages

of behaviors Simple to

Can give

Ability to link

administer

complex

behavior to

answers

concept

Clarifications

Can be

Cost effective

Simplicity

Cost effective

anonymous Anonymous

Increase

Anonymous

response rates Comparable

Nonverbal cues

Comparable

Disadvantages Misinterpretation

Misinterpretation

Observer

Misinterpretation

influence Nonresponses

Interviewer

Current

Researcher

biases

behaviors

biases

only Low return rates

Costly

Costly

Limited to availability

No clarifying

Unobservable

No clarifying

behaviors Limited answers

Limited answers

© Jones & Bartlett Learning.

Summary Data can be collected using a variety of methods. Survey research is the most popular method in criminal justice and criminological research. Surveys are often mailed but increasingly are also electronically sent using the Internet. Data can also be collected through interviews, which can be conducted in person (face-to-face) or over the telephone. The interview may be structured, semistructured, or unstructured depending on the level of control within the design. Other methods of collecting data include field observation, analysis of secondary data, and content analysis.

APPLICATION EXERCISES 1. As an employee of a violence prevention program, you would like to do a study on the topic of school

bullying. When you take the idea to the program directors, they ask you to expand the topic to workplace and school bullying and to provide different strategies for the collection of data. How would you begin? 2. Describe how data could be collected using surveys, and provide the advantages and disadvantages of this strategy. a. Describe how you would increase the response rate for a mail survey. b. Describe how data could be collected using the Internet. 3. Describe how data could be collected using interviews (structured, semistructured, and unstructured), and provide the advantages and disadvantages of this strategy. 4. Describe how the data collection process for interviews would differ between face-to-face interviews and telephone interviews. 5. Describe how data could be collected using the different strategies for field observation (full participant, participant

researcher, researcher participant, and complete researcher), and provide the advantages and disadvantages of this strategy. 6. Describe how data could be collected using secondary data, and provide the advantages and disadvantages of this strategy. a. Describe the possible source of data available for secondary analysis on bullying.

RESEARCH EXERCISES 1. Using the hypothesis you developed in Chapter 7, determine the best data collection strategy for this hypothesis and explain your choice. 2. Describe in detail how you would collect the data to test your hypothesis. 3. Find a possible source of existing data related to your hypothesis. It may be necessary for you to look for a more general topic related to your hypothesis.

SECTION III: Final Steps

CHAPTER 11: Data Preparation and Analysis What You Should Know! At this point, all the necessary steps have been established for finding a topic, developing a hypothesis, selecting a research design, and collecting data. The next step is to prepare the data for analysis and actually analyze the data. This analysis is where researchers will start to see the final results of their hard work. Chapters 11 and 12 focus on quantitative analysis, which is used more often than qualitative analysis in criminal justice and criminological studies. After completing this chapter, the reader should be able to: 1. Discuss the importance of data preparation. 2. Describe what occurs during the data coding process. 3. Describe the data-cleaning process. 4. Compare the different methods of dealing with missing data. 5. Explain why data may need to be recoded. Provide an example of how recoding may be

6.

7.

8.

9.

10.

11.

12.

done. Identify and describe the three types of data analysis, including univariate, bivariate, and multivariate. Recognize and explain the three types of statistical analysis, including descriptive, comparative, and inferential. Present and discuss the four types of frequency distributions, including absolute frequency, relative frequency, cumulative frequency, and cumulative relative frequency. Describe the various methods of visually presenting data in addition to frequency tables, including pie charts, bar charts, histograms, polygons, and line charts. Discuss the three measures of central tendency, including mode, median, and mean. Discuss the three measures of variability, including range, variance, and standard deviation. Discuss what is meant by the terms skewness and kurtosis.

The first two sections of this textbook have provided a wealth of information regarding how to conduct research. Now it is time to address what to do with that data once they are collected. Following data

collection, the next step is to prepare the data, which must occur before the data can be analyzed. This chapter is not intended to teach students how to perform statistical analyses or serve as a replacement for a statistics class. The purpose of this chapter is to provide a general overview of preparing and analyzing data. Keep in mind that whenever referring to data in research, it should be used as a plural (datum is the singular version of data).

Data Preparation Taking the time to properly prepare data for analysis is a vital step in the research process. Nevertheless, there is a tendency for researchers, even experienced researchers, to want to jump straight to the analysis step once they have the data. Will the results support the hypothesis? Has something new been discovered? Are these findings consistent with previous research? These are examples of questions one hopes to answer through the data analysis. However, not preparing the data for analysis would be like cooking with dirty utensils: You would have a meal at the end, but would you really want to eat it? The first set of tasks in data preparation generally consists of data coding, data entry, and data cleaning. After these tasks have

been completed, then issues related to missing data and recoding of data can be addressed.

Data Coding Coding data refers to the process of assigning values to the data for quantitative statistical analyses. The first step in coding is to determine the variable’s level of measurement. In Chapter 6, we discussed the four levels of measurement: nominal, ordinal, interval, and ratio. In the data analysis process, interval-level variables are usually grouped with and treated as ratio-level variables. Keep in mind the questions for determining a variable’s level of measurement. Are scores or categories the possible outcomes of a variable? If the possible outcomes are scores, then the variable is ratio. If the possible outcomes are categories, then the next question is whether the categories have an order. If the categories have an order, then the variable is ordinal. If the categories do not have an order, then the variable is nominal. Not all data require coding. Since ratio-level variables are based on scores, they can be left as is (e.g., age or income scores). Nominal- and ordinallevel variables are based on categories and require coding. For example, marital status (married, widowed, divorced, separated, never married) is a

nominal-level variable. During the coding process each category is assigned a number, such as 1 for married, 2 for widowed, 3 for divorced, 4 for separated, and 5 for never married. Since marital status is a nominal-level variable, the order does not matter. However, the order is important when coding ordinal-level variables. For an ordinal-level variable such as socioeconomic status (low, middle, high), the lowest number should be associated with the lowest category. An example of coding for socioeconomic status would be 1 for low, 2 for middle, and 3 for high. It is usually easier for researchers if the coding scheme is developed in advance. Precoding refers to a coding scheme that has been incorporated into the questionnaire or observational checklist. Such a data collection device may enable the researcher to enter data directly from the instrument rather than having to code the data first. The field research checklist in Chapter 10 (Figure 10-1) provides an example of precoding. Specifically, this form of precoding is known as edge coding, in reference to the responses lined up in the margin, which simplifies the data entry process. During the coding process, qualitative data may also be converted into quantitative data. For example, an

open-ended question about prior arrests may later be converted into numerical groupings based on the number of arrests, type of offenses, seriousness of the charges, conviction versus acquittal, or other logical groupings created by the researcher. Remember the previous warning about false precision when doing such a conversion. Be certain that there is an explainable logic to the numerical assignments. Ultimately, coding decisions are up to the individual researcher, and researchers will develop their own style over time. One piece of advice is to keep the coding process as simple and straightforward as possible to avoid confusion. To ensure clarity, researchers should create codebooks or coding keys to show details related to the variables, such as the question (if part of a survey), level of measurement, and other coding details. Figure 11-1 is an excerpt from the codebook for Dantzker’s (2010) study of police pre-employment psychological screening.

Data Entry Once the data are coded, they can then be entered into a computer software program dataset for analyses. Figure 11-2 is an excerpt of the coded data after they have been entered into a dataset. In

this figure, each column represents a single variable and each row a single respondent. For example, if marital status was entered into a dataset, then the numbers 1–5 would be displayed in the marital status column. If respondent on line 1 was divorced, then 3 would be displayed on line one in the marital status column.

Figure 11-1 Code Book Excerpt from Prescreening of Police/Law Enforcement Candidates’ Psychologist Survey

Source: Reproduced from Dantzker, M. L. Psychologists’ role and police pre-employment psychological screening. ProQuest Company, 2010.

Figure 11-2 Coded Data Excerpt. Source: Reproduced from Dantzker, M. L. Psychologists’ role and police pre-employment psychological screening. ProQuest Company, 2010.

At various points in the text, it has been suggested that a computer can assist in conducting different aspects of the research process. The same is true for the analysis portion. Several statistical software packages are available to researchers, such as Excel, QuattroPro, Statistical Package for the Social Sciences (SPSS), and Statistical Analysis System. Each has its own quirks and specialties that require individuals to choose what best suits their research needs and abilities. Regardless of which statistical software package is chosen, each allows for the entry, analyses, and storage of data. The authors recommend SPSS, which is available for personal computers after years of only being accessible through mainframes. This program provides the ability to conduct almost every type of statistical technique required, facilitated through a “point and click” interface that is less intimidating to new users than other statistical software packages. Before choosing one of the packages, researchers should examine their research design to determine which package will work best for their needs. Once the choice is made, it is simply a matter of learning how to use it and being able to interpret the results. Most statistical packages not only have handbooks but also excellent tutorials.

The key to data entry is accuracy. Typically, an individual will read and type the information from the data collection device (e.g., survey) into a dataset. Another possible strategy is to have a reliable person read the data aloud as it is entered into the computer to avoid having to continuously switch one’s viewing from the code sheet to the computer screen. This process helps prevent mistakes. However, researchers should always review what has been entered to make certain that it is accurate. Data entry can be time consuming and tedious. For many researchers, it is their least favorite part of the entire research process. It is not unheard of for university professors to assign graduate students this task rather than completing it themselves. However, there are several strategies that can ease the data entry process. The use of the Internet in data collection for surveys has greatly simplified the data entry process since the data is typically entered directly from the survey into a dataset, which can then be uploaded to one of the statistical software packages. If a telephone survey was conducted, it may have been possible to enter the data directly into the computer as the questions were asked. Another shortcut is to use optica-scan sheets, such as the Scantron sheets used in taking multiple-choice examinations. This strategy permits

the data to be entered directly from the sheets marked by the respondents. Regardless of which technique is used, the data need to be cleaned before analysis can begin.

Data Cleaning Data cleaning is the process of examining and reviewing the data for errors (Adler & Clark, 2007; Frankfort-Nachmias & Leon-Guerrero, 2008; Shaughnessy, Zechmeister, & Zechmeister, 2008). As in data entry, data cleaning can be tedious but is a crucial important step since much of the data entry in research is still done by hand. Mistakes that might have occurred during the initial recording of data or data entry are corrected, resulting in data that have been “cleaned.” The first step in data cleaning is reviewing the data that has been entered for accuracy. If a researcher is using a computer program that is programmed to check for errors, it may either beep or refuse to accept data that does not meet the coding requirements for that variable. This process is referred to as automatic data cleaning. For example, assume that a Likert scale was used where 1 represents strongly agree, 2 agree, 3 neither agree nor disagree, 4 disagree, and 5 strongly disagree. If the researcher tries to enter a

value of 6, the program may beep as an alert of the error. Not all statistical software packages have this automatic function. Therefore, cleaning must be done by the researcher. In addition to visually checking the data, creating a frequency table for each variable can be helpful. In a frequency table, all scores and categories are displayed for the selected variable, which allows the researcher to quickly see if there are errors. Using the frequency table, any incorrectly entered value should be easily observable. Another type of data cleaning is contingency cleaning. In this technique, review of the data is expedited by the knowledge that certain responses should only have been made by certain individuals. Not all females would indicate having given birth to a child, but no males should have indicated such. As one is reviewing the data, there are certain questions that logically lead to similar responses on following questions. If they do not match the expectation, then that particular questionnaire or form should be reviewed to check the accuracy of what was entered. After the data have been cleaned, the next step is to determine how to handle missing data.

Missing Data It is not unusual for data to be “lost” during the data collection process. Missing data may be caused by an oversight in data entry, which can usually be easily corrected during the data cleaning process by obtaining the right information from the survey form or code sheet. It also may occur because the respondents accidentally overlooked or deliberately chose not to answer a particular question. One strategy to decrease the likelihood of missing data from occurring is to provide an “escape” response, such as “no response” or “does not apply.” These responses will not provide information applicable to the actual variable but are preferable to having missing data. If missing data does occur, there are several accepted strategies for handling this issue. If the data have been left out on a single question, researchers may choose to enter it as a nonresponse. To do so, a value that is not being used within the variable itself is selected to represent a nonresponse. For example, the marital status variable has valid categories numbered 1 through 5. Therefore, 0 could be used to indicate a nonresponse. However, researchers frequently use a value that differs from the valid responses so that the nonresponse is not confused with the valid responses. Instead of 0, researchers may have

used 99 (assuming the researchers are not using continuous variables in which 99 could be a possible response) to indicate a nonresponse. This strategy is the preferred method of dealing with nonresponses. Another option for dealing with nonresponses is to assume that the missing data are caused by an oversight rather than an intentional omission. In this situation, researchers might look at the other responses to try to determine what the missing response would most likely have been. For example, if on related questions a respondent had indicated support for strict law enforcement, one might assume a similar answer on the unanswered question and enter this information into the dataset. Additionally, if the instrument used is a Likert scale, then the nonresponse can be classified as a “neither agree nor disagree” or “do not know” (depending on how the scale is worded). Although some researchers use these strategies, it is not recommended because it can lead to challenges regarding the objectivity and validity of the analysis. Yet another option is to exclude from the analysis the data collection instrument containing the omission. If respondent 1 did not answer one of the questions, then all of his or her responses are

eliminated from the analysis. This option may be appropriate if respondent 1 did not answer multiple questions. However, if only one or two questions are not answered, then this solution can lead to the loss of worthwhile information.

Recoding Data Depending on the goal of the analysis, it may be necessary to recode certain variables. For example, using a ratio-level age variable to create a frequency table or a cross-tabulation (Table 11-1) may result in a table too large to be useful due to the large number of possible scores. A comparison of age by income could result in a table that is lengthy and confusing. Assuming that age was presented in columns and income in rows, if the people surveyed ranged from 20 years of age up to 70 years of age, and if their incomes ranged from $10,000 to $100,000, the resulting table would be enormous. Because each individual year of age would have to be represented, there would be 51 columns in this hypothetical table. Because there would likely be as many different incomes as there were respondents, the table could end up with thousands of rows. By collapsing the categories into logical groupings, the data can be presented clearly and concisely (Table 11-1).

TABLE 11-1 Hypothetical Age by Income Comparison

(Ordinal Data)

20–

30–

40–

50–

60–

70+

Age

29

39

49

59

69

50

30

20

10

05

15

100

80

50

30

50

80

80

100

80

60

80

60

50

100

160

120

100

80

30

80

120

150

120

50

20

50

100

120

100

40

10

40

80

100

80

30

05

30

60

80

60

20

05

20

40

60

40

10

Income $10,000 to $19,000 $20,000 to $29,999 $30,000 to $39,999 $40,000 to $49,000 $50,000 to $59,000 $60,000 to $69,000 $70,000 to $79,999 $80,000 to $89,000 $90,000 to $99,000

$100,000+

00

10

30

50

30

10

© Jones & Bartlett Learning.

The scores or categories within an existing variable can be grouped to create a new variable. A ratiolevel variable can be recoded into a nominal- or ordinal-level variable. For example, scores in a ratio-level age variable (e.g., 28, 30, 45, 62) can be recoded into an ordinal-level variable with three categories: less than 30, 30–50, greater than 50. An ordinal-level variable can be recoded into a different ordinal-level variable or into a nominal-level variable. A nominal-level variable can only be recoded into a different nominal-level variable. For example, a nominal-level variable based on vehicle type may have four categories: cars, trucks, SUVs, and vans. This variable could be recoded into a different nominal-level variable based on whether the vehicle is a car with the possible response categories of yes or no. Scores and categories can be grouped to create a new variable, but existing categories cannot be subdivided to create a new variable.

Data Analysis

Now that the data have been entered into a dataset and cleaned, analysis of the data may begin. There are three types of data analyses: (1) univariate, (2) bivariate, and (3) multivariate. With univariate analysis, one variable at a time is examined (Frankfort-Nachmias & Leon-Guerrero, 2008; Gavin, 2008; Walker & Maddan, 2009). Descriptive statistics are used to analyze individual variables and may include frequencies and measures of central tendency and variability. Descriptive statistics are discussed in detail later in this chapter. Bivariate analysis is when the relationship between two variables is examined. This examination may be comparative or inferential depending on the nature of the study. If a researcher is exploring the relationship between the two variables, then comparative statistics are used. Comparative statistics usually involve an analysis of the characteristics of the two variables to better describe the relationship between them. Even though the relationship between two variables is being examined, the variables themselves are not being labeled as independent and dependent variables. In other words, comparative statistics does not try to indicate the influence one variable has on another variable. Depiction of crime rates over time, noting the percentage of change in

criminal activity, and trend analyses are examples of comparative statistics. Comparative statistics exist in a gray area between descriptive and inferential statistics and are categorized as a separate type of statistics here, according to the authors’ practice. If the hypothetical Table 11-1 was real, it would be an example of comparative statistics. If one of the variables is identified as being dependent and the other variable is identified as being independent, then the analysis of their relationship becomes inferential. Inferential statistics indicates that the researcher is trying to use the knowledge gained from the inferential statistics to make predictions or inferences about the relationship between the two variables. Specifically, inferential statistics are used to predict the outcome of the dependent variable based on the influence of the independent variable. The final type of statistical analysis is multivariate analysis. Multivariate analysis is the examination of three or more variables. This technique is inferential in nature in that descriptive and comparative statistical analyses (through univariate and bivariate analyses) of the data have already been conducted and now the focus is on examining the relationships among several variables. From this examination, a researcher is trying to develop explanations for the

observed relationships. Inferential statistics are covered in detail in Chapter 12.

Statistical Analysis Statistics are data presented in a manner that best represents what it is the researcher wants to discuss. Statistical analyses might best be viewed as processes for problem solving (FrankfortNachmias & Leon-Guerrero, 2008; Gavin, 2008; Walker & Maddan, 2009). Statistics are used in criminal justice and criminology to help describe a variety of associated aspects, such as crime rates, number of police officers, or prison populations. In addition, statistics can be used to make inferences about a phenomenon. As discussed in the preceding section, there are three types of statistics: (1) descriptive, whose function is describing the data; (2) comparative, whose function is to compare the attributes (or characteristics) of two variables; and (3) inferential, whose function is to make an inference, estimation, or prediction about the data. The remainder of this chapter focuses on providing an overview of the processes and terminology of descriptive statistics. Understanding the basic principles is important so that one can comprehend the work of other researchers.

Descriptive Statistics When researchers are interested in knowing selected characteristics about their sample, they require some form of descriptive statistics. One of the most common descriptive statistics is the use of frequencies. When data are first collected, they are referred to as raw data. In other words, they have not been prepared and are not neatly organized. A first step to organizing the data, after coding and entry, is through a frequency distribution, which is commonly used to describe sample characteristics. The descriptive data in Table 11-2 come from Dantzker’s (2010) study. These data describe the sample’s characteristics of interest.

Frequency Distributions Frequency distributions simply indicate the number of times a particular score or characteristic occurs in the sample, which can be reported in whole numbers and percentages. There are four types of frequency distributions: (1) absolute, (2) relative, (3) cumulative, and (4) cumulative relative (FrankfortNachmias & Leon-Guerrero, 2008; Gavin, 2008; Walker & Maddan, 2009). Each offers a statistically sound indication of the sample’s composition. Absolute and relative frequencies are the most reported. It is recommended that the type that

seems to be the most desirable for reporting the data contained in the research be selected. TABLE 11-2 From the Real World: Sample Demographics

Police

Clinical

psychologists

psychologists

11(31)a(50)b

11 (28) (50)

Total

Gender Female

22 (29) (100)

Male

25 (69) (47)

28 (72) (53)

53 (71) (100)

Total

36 (100) (48)

39 (100) (52)

75 (100) (100)

Years providing police services 10 years

20 (55) (44)

25 (65) (56)

45 (59) (100)

Total

37 (100) (47)

39 (100) (53)

76 (100) (100)

Agency types served Municipal

5 (14) (29)

12 (32) (71)

17 (23) (100)

County

0

7 (18) (100)

7 (9) (100)

State

1(3) (25)

3 (8) (75)

4 (5) (100)

Combo

31 (83) (66)

16 (42) (34)

47 (63) (100)

Total

37 (100) (49)

38 (100) (51)

75 (100) (100)

aColumn bRow

Percentage

Percentage

Source: Reproduced from Dantzker, M. L. Psychologists’ role and police pre-employment psychological screening. ProQuest Company, 2010.

Absolute frequency distributions (also known simply as frequencies) display the data based on the number of times a score or category appears. Relative frequency distributions (also referred to as percentages) are the percentage equivalent of absolute frequency distributions and are determined by dividing the absolute frequency by the total number of cases. When dealing with large numbers, it is easier for readers to interpret percentages than raw numbers. Cumulative frequency distributions enable readers to see what the products of the grouping are in the frequencies table by adding the absolute frequency of each previous category or score. Cumulative relative frequency distributions do the same as cumulative frequency distributions but add relative frequencies rather than numbers. Table 11-3 demonstrates how the four types of frequencies are related.

Displaying Frequencies Frequencies are just one means of describing the data. Other means include pie charts, bar graphs, histograms and polygons, line charts, and maps. As demonstrated in Figure 11-3 (a pie chart), the frequencies or percentages can be depicted in a simplified picture form. Such a display is clear and easily interpreted. Bar charts are also easily interpreted (Figure 11-4). Pie and bar charts are

used to display nominal- and ordinal-level data. Histograms are visually similar to bar graphs but are used to display interval- and ratio-level data (Figure 11-5). To indicate the continuous nature of the variable, the bars are adjacent to each other. In a bar chart, there is space between each bar, which indicates a categorical rather than continuous variable. Polygons display the same data as histograms but use dots instead of bars. Lines are then drawn between the dots to reveal the shape of the distribution. Figure 11-6 is an example of a frequency polygon. TABLE 11-3 Absolute, Relative, Cumulative, and Cumulative Relative Frequency Distributions of Targeted Sample of Psychologist Members of APA by Region

Absolute

Relative

Cumulative

Cumulative

frequency

frequency

frequency

relative frequency

EN

531

14.7

531

14.7

176

4.9

707

19.6

672

18.6

1379

38.2

Central ES Central Mid Atlantic

Mountain

210

5.8

1589

44

New

351

9.7

1940

53.8

Pacific

607

16.8

2547

70.6

South

650

18

3197

88.6

221

6.1

3418

94.7

191

5.3

3609

100

3609

100

England

Atlantic WN Central WS Central TOTAL

Source: Reproduced from Dantzker, M. L. Psychologists’ role and police pre-employment psychological screening. ProQuest Company, 2010.

In addition to these frequency-presentation techniques, criminal justice and criminological researchers may also use line charts. Line charts are polygons that demonstrate scores across time (Figure 11-7). As such, they are useful to reflect changes over time in the same dot-and-line format as frequency polygons.

In addition to the tables, graphs, and charts discussed previously, there are four other ways to describe the properties of the data: (1) measures of central tendency, (2) measures of variability, (3) skewness, and (4) kurtosis.

Measures of Central Tendency Because frequencies can be quite cumbersome, researchers sometimes require a way to summarize the data in a simpler manner. Measures of central tendency are one way to summarize data. The three most common measures are the (1) mean, (2) median, and (3) mode. The mean is the arithmetic average. The median is the midpoint or the number that falls in the middle. The mode is the number that occurs most frequently. The mean is usually used as a measure of central tendency for interval- or ratio-level data. The median is used mostly for ordinal-level data. The mode is generally used for nominal-level data. Figure 11-8 shows how the three measures are obtained.

Figure 11-3 Pie Chart. © Jones & Bartlett Learning.

Figure 11-4 Bar Chart. © Jones & Bartlett Learning.

Figure 11-5 Histogram. © Jones & Bartlett Learning.

Figure 11-6 Polygon. © Jones & Bartlett Learning.

Figure 11-7 Line Graph. © Jones & Bartlett Learning.

Using a measure of central tendency allows the researcher to simplify the numbers in a summary manner. In addition, these measures are often used in conjunction with measures of variability. Gaining a familiarity with the characteristics of these measures can help researchers identify when it might be appropriate to use measures of central tendency.

Measures of Variability Despite their similarities, all statistics are not the same. Differences that occur among scores in a variable is called variability. The three main measures of variability are (1) range, (2) variance, and (3) standard deviation. The range is simply the

difference between the highest and lowest scores. Variance is a summary measure of the differences between each of the scores and the mean. Standard deviation is also a summary measure of the differences between each of the scores and the mean, but it has been calculated to be the same unit of measurement as the original variable. For example, if the original variable was age measured in years, then the standard deviation would also be measured in years, making it easier to interpret than the variance.

Figure 11-8 Measures of Central Tendency. © Jones & Bartlett Learning.

Skewness The discussion thus far has alluded to the distribution of the data and has depicted it somewhat in sample figures. Many of the statistical

techniques discussed in this chapter (and which will influence the discussions of inferential statistics in Chapter 12) are based on the assumption of a normal distribution. A normal distribution may also be referred to as a normal, or bell, curve. You may remember cutting hearts from paper as child for Valentine’s Day. To do so, you folded the paper in half so that both sides of the heart would be identical or mirror images. If you were to cut a normal curve in half at the center point, which is the mean, then the sides would be mirror images of each other. By using polygons, line charts, or scatterplots (discussed in Chapter 12), researchers are able to visualize the distribution of the data. If it is a normal distribution, the researcher is able to use a broader range of statistical techniques. If it is a non-normal (also known as nonparametric) distribution, the statistical techniques that may be used are more limited. The measures discussed previously (central tendency and variability) indicate the location of the center of the curve (central tendency) and the spread of the curve (variability). They may also differ based on the symmetry of the curve. If one side has more values so as to cause the slope of that side to stretch further outward and no longer create a mirror image, the distribution is said to be

skewed. Skewness alerts the researcher to the presence of outliers, which are cases that have extremely high or low values. For example, if a line chart based on net worth included individuals such as Bill Gates and Warren Buffett, their fortunes would skew the data distribution. Gates and Buffett are the outliers in this data and pull the distribution to one side. Such knowledge aids the statistician in conducting analysis of the data.

Kurtosis Kurtosis is another distribution consideration that warrants some discussion. Kurtosis refers to the amount of smoothness or pointedness of the curve. A tall, thin curve is described as being leptokurtic. A short, flat curve is platykurtic. For the purposes of this text, one need only recognize the terms if a statistician states that the data distribution exhibits such features. These statistics, in tandem with measures of central tendency and variability, provide useful ways to describe the data. In general, descriptive statistics are primarily used to describe how the data are comprised. However, describing the data is usually just a small portion of making inferences from it.

Summary

Once the data collection is accomplished, the next two phases in completing the research are data preparation and data analysis. Because the data are often obtained in a raw manner, they must be coded for entry into a statistical software program. After the data have been entered, they must be cleaned to ensure accuracy and appropriateness for subsequent statistical analysis. At this point, missing data are dealt with, and recoding may also be required. Univariate analysis using descriptive statistics is typically conducted to gain knowledge about the data before any bivariate and multivariate analyses are completed. Descriptive statistics may involve the use of frequency distribution tables and measures of central tendency and variability. Inferential statistics used in bivariate and multivariate analyses are discussed in more detail in Chapter 12.

APPLICATION EXERCISES 1. In Chapter 10, you described different data collection strategies for a study of workplace and school bullying. Of the ideas you presented to the directors of the violence prevention program, they approved data collection through the use of a mail survey. After receiving

2.

3. 4. 5. 6.

thousands of responses, you now need to analyze the data. The program directors have asked for results from the raw data. Explain the importance of first preparing the data for analysis. One of the questions on the survey asked respondents how often they experience bullying by selecting a specific response from the following: did not experience bullying, once a month, 2–3 times per month, 4–5 times per month, and more than 5 times per month. Describe how you would code this information into a quantitative variable and identify the level of measurement. Describe the process you would use to enter the data into a dataset. Describe the process for cleaning the data. Describe how you would address the issue of missing data. Describe how you would recode the variable from step 2 into a nominallevel variable that indicates the respondent’s experience with bullying and has the possible response categories of yes and no.

7. Describe the possible uses of univariate, bivariate, and multivariate analyses within the study. 8. Draw an example of each of the following methods for displaying frequencies: a. Pie chart b. Bar chart c. Histogram d. Line chart

RESEARCH EXERCISES 1. Using the search procedures you developed in Chapter 3, locate an article or government report that uses quantitative data and one of the previously identified methods for displaying frequencies. 2. Describe the methods for displaying frequencies, such as pie or bar charts, used within the article or government report. 3. Describe the measures of frequencies and measures of central tendency and variability used within the article or government report.

CHAPTER 12: Inferential Statistics What You Should Know! The purpose of inferential statistics is to acquire knowledge that can be used to make inferences or predictions about the data. The type of inferential statistic used largely depends on the variables’ levels of measurement within the analysis. After completing this chapter, the reader should be able to: 1. Differentiate between descriptive, comparative, and inferential statistics. 2. Discuss measures of association, including lambda, gamma, and Pearson’s r, and provide examples. 3. Explain what is meant by statistical significance, and describe how tests of significance are used. 4. Describe the commonly used comparative statistics techniques, including crime rates, crime-specific rates, percentage change, and trend analyses.

5. Discuss bivariate analysis and provide examples. 6. Discuss multivariate statistics, and provide an example of a multivariate technique for nominal- and ordinal-level data. 7. Describe the various multivariate techniques for ratio-level data.

Statistical Analysis The previous chapter discussed the three types of statistics: descriptive, comparative, and inferential. Descriptive statistics describe the data being analyzed. A variety of descriptive statistics were presented in Chapter 11, including measures of central tendency and variability. Comparative statistics, whose function is to compare the attributes of two variables, were also introduced. Comparative statistics are within a “gray area” that moves from description to inference. As such, comparative statistics are often classified as either descriptive or inferential rather than as a separate category. Inferential statistics were briefly touched on in Chapter 11. The purpose of inferential statistics is to make inferences, estimations, or predictions about the data. This chapter shows how comparative statistics may be used to begin making

inferences about the data and provides an overview of inferential statistics. Criminal justice and criminological researchers use a variety of inferential techniques. The selection of the proper inferential technique is based on the characteristics of the data, including the variables’ levels of measurement.

Overview of Inferential Statistics Inferential statistics allow the researcher to develop inferences (predictions) about the data. If the sample is representative (Chapter 5), then these predictions may be extended to the population from which the data were drawn. Inferential statistics allow criminal justice and criminological researchers to conduct research that can be generalized to populations within society. When making inferences about datasets, researchers rely on measures of association to determine the strength and direction (if applicable) of the relationships between or among variables. Tests of significance are also used to determine whether the sample that was examined is representative of the population from which it was drawn.

Measures of Association Measures of association are used to determine the strength of relationships and direction (if

appropriate) among the variables that are being studied. The measure of association that is used is dependent on the type of analysis being conducted, the distribution of the data, and the level of data under analysis. Many measures of association are used in criminal justice and criminological research, but some measures are more commonly used and are considered to be standards, such as lambda, gamma, and Pearson’s r and r2. Lambda is commonly used for nominal-level data. Gamma is commonly used for ordinal-level research. Pearson’s r and r2 are commonly used for intervaland ratio-level data. Measures of association for nominal-level data, such as lambda, tell researchers how strong the relationship is but do not indicate directionality. Lambda is based on a cross-tabulation (Chapter 11) and determines the strength of the relationship. The outcome of lambda will range from 0 (no relationship) to 1 (a perfect relationship). A 0.00001 indicates a very weak relationship. A 0.9999 indicates an extremely strong relationship. Researchers can assess the strength of the relationship as being weak, moderate, or strong. For example, a lambda score of 0.5367 indicates a moderate relationship, whereas a lambda score of 0.8249 is an indicator of a relatively strong

relationship. Since nominal-level variables do not indicate direction, the lambda also does not indicate direction, and researchers have to assess the crosstabulation (also known as a contingency table) to determine the structure of the relationship. For ordinal-level data, measures of association, such as gamma, tell the researcher both the strength and the direction of the relationship. Gamma is based on a cross-tabulation and will range from −1.00 to +1.00. A zero (0) indicates no relationship; a negative one (−1) indicates a perfect negative relationship (as one variable increases, the other variable decreases); and a positive one (+1) indicates a perfect positive relationship (as one variable increases, the other variable increases). A score of −0.8790 indicates a strong, negative relationship, whereas a score of +0.2358 indicates a weak, positive relationship. Measures of association for ratio-level data, such as Pearson’s r and r2, indicate both the strength and direction of the relationship. Keep in mind that for analysis purposes, interval-level data are treated as ratio-level data. The results for Pearson’s r range from −1.00 to +1.00 and are interpreted the same as the results for gamma. Pearson’s r2 is determined by squaring (multiplying the number by itself) and

indicates the percent of variance explained in the dependent variable by the independent variable. A higher percentage indicates a stronger relationship. In addition to the strength and direction of the relationship, statistical significance is also of importance.

Statistical Significance The presence of statistical significance indicates that the sample findings are representative of the population being studied. If a researcher is using a complete enumeration (the entire population is studied rather than just a sample from that population), then determining statistical significance is unnecessary because it is already known that the population is accurately represented. However, complete enumerations are rare because studying entire populations is too costly and time consuming. Statistical significance is based on probability sampling and is used when trying to determine a nomothetic explanation for a phenomenon (Chapter 6). It is this use of probability that enables researchers to make inferences based on relatively few observations. Generally, social science researchers require a statistical significance of 0.05 or better, which indicates that they are 95% confident that the findings from the sample

represent the population, to state that a result is statistically significant.

Comparative Statistics Several comparative techniques are available to criminal justice researchers. Briefly discussed here are crime rates, crime-specific rates, percentage change, and trend analyses. These examples of comparative statistics are widely used within the field of criminal justice and provide the basis for the FBI’s Uniform Crime Reports (Chapter 11).

Crime Rates Crime rates are perhaps the most frequently presented data within criminal justice and criminological research and are simply the number of crimes that occurred in an area (e.g., city, county, state) divided by the population for that area and then typically multiplied by 100,000. The purpose of calculating crime rates is to allow the comparison of the occurrence of crime between places with different population sizes. Crime rates for index crimes (i.e., murder and nonnegligent homicide, forcible rape, robbery, aggravated assault, burglary, motor vehicle theft, larceny-theft, and arson) are commonly displayed in the Uniform Crime Reports. Table 12-1 provides an example of violent crime rates. By viewing the rates presented in that table,

the reader may easily compare the rates among the various states even though the states have different populations.

Crime-Specific Rates Crime-specific rates differ from crime rates in that they use a different base than population within the computations. For example, one could look at motor vehicle thefts by number of registered vehicles, burglaries by number of households, or gun-related crimes by number of registered guns. This method can also be used to calculate victimization rates, arrest rates, and clearance rates. TABLE 12-1 Violent Crime Rates per 100,000 Population for Select States, 2014

State

Population

Violent

Murder and

crime

nonnegligent

Rape

Robbery

manslaughter Arizona

6,731,484

399.9

4.7

50.2

92.8

California

38,802,500

396.1

4.4

29.7

125.5

Hawaii

1,419,561

259.2

1.8

31.3

78

Nevada

2,839,099

635.6

6

47.8

209.7

Texas

26,956,958

405.9

4.4

42.3

115.7

Washington

7,061,530

285.2

2.5

38.2

79.9

Wyoming

584,153

195.5

2.7

29.8

9.1

Source: Data from “Crime in the United States, 2014: Violent Crime,” Federal Bureau of Investigation, 2015.

Percentage Change Percentage-change statistics allow researchers to compare data over time. The computation is straightforward: Subtract the earlier number from the later number, and then divide the difference by the earlier number. This statistic allows one to determine whether there has been an increase or decrease in particular crimes between the two points in time. The bottom of Table 12-2 demonstrates this statistic for three index crimes in the United States for three different sets of years.

Trend Analyses Trend analyses are yet another way of comparing differences over time. Researchers may use a visual representation, such as a histogram, to show how rates have increased for a particular offense. Trend analyses are quite useful in assessing the impact of crime prevention strategies. The reader should refer to Table 12-2, which shows the number of specific crimes from 1998 to 2007. From this table, one can determine what trends may have

existed for these crimes during the reported time frame. TABLE 12-2 Specific Crimes in the United States by Volume, Rate per 100,000, and Percent Changes, 1998–2007

Year

Population

Forcible

Forcible

Robbery

Robbery

Aggravated

rape

rape

volume

rate

assault

volume

rate

volume

1998

270,296,000

93,144

34.5

447,186

165.4

976,583

1999

272,691,000

89,411

32.8

409,371

150.1

911,740

2000

281,421,906

90,178

32

408,016

145

911,706

2001

284,796,887

90,491

31.8

422,921

148.5

907,219

2002

287,973,924

95,235

33.1

420,806

146.1

891,407

2003

290,788,976

93,883

32.3

414,235

142.5

859,030

2004

293,656,854

95,089

32.4

401,470

136.7

847,381

2005

296,507,061

94,347

31.8

417,438

140.8

862,220

2006

299,398,484

92,757

31

447,403

149.4

860,853

2007

301,621,157

90,427

30

445,125

147.6

855,856

Percent change in volume and rate per 100,000 inhabitants for 2, 5, and 10 years

2007/2006

–2.5

–3.2

–0.5

–1.2

–0.6

2007/2003

–3.7

–7.1

+7.5

+3.6

–0.4

2007/1998

–2.9

–13

–0.5

–10.8

–12.4

© Jones & Bartlett Learning.

Inferential Statistics Inferential statistics make inferences, estimations, or predictions about the data. Because of the nature of inferential statistics, to do justice in explaining them requires a text of its own. Extensive explanations are left for statistics courses, but brief descriptions of selected statistics follow.

Bivariate Analysis Bivariate analysis is the examination of the relationship between two variables. Usually this analysis involves attempting to determine how a dependent variable is influenced by an independent variable. The more commonly used bivariate techniques are contingency tables and bivariate (simple linear) regression. Assessing the relationship between two variables addresses the strength of the relationship, the direction of the relationship (if applicable), and the level of significance.

Contingency Tables (or CrossTabulations) With nominal- and ordinal-level data, two popular statistical techniques are contingency tables (crosstabulations) and chi-square, which is a common statistic for a contingency table. A contingency table is a set of interrelated cells that displays the relationship between two variables. Each cell can display a variety of data (Table 12-3), such as frequencies or percentages. From these data, a chisquare (χ2) statistic can be calculated. Chi-square is one test of statistical significance that is popular because it can be used with data at any level of measurement. The contingency tables (crosstabulations) and chi-square measure offer an analysis of the statistical relationship from which one might make an inference.

Bivariate Regression Bivariate regression, also known as simple linear regression, is based on the principle that over time things tend to regress toward the mean. For example, if one were to measure the heights of female students enrolled at a school, one would find that they range from well above average to well below average height for women. Assuming a normal population, most of the heights would tend to cluster around the mean (average) height.

Bivariate regression is appropriate to examine the relationship between two variables when using ratiolevel data with data that are normally distributed (Chapter 11) and have a linear relationship. If two variables have a linear relationship, then as the independent variable (X) increases, so does the dependent variable (Y). Conversely, if a negative relationship exists (e.g., a crime prevention strategy), then as X (the strategy used) increases, Y (the specific crime targeted) is expected to decrease. TABLE 12-3 Example of Contingency/CrossTabulations: Overall Job Satisfaction by Level of Education/Degree

EDUCATION Count

H.S.

Assoc.

Bach. Deg.

Master’s

Exp

Diploma

Deg. 2

3

Deg. 4

Val

1

1

0

0

0

.5

.2

.3

.0

Row Pct Col Pct Tot Pct Overall

2

Neutral

3

4

Extremely

5

100.0%

.0%

.0%

.0%

4.5%

.0%

.0%

.0%

2.3%

.0%

.0%

.0%

11

4

4

1

10.2

4.2

5.1

.5

55.0%

20.0%

20.0%

5.0%

50.0%

44.4%

36.4%

100.0%

25.6%

9.3%

9.3%

2.3%

9

3

6

0

9.2

3.8

4.6

.4

50.0%

16.7%

33.3%

.0%

40.9%

33.3%

54.5%

.0%

20.9%

7.0%

14.0%

.0%

1

2

1

0

2.0

.8

1.0

.1

25.0%

50.0%

25.0%

.0%

4.5%

22.2%

9.1%

.0%

Satisfied

2.3%

4.7%

2.3%

.0%

Column

22

9

11

1

Total

51.2%

20.9%

25.6%

2.3%

Chi-Square

Value

DF

Significance

Pearson’s

5.12525

9

.82326

Likelihood

5.53040

9

.78584

.61069

1

.43453

Ratio Linear-byLinear Association Minimum Expected Frequency—.023 Cells with Expected Frequency < 5 – 13 of 16 (81.3%) Approximate

Value

ASE1

Val/ASEO

Significance

Statistic Phi

.34524

.82326

Cramer’s V

.19933

.82326

Contingency

.32634

.82326

Coefficient © Jones & Bartlett Learning.

Multivariate Analysis

Multivariate analysis is the examination of the relationship between three or more variables. Usually this involves attempting to determine how a dependent variable is influenced by more than one independent variable. This methodology offers more insights than bivariate analysis, in that it is possible to study the relationships among several variables at one time. The more commonly used multivariate techniques are Student t test, correlation, analysis of variance (ANOVA), and multiple regression (Frankfort-Nachmias & Leon-Guerrero, 2008; Gavin, 2008; Walker & Maddan, 2009; Weisburd & Britt, 2007). Again, the purpose here is not to teach how to compute these measures but rather to provide a brief overview of how these tools might be used in criminal justice and criminological research and what to discuss with a statistician if using this strategy.

Student t Test The Student t test is used to compare groups’ means for a particular variable and hypothesis testing. Computing Student t is a fairly complex process that contrasts expected outcomes with observed outcomes. The differences among the means of the variables are then assessed. The Student t indicates whether the relationship between the groups is statistically significant. As seen in

Table 12-4, the differences between the means are obvious. The t-values do not mean anything at this point, but indicate the values used by the computer to calculate the significance.

Correlation Correlation is a commonly used technique for evaluating ratio-level data. The previous discussion of bivariate regression explained how relationships between two variables are examined based on the assumption of a linear relationship. In correlation, relationships are assessed based on covariation. Covariation simply means that as changes occur in one variable, X, there will also be changes in another variable, Y. Assessing correlations is based on how the variation in one variable corresponds to variation in the other variable. The means of assessing the correlation of ratio-level data is Pearson’s r, which is used to determine the strength of the association among variables by dividing the covariance of X and Y by the product of the standard deviation of X and Y. The statistical software packages calculate these numbers. Of interest here is the strength, direction (recall one is looking at a number between −1 and +1), and level of significance. Table 12-5 is an example of the use of correlations in analyzing ratio-level data.

Analysis of Variance ANOVA is another means of examining ratio-level data. Where correlation uses Pearson’s r to determine the nature of relationships among variables, ANOVA uses something known as an F ratio to compare the means of three or more groups. This technique helps avoid committing errors that might occur when using multiple t tests (FrankfortNachmias & Leon-Guerrero, 2008). ANOVA allows statisticians to determine significance by assessing the variability of group means. Therefore, it is a useful method for evaluating grouped data (e.g., the outcomes of correctional treatments on inmate groups). Table 12-6 is an example of ANOVA findings. TABLE 12-4 Student’s t for Student’s Perceptions of Policing Comparison of Selected Means Scores, by Time

Variable

Mean (t1)

Mean (t2)

t-value

Primary Role

.554

–.458

5.49*

Level of Competency

–.747

–1.289

4.29*

Serve and Protect

–1.000

–.800

–1.27

Corrupt Act

–.598

–.390

–1.33

*

Strike a Minority

–.171

–.500

2.37*

Ignore Needs

–1.000

–1.060

.55

Preventing Crime

–.476

–.951

2.97*

Harass or Help

–1.374

–1.470

.96

Unknown Reaction

.482

.716

–1.65

Professionalism

.183

.598

–3.38*

Help Society

.627

.928

–1.84

Benefit of Doubt

.256

.646

–2.39*

* = p