Sociology

Doing Sociology with Student CHIP Data Happy!

Fifth Edition

Gregg Lee Carter Bryant University

Allyn & Bacon

Boston Columbus Indianapolis New York San Francisco Upper Saddle River Amsterdam Cape Town Dubai London Madrid Milan Munich Paris Montreal Toronto

Delhi Mexico City Sao Paulo Sydney Hong Kong Seoul Singapore Taipei Tokyo

Copyright © 2010, 2004, 2001 Pearson Education, Inc., publishing as Allyn & Bacon, 75 Arlington St., Suite 300, Boston, MA 02116. All rights reserved. Manufactured in the United States of America. This publication is protected by Copyright, and permission should be obtained from the publisher prior to any prohibited reproduction, storage in a retrieval system, or transmission in any form or by any means, electronic, mechanical, photocopying, recording, or likewise. To obtain permission(s) to use the material from this work, please submit a written request to Pearson Higher Education, Rights and Contracts Department, 501 Boylston Street, Suite 900, Boston, MA 02116, or fax your request to 617-671-3447. Many of the designations by manufacturers and sellers to distinguish their products are claimed as trademarks. Where those designations appear in this book, and the publisher was aware of a trademark claim, the designations have been printed in initial caps or all caps. 10 9 8 7 6 5 4 3 2 1 13 12 11 10 09 ISBN-10: 0-205-78001-6 www.pearsonhighered.com ISBN-13: 978-0-205-8001-3

iii

CONTENTS

Preface: Doing Sociology with Student CHIP …………………………………………………………………………………… v

About Student CHIP………………………………………………………………………………………………………………………………..1

A Primer on Critical Reading……………………………………………………………………………………………………………….3

A Primer on Elementary Data Analysis …………………………………………………………………………………………………… 9 Measures of Central Tendency ………………………………………………………………………………………………………9  Measures of Dispersion……………………………………………………………………………………………………………….10  Scatterplots and the Correlation Coefficient ………………………………………………………………………………..11  Basic Tabular Analysis ………………………………………………………………………………………………………………..14  Criteria for Establishing Causality in Nonexperimental Situations………………………………………………..16  The Art of Reading Partial Tables…………………………………………………………………………………………………18

Causality…………………………………………………………………………………………………………………………………..18            Spuriousness ……………………………………………………………………………………………………………………………19  Multivariable Model ……………………………………………………………………………………………………………………20            Intervening Variable …………………………………………………………………………………………………………………..21            Interaction Effects ……………………………………………………………………………………………………………………..21

Chapter 1. The Problem of Social Order……………………………………………………………………………………………29 1. Social Order and Control via Close Social Ties: The Example of Suicide ……………………………………30  2. Social Characteristics of Happy Individuals ……………………………………………………………………………….43  3. Trust and the Social Order …………………………………………………………………………………………………………51  Exploratory Exercises……………………………………………………………………………………………………………….59  Answers for Selected Chapter 1 Exercises………………………………………………………………………………..63

Chapter 2. Issues in Sociological Research …………………………………………………………………………………….75  4. Attitudes vs. Actions: Do Religiosity and Church Attendance Go Hand-in-Hand? ………………………75  5. The Idea of Contextual Effects: The Social Context of Working Full-Time for Women…………………81  Exploratory Exercises……………………………………………………………………………………………………………….85

Chapter 3. Culture………………………………………………………………………………………………………………………………….89  6. The Problem of Ethnocentrism ……………………………………………………………………………………………………..89  Exploratory Exercises……………………………………………………………………………………………………………….99

Chapter 4. Society………………………………………………………………………………………………………………………………..103  7. Changes in the Occupational Structure of the United States (Post-Industrial Society)………………103  8. HIV/AIDS—A Crossnational Examination………………………………………………………………………………….109  Exploratory Exercises……………………………………………………………………………………………………………..119

Chapter 5. Socialization………………………………………………………………………………………………………………………123  9. Social Class and Parental Values……………………………………………………………………………………………..123 Exploratory Exercises……………………………………………………………………………………………………………..133

Chapter 6. Groups………………………………………………………………………………………………………………………………..137  10. Physical Health, the Quality of Primary-Group Ties, and Social Class ………………………………………137  11. Psychological Health, the Quality of Primary-Group Ties, and Social Class………………………………147  12. Family Life and Adolescent Academic Success ……………………………………………………………………….153 13. Who’s Most Likely To Be Divorced? ………………………………………………………………………………………..163 Exploratory Exercises……………………………………………………………………………………………………………..171

iv

Chapter 7. Interaction………………………………………………………………………………………………………………………….175  14. The Relationship of Internet Use to Social Background and Emotional Well-Being…………………..175  Exploratory Exercises……………………………………………………………………………………………………………..187

Chapter 8. Crime, Deviance, and Social Control……………………………………………………………………………191  15. Social Class and Deviance (A Test of Anomie Theory) ……………………………………………………………….192  16. Social Integration and Deviance (A Test of Control Theory) ……………………………………………………..197  Exploratory Exercises……………………………………………………………………………………………………………..203

Chapter 9. Inequality……………………………………………………………………………………………………………………………207  17. An Examination of the “Status Attainment” Model and the Predictors of Individual Economic Success…………………………………………………………………………………………..207  18. Up and Down Opportunity’s Ladder—Generational Social Mobility …………………………………………….223  19. Social Class and Political Participation…………………………………………………………………………………….233  20. The Wealth of Nations: Associations Among Economic Development, Political Structure, Age Composition, and Life Expectancy ……………………………………………………………………………245 Exploratory Exercises……………………………………………………………………………………………………………..253

Chapter 10. Race and Ethnicity…………………………………………………………………………………………………………257  21. Ethnic Tolerance: A Function of Social and Personal Control…………………………………………………..257  22. Race, Ethnicity, and Poverty …………………………………………………………………………………………………….263  23. Predicting Minority Success in School—Individual and Contextual Factors ……………………………..273 24. Intergroup Contact and the Reduction of Prejudice ………………………………………………………………….281 Exploratory Exercises……………………………………………………………………………………………………………..289

Chapter 11. Gender ……………………………………………………………………………………………………………………………..293  25. The Malleability of Gender Roles (Are Men Better Suited for Politics?)……………………………………..293  26. Sex Differences in Income………………………………………………………………………………………………………..301 27. The Status of Women: Crossnational Comparisons …………………………………………………………………310 Exploratory Exercises……………………………………………………………………………………………………………..322

Chapter 12. Social Change and Social Conflict……………………………………………………………………………..326  28. The Strategy of Social Protest ………………………………………………………………………………………………….326  29. The International Scene: Ethnic Diversity and Civil War ……………………………………………………………332 Exploratory Exercises……………………………………………………………………………………………………………..339

About the Author………………………………………………………………………………………………………………………………….343

v

Preface

Doing Sociology with Student CHIP

Over the past two decades, those of us believing that an appreciation of how data and theory fit together is crucial to every sociology course— indeed, that it is part and parcel of thinking sociologically and thinking critically—have had much to celebrate. Sociology texts and antholo- gies have become rich in data, and many authors try hard to demonstrate the interplay between sociological insight and the relevant empirical observations. Part of this trend has been the appearance of workbooks and related materials that get students involved in the analysis of soci- ological data as early as their very first introduc- tory sociology course. Doing Sociology with Stu- dent CHIP: Data Happy! is now an established part of the trend. It does, however, improve upon its competitors in at least three ways:

First, Doing Sociology with Student CHIP pro- vides students with a primer for critiquing socio- logical writing. At first glance, this primer may seem biased toward quantitative thinking. A deep- er look, however, reveals that it will carry students not only through quantitative arguments, but through qualitative ones as well; not only through sociology, but through any empirical discipline (from history to physics). In short, critical thinking is critical thinking. Its essentials are universal and are structured to assess proposed answers to the question “Why?” and to suggest alternatives. (Why does the world look the way it does and why do things unfold the way they do? Why do earthquakes happen? Why do people get cancer? Why are some people poor? To what extent do the proposed explanations seem plausible?)

Second, Doing Sociology with Student CHIP provides a primer on elementary data analysis and its connection to the problem of establishing caus- ality. As with the chapter on critical reading, stu- dents should read through the data analysis pri- mer slowly and carefully, and they should do all of the exercises.

Finally, compared to its competitors, Doing Soci- ology with Student CHIP employs a much broad- er range of data sources. The data are of the high- est quality and include the General Social Survey, census and vital statistics reports, FBI crime summaries, and crossnational statistics from the

United Nations, World Bank, and other highly re- spected organizations. The data are used in com- puter exercises organized around the major sub- fields of sociology. Each set of exercises is intro- duced by a brief summary reviewing some major concerns of the particular subfield and, when appropriate, clarifying related concepts and ter- minology.

In sum, Doing Sociology with Student CHIP con- tains dozens of computer exercises that allow stu- dents to use serious data to explore and test the in- sights of sociology. The exercises cover the major subfields of sociology (e.g., gender, race, inequali- ty) and can be adapted for most undergraduate sociology courses. They will hold special interest for empirically oriented instructors and students. The exercises will appeal to instructors trying to bridge the gap between what they do as research- ers and what they teach in the classroom.

Important Features of the Fifth Edition

Many instructors and students who used the first four editions of this workbook were generous in offering suggestions for its improvement. Among the suggestions that are apparent in the new edition:

 The strong emphasis on social problems has been continued. The new edition has been es- pecially designed for the introductory course and for courses on contemporary social prob- lems. Included among its many exercises are those on HIV/AIDS, education, anomie, crime, divorce, physical health, psychological health, minority education, literacy, fertility, hunger, infant morality, immigration, race and ethnic relations, religiosity impacts, separatist movements, social capital and social trust, and the status of women—both in the United States and crossnationally.

 The strong emphasis on international com- parisons, with many of the new social prob- lems being analyzed with crossnational data, has been continued.

 Wherever possible, data sets have been up- dated, including the use of the 2006 General Social Survey, 2000–2008 U.S. census data, the

vi

2005 NCCJ TAP-III survey on race and ethnic relations, and the most recent Human Devel- opment Reports of the United Nations.

 Many professors want their students to have more in-depth reading on the particular topic at hand than is offered in the introductory comments to each chapter of this workbook; to this end, the new edition has been closely coupled to the latest (fifth) edition of Empirical Approaches to Sociology, an edited anthology that provides this in-depth reading. At the end of this Preface (Box 1, p. vii), I have presented a simple table to assist instructors on how they might topically coalesce the 39 essays in Empirical Approaches with Doing Sociology with Student CHIP. However, in general, the table will serve the same purpose for any instructor using a standard introduc- tory sociology or social problems text.

 Because sociological analysis can sometimes seem quite abstract and distant from our everyday interactions (we tend to deal with people on a name basis rather than as “Case- IDs”), the new edition of this workbook main- tains the practice used in earlier editions of labeling all cases whenever aggregate levels are used; thus, students can locate particular states or nations on a particular plot line or see where the aggregate unit fits in a crosstabula- tion (putting it in an exact cell, e.g., being able to put Nevada in the “high divorce” column and “high suicide” row to see its exact location in the table).

 Because the power of interactive data analysis is never fully realized if students are doing only those analyses they are told to do, the new edition of this workbook continues the custom of earlier editions of having Explor- atory exercises at the end of each chapter; these exercises allow students to come up with their own hypotheses and their own interpretations (whether grounded in their readings, their professors’ lectures, or their own intuitions). The exercises are organized around a 12-step protocol that begins by asking the student to state a hypothesis about the relationship be- tween two variables; it then asks the student to choose and defend either (a) an antecedent variable that might reveal the original relation-

ship to be spurious, or (b) an intervening vari- able that might act as a causal mechanism con- necting their independent and dependent vari- ables. Some instructors will recognize the pro- tocol and its use of control variables as an outgrowth of the famous “elaboration model” developed at Columbia University during the 1950s. The model is so named because its goal is to elaborate on the relationship found be- tween two variables; the elaboration process involves adding control variables (antecedent or intervening), with the ultimate aim of being able to accurately interpret why the X–Y rela- tionship exists. It lays the groundwork for understanding causal analysis in the social sciences. Students who become adept with the elaboration model will have the foundation they need to understand much more sophisti- cated statistical approaches such as multiple regression and structural equation models. However, even if the student never takes another social science course, learning the elaboration model will provide him or her a valuable set of intellectual tools for dealing with the world critically.

 Because the most important part of the exer- cises in Doing Sociology with Student CHIP is the “white space” where students give their interpretations of the particular tables and plots at hand, I have updated several ex- amples of how this should be done in the first chapter (“The Problem of Social Order”).* I have also included a basic “footer” to each table that includes a “prediction” for what the student expects to find and a “finding” to allow the student to state what was actually found. Over the years, I have discovered that when students prepare such a footer the quali- ty of their answers improves significantly.

 Relatedly, to improve the “white-space” per- formance of students, I have embellished the section on “The Art of Reading Partial Tables” at the end of the introductory chapter entitled

* Note to the Instructor: Chapters may be used in any order; however, because Chapter 1 provides students with several concrete examples of how they should write up their findings, it is recommended that it be assigned first.

vii

A Primer on Elementary Data Analysis (see pp. 18–22). Many of the advanced exercises in this workbook require that students be able to recognize what has happened to the original relationship after the introduction of a control variable; for example, did the relationship maintain itself? is a multivariable model evident? is there an obvious interaction effect? All discussion is kept at an intuitive level and no statistics beyond simple percentages are used; in short, the emphasis is on analysis, reasoning, and the ability to recognize patterns in one’s findings—and not on high- level statistical analysis techniques.

 The software and data files for Doing Sociology with Student CHIP are now Web-based: www.zetadata.com/chipdh5.html. This has freed the use of the workbook from the vari- eties and vagaries of operating systems, as well as allowing for a return to both PC and Mac platforms.

Acknowledgments

Although many individuals—each of whom I am grateful to—provided suggestions for this and earlier editions, the following were particularly helpful and deserve special thanks: Louis Ander- son (emeritus, Kankakee Community College), Diane Balduzy (Massachusetts College of Liberal Arts), James Coverdill (University of Georgia), James A. Davis (National Opinion Research Cen- ter and University of Chicago), William H. Frey

(Brookings Institution and University of Michi- gan), Lawrence Hazelrigg (emeritus, Florida State University), Sharlene Hesse-Biber (Boston Col- lege), Margot Kempers and her Urban Sociology students (Fitchburg State College, Massachu- setts), Walter Schafer (California State University, Chico), and the several hundred unnamed-but- much-appreciated sociology students at Bryant University who have given me invaluable feed- back over the years. Joanne Socci performed many essential secretarial tasks for the prepubli- cation version of this workbook, for which I am grateful. I would also like to thank Bryant Uni- versity reference librarians Colleen Anderson, Samantha Cabral, Maura Keating, Laura Kohl, Cheryl Richardson, and Patricia Schultz for their help in tracking down reference materials, as well as my assistant, Erica Baldacchino. Ruth Bogart of Zeta Data (the producer of Student CHIP) is eternally accessible and helpful in opti- mizing the use of her software. Finally, I appreci- ate the encouragement and advice of my editor at Allyn and Bacon, Jeff Lasser.

I welcome your comments and suggestions on any aspect of this fifth edition of Doing Sociology with Student CHIP:

Gregg Lee Carter Department of History & Social Sciences Bryant University Smithfield, RI 02917 gcarter@bryant.edu http://web.bryant.edu/~gcarter

viii

Box 1. Topics for Introductory Sociology and Social Problems Courses, with Corresponding Readings

in Empirical Approaches to Sociology (5th edition) and Corresponding Data Analysis Exercises in Doing Sociology with Student CHIP: Data Happy! (5th edition)

Topic

Reading Number in Empirical Approaches to Sociology, 5th ed.

Corresponding Exercises in Doing Sociology with Student Chip, 5th ed.

Sociological Perspective (Social Theory) 1–4, 6, 7–12, 25, 27, 38 1, 3, 5, 15–16, 24 Research Methods 5–7, 10, 17, 21–22, 26, 31, 36–38 4–5 Culture 8–10, 15, 33, 35, 38–39 6, 29 Society 1–4, 7,–9, 11–14, 24, 28, 31–33,

35–36, 38–39 7–8, 20, 29

Socialization 15–16, 20, 25, 30, 32–35, 38 9, 12 Groups 4, 10, 15–20, 25, 27, 35, 37–38 10–13, 24 Interaction 4, 10–12, 17, 21–24, 33, 35–38 14, 24 Crime, Deviance, and Social Control 1–3, 6–7, 21, 25–28, 36–39 1, 3, 13, 15–16, 28–29 Inequality/Stratification 2, 4, 7, 13–16, 19–20, 22, 23, 29–

33, 35–36 17–20, 22, 26, 29

Race and Ethnic Relations/Inequality 4, 8, 10, 13, 16, 32–33, 38–39 21–24, 29 Gender and Gender Inequality 19, 22, 34–35 25–27 Social Change and Social Conflict 2, 4, 10, 13, 28, 36–30 3, 8, 28–29 Family 1, 3, 15–16, 20, 27, 35, 38 1, 2, 9–13 Economy 2, 14 7, 20, 29 Polity 2, 14, 31, 36, 39 19, 25, 27–29 Education 7, 20, 25, 39, 31, 35 9, 12, 17, 23, 27 Religion 1, 3, 7– 9, 39 1, 4, 16 Health/Healthcare/Aging 13–14, 18–19 2, 5, 8, 10–11, 20, 27 Population Change 1, 3–4 1 Work/Corporate America 13–14, 23, 35 7, 14, 18 Mass Media/The Internet 24, 35 14 Collective Violence/War 2, 36, 38–39 28–29

1

About Student CHIP

Intended for use by novices in social science data analysis, Student CHIP creates contingency tables (crosstabulations) and graphical displays associated with them. The following passages give a detailed description of the program. How- ever, it is easy to use, and even in the early-going instructors and students need probably do no more than skim over these details.

Although CHIP does not compute chi square or other significance tests normally calculated for contingency tables, most instructors can safely avoid any but the most cursory discussions of statistical significance (see my discussion on p. 12 in A Primer on Elementary Data Analysis). For data sets using the General Social Survey, the number of cases is very large and percentage differences of 9 or 10 are well above conventional standards for statistical significance (the “.05” level). For data sets using fewer cases, we would generally want percentage differences to be larger than 10 before getting too enthusiastic about the substantive or statistical significance of the rela- tionship.

URL

Earlier editions of Doing Sociology with Student CHIP used the desktop version of the software, but starting with this edition, the workbook uses a Web-based version—located at:

www.zetadata.com/chipdh5.html

Please note that you need to have a relatively recent version of Java Runtime on your computer to run the Web version of Student CHIP. If your machine does not open the above URL, you can get a free download of Java at:

www.java.com/en/

Menu You will begin all CHIP sessions by clicking on File. Note that all of your data files are on the http://www.zetadata.com/chipdh5.html URL.

The File menu offers the following options that you will use in this workbook:

Open lets you retrieve a data file.Student CHIP will display a list of the available files. Simply move the highlighting bar to the file name of your choice, and choose OK or press <Enter>.

After you have opened a data file, Student CHIP will display the descriptive title of the data set and the number of cases.

Print provides a screen dump of your current screen.

Print Graph is used when you screen includes a graphical display—such as line chart.

Clear clears your current screen.

Command Menu

The Command menu offers the following op- tions:

Info prints the names of the variables, the number of categories in each variable (e.g., Sex will have two categories), the current “causal order,” and the data description (if one is provided).

Marginals gives you the percentage distribu- tions for all of the variables in the data set.

Crosstab creates a crosstabulation of two variables. Crosstab is the workhorse for data analysis once you have entered or retrieved a data set and have modified it to your satis- faction.

First, consider the bivariate case (when you have only two variables) with no control variables. As a hypothetical example, you might be looking at the relationship between Region and Religion.

After you choose Crosstab, Student CHIP asks you to choose your two variables. Do this by selecting the first variable, then the second, from the list boxes that appear. You must decide which variable goes in the rows

Doing Sociology with Student CHIP

2

and which in the columns. Generally, you will want your independent variable to ap- pear in the columns and your dependent variable in the rows. Therefore, choose the dependent variable first, so it will be in the rows, and the independent variable second, so that it will appear in the columns.

Table Menu: Crosstab Options After you have chosen the variables, the options in the Table menu will be available:

Frequency prints a crosstabulation in terms of frequencies—that is, the number of cases in the categories and cells.

Percent Across displays tables based on per- centages across of the total or subtable.

Percent Down displays tables based on per- centages down of the total or subtable.

Control enables you to perform multivariable analyses.

Beyond two, CHIP views additional vari- ables as conditions or special cases of the relationship defined at the beginning of Crosstab. Thus, if you have a table with Race, Sex, Region, Education, and Income, and you choose Income and Region as your row and column variables, Student CHIP thinks of Race, Sex, and Education as defining sub- groups in which the relationship can appear. That is, it will let you look at the Income/ Region relationship among black males with high education, among black males with low education, black females with high educa- tion, and so on. Of course, you may change at any time your bivariable pair by choosing Crosstab from the Command menu.

After you choose Control, Student CHIP will ask you to select among the available control variables. You can highlight more than one variable at a time. Highlight the variables you want to use as controls, then click the Select button. You will see many choices—

including Frequency, Percent Down, and sever- al graphical display options. CHIP will pre- sent you with all possible control conditions (combinations of categories of your control variables) one at a time. You can decide whether you want to have any given sub- group displayed or whether you want to choose another option. Your control vari- able(s) will be in effect until you choose the Release option to return to the bivariate case. (Note: If you have used a control variable on one data set and then open up another data set, you must use the Release command before working with the second data set.)

Bar Chart / Pie Chart / Line Chart / Stacked Bar let you plot a graph of percentages in any of the four given styles. To print a graph, use the File menu as noted above (File / Print Graph).

Modify Menu

The Modify menu offers the following options:

Omit allows you to delete a category from a variable. From then on, that category does not exist as far as Student CHIP is con- cerned. For example, give the values republi- cans, independents, and democrats for the vari- able PartyID, you might omit independents and limit your analysis to differences be- tween partisans of the two major parties.

Combine allows you to put into a single, pooled group two or more categories. For example, if you have republicans, democrats, and independents, you might combine republi- cans and democrats and call them partisans, so your variable Partyid would now include the two categories independents and partisans.

The options of the Modify menu affect only the working data set; they do not auto- matically change any of the files on your disk.

3

A Primer on Critical Reading*

*Sections of this and subsequent chapters are taken from Gregg Lee Carter (ed.), Empirical Approaches to Sociology: Classic and Contemporary Readings 5th Edition (Boston: Allyn & Bacon, 2010) and are used by permission.

This chapter sets forth the intellectual steps for critiquing writing based on research. Together these steps form a foundation for reading and thinking critically; they have nearly universal applicability. You can use them not only in soci- ology, but in all courses that have empirical con- tent (virtually all the natural, social, and business sciences), whether or not the subject matter is quantitative. Moreover, once incorporated into your thinking, these steps can help you critique studies reported in the popular media and the arguments presented to you in everyday dis- course.

The fundamental strategy for explanation in all empirical disciplines, including sociology, is the following: To explain is to account for change in one phenomenon (variable, thing) with changes in another phenomenon or set of phenomena (variables, things). If you reflect for a moment, it should become apparent that this strategy is also fundamental in everyday explanations of life’s events. For ex- ample, why are you sad at some times and happy at others? Most likely, your mood changes in reaction to changes in the events, people, and situations in which you are involved. To give a concrete example from my own life: a few years ago, I was unhappy because a knee injury halted my running exercise program. In sum, changes in my physical well-being created changes in my training that, in turn, changed my mood. When one thing is explained by another, we say that the first is dependent on the second. The second is independent. Accordingly, my foot injury was an independent variable that explained the halting of my training (a dependent variable). Whether a vari- able is independent or dependent hinges upon the slice of reality under investigation. Thus, as we continue along the pathway of causation, the halting of my training becomes the independent variable that explained the change in my mood.

Because the world is complex, we can quickly begin discussing many variables simultaneously. As such, it is often helpful to diagram the causal connections among variables. In its simplest form,

such a diagram reads from left to right, with arrows imparting the causal order. For example, we would diagram the situation presented here as: Knee Injury  Stopping Training  Unhappiness. Social scientists use the term model to describe an interrelated set of variables that represent a slice of reality.

The most important models are those which are explanatory (recall that to explain is to account for change in one variable with changes in another variable or set of variables). Much of the research in soci- ology—as in all the sciences—is organized around developing and testing explanatory models. How- ever, even when this is not a primary aim, virtual- ly all nonfiction writing is organized around some sort of model of the world and how it operates. As a critical reader, you must develop the ability to discern the model at hand.

The first step in critiquing an article or argument is thus to identify the model. This involves identify- ing the dependent and independent variables and ascertaining the relationship between them. Let us illustrate with the following article:

Poverty More Than Race Increases Risk of Cancer, Panel Concludes Bethesda, MD (AP). A federal advisory panel yesterday focused new attention on poverty—far more than race— as one of the most powerful and under-estimated risk factors for cancer in America.

Both inadequate access to health care and unhealthy habits among the poor contribute to the problem, the President’s Cancer Panel was told.

“Poor people are more focused on day-to-day survival— and I’m afraid that health care more often takes a back seat,” [the Secretary of] Health and Human Services… told the panel.

Added Dr. Samuel Broder, Director of the National Cancer Institute: “It is difficult for an individual to say, ‘I’ll go for a mammogram today’ when you are worried about how to pay for dinner.”

He linked the higher cancer rates for poor people to what he called “poverty-driven lifestyles” that may include unhealthy diets, greater use of alcohol and tobacco, occupational risks, and less access to medical care.

Doing Sociology with Student CHIP

4

People living below the federal poverty level have a death rate from cancer that is twice as high as that for the rest of the population. Black men, who outnumber whites in poverty three-to-one, have a 25 percent higher risk of contracting cancer.

Dr. Harold Freeman, a surgeon at Harlem Hospital in New York and the first black person to lead the panel, said, “Race in itself is not the cause of death. It is a circumstance in which people live, basically defined by poverty.” Step 1: What is the model? In identifying the model, you must first determine the key dependent variable(s). A dependent vari- able is the phenomenon or event that the research- er or author is trying to explain. It’s what is being caused. When sociologists think about individuals (as opposed to a collection of persons, such as a group or the residents of a city), the most impor- tant dependent variables involve behaviors and attitudes. (Why do some people drink heavily, while others abstain totally? Why do some whites hate blacks, while other whites are indifferent to skin color?) In the newspaper article quoted above, the key dependent variable is cancer (or the odds of dying from cancer).

Next, you must identify the key independent vari- able(s). An independent variable is the thing that explains, in part, the dependent variable. It is a reason for, or a cause of, the dependent variable. When trying to explain individual behaviors and attitudes, sociologists are drawn strongly to inde- pendent variables that indicate the groups and social networks to which people belong. In the article above, the key independent variable is poverty (purportedly much more important than race, which once was considered another key determinant of cancer risk). Poverty is linked to cancer via the intervening variables (variables that are simultaneously independent and depend- ent) of unhealthy diets, use of alcohol and tobacco, occupational risks, and access to medical care. In sum, poverty partly determines this latter set of vari- ables, which, in turn, partly determines the odds of dying from cancer.

Next, you must be able to recognize the form of the relationship between the independent and the dependent variables. The easiest functional forms to conceptualize and to identify are the straight-

line or linear. If two variables are linearly posi- tively related, then increases in one are associated with increases in the other. If two variables are linearly negatively related, then increases in one are associated with decreases in the other. In the above article, poverty is positively associated with unhealthy diets, the use of alcohol and tobacco, and occupational risks, whereas it is negatively related to access to medical care.

Especially when our subject of empirical inquiry is people and their behavior, thoughts, and feelings, explanation (i.e., accounting for variation in a dependent variable with variation in one or more independent variables) is not the same thing as understanding. Understanding is deeper; it is that “aha, I see” experience within us. As a critical reader, you must understand why the independ- ent variables are important. This understanding depends upon your ability to empathize with the people under study. In other words, you must not only identify each key independent variable, but also recognize the interpretation showing why it has the effect that it does. This interpretation may or may not be explicit. In the Associated Press article, Dr. Broder’s interpretive comments give us a deeper understanding of the relationship be- tween poverty and access to health care when he states: “It is difficult for an individual to say ‘I’ll go for a mammogram today’ when you are worried about how to pay for dinner.”

A picture is worth a thousand words, and the model becomes clearer when we diagram it. The rules for sketching models are as follows:

 Begin the sketch on the left and follow the causal flow to the right until you have enter- ed the final dependent variable; that is, the independent variable(s) on the left, the inter- vening variable(s) in the middle, and the de- pendent variable(s) on the right:

 Use single-headed, straight arrows to denote causal relationships:

 Use a double-headed, curved arrow to denote an association between two variables that

A Primer on Critical Reading

5

exists but is not causal (see the forthcoming discussion on the “criteria for establishing cau- sality”):

 If there is a feedback effect or two-way causa- tion, e.g., X affects Y, which, in turn, affects X, then use a second arrow:

 If a causal relationship is hypothesized but is not confirmed by the data, denote it with a cross-hatched, straight arrow; alternatively, you may use a broken arrow:

 Note the functional form of the relationship between two variables by placing an appro- priate symbol above a straight arrow and to the left of a curved double-headed arrow; the most common forms will be the linear, either positive (“+”) or negative (“–“):

Although there is no rule for symbolizing non- linear relationships, one simple strategy is to use a symbol that looks like the curve (e.g., a U for a parabolic association):

(the X,Y relationship is U-shaped)

 Nominal variables (those whose values do not have magnitude, i.e., cannot be ranked “greater than” or “less than,”) need to be set at a given value. For example, if using the vari- able sex, you must specify either “male” or “female”:

(Males are more likely to have high incomes.)

 If a nominal variable has only two categories, e.g., Sex (male/female) or Race (white/ black), then either category may be used in the sketch; the following sketch is equivalent to the preceding one:

(Females are less likely to have high incomes.)

If we follow the above rules, our model sketch comes out like this:

There are other ways we could have sketched this model. For example, it is not clear from the article whether race and poverty are simply as- sociated or whether race is a determinant of pov- erty. If we think that the argument was for an association only between these two variables, then our sketch would look like this:

Doing Sociology with Student CHIP

6

Step 2: How is each variable measured? Variables are concepts. Theories clarify the causal links among concepts. To apply or test a theory, it is necessary to specify these concepts in terms of their manifestations in the concrete world. This process is called measurement. Scientific articles are very good at alerting us to the difference between conceptual variables and their measure- ments. Articles in the popular media and everyday arguments are not so good at this. Regardless, as a critical reader you must have the ability to distin- guish a concept from its measurement, realizing that they are not the same thing. Measurement answers in concrete empirical terms the question: “By X, what do you mean?” (“By ‘educational quality’ of a college, what do you mean? Do you mean the grade point average of the student body? Do you mean the percentage of faculty members with doctoral degrees?”). Models tested with dif- ferent measurements will yield different results, though the results will not differ by much if both measurements are fairly valid—that is, if they are truly indicative of the phenomena under study. And what if a researcher finds that a particular X and a particular Y are not related as hypothesized? Then one must ask, Did the research turn out this way because the model is wrong (i.e., does not truly represent reality)? Or is it that the model is perhaps right but simply was tested incorrectly (e.g., with poor measurements)?

In the newspaper article quoted above, measure- ment issues are given short shrift (and this should raise the eyebrows of the critical reader). Poverty is measured as living below the federal poverty line; although this line is not given here, it is readily available from the U.S. Bureau of the Census (in 2009, the poverty line for a family of 4 was a cash income of approximately $21,200; this line is up- dated each year to reflect changes in the Consumer Price Index). Cancer is measured as a group death rate from the disease (e.g., number of blacks dying from cancer per 100,000 blacks); we must assume that all cancers are involved. We do not know the data source, but we could obtain it by contacting the panel at the National Cancer Institute in Bethesda, Maryland. No measurements are given for any other variable. Thus, we cannot specifically answer the question “By X, what do you mean?” for unhealthy diets, occupational risks, or access to

health care. Similarly, we are not really sure what constitutes a dangerous level of alcohol or tobacco use.

Step 3: How well do the data fit the model?  Critical thinkers realize that data rarely, if ever, fit a model perfectly. In sociology a perfect fit is even rarer because our measurements tend to be crude and because human affairs are so complex that any one data set is unlikely to contain the full range of needed independent variables. Thus, the fit between models and data is a matter of degree.

One standard that many scientists use to judge how well data fit models is statistical signifi- cance. A pattern in the data (say, a positive cor- relation between X and Y) is considered statis- tically significant if it is unlikely that it could have occurred by chance alone. For example, about a half million females and about a half million males live in Rhode Island. If it were simply a matter of chance who ended up in prison, then we would expect about half of the state’s prison population to be male and about half female. But that is not the case. The prison population of Rhode Island (as in every state) is overwhelmingly male (95% or higher). Thus, the data contain a strong pattern that is unlikely to be due to chance and that would support this model: Sex (M)Crime.

Researchers typically report whether their find- ings are statistically significant. At this point in your intellectual career, you should conclude that the data fit the model well if statistically significant associations are reported. At the same time, how- ever, you should be developing an internal set of standards and an intuitive sense of how well the findings fit the model. For example, in the above newspaper article few data are reported (and none for the intervening variables and cancer rates), but those that are given convincingly support the following: (1) poverty is linked to dying of cancer (“People living below the federal poverty level have a death rate from cancer that is twice as high as that for the rest of the population”), and (2) race is linked to poverty and cancer (“Black men, who outnumber whites in poverty three-to-one, have a 25 percent higher risk of contracting cancer”). Imagine, however, that the article had reported that those below the poverty line had a 27.2

A Primer on Critical Reading

7

cancer would increase the odds of working less or quitting one’s job altogether; it would also seem likely that a person’s expenses would rise, as any insurance one had—private or public—would be unlikely to cover all medical costs. Thus, it seems reasonable to argue that cancer feeds back into poverty; in short, not only does poverty lead to cancer, but cancer leads to poverty. My alternative model can be sketched as follows:

percent chance of suffering cancer, while those above the line had a 26.4 percent chance. Even though this difference might be statistically signifi- cant given a population of tens of millions of people both below and above the poverty line (all things equal, the larger the sample size, the greater the likelihood of finding statistical significance), substantively and intuitively we would not be per- suaded that the PovertyCancer model has much explanatory worth. Indeed, we would be prompt- ed to seek out other potential predictors of cancer with which to construct more powerful models. Finally, you should be developing internal standards and an intuitive sense of how well data fit models because some scholarly essays and many journalistic articles do not contain quantita- tive information and therefore will not use tests of statistical significance.

Step 4: Propose an alternative model based on your assessments in Steps 1–3. By definition, a model represents only a thin slice of reality. It can always be made more or less complex (say, by adding or deleting additional independent variables). Up to this point, in Steps 1–3, all the critical readers of an article should have made similar assessments (the various compon- ents of the model either are there or are not; the measurements are either presented or are not; the data are significant in total, in part, or not at all). At Step 4 you can make your unique contributions as a reader and a critic. If you are fairly well convinced by the model, its measurements, and the findings, then your contributions in Step 4 will be minimal (adding another independent variable, for example, along with your rationale). On the other hand, if you find the model unconvincing, either on theoretical grounds or because the data simply do not fit it very well (especially if you think the measurements of the variables seem adequate), then your Step 4 contribution may result in a complete reworking. In the newspaper article at hand, the theoretical model rings true to me. Further, the data presented—though scant— fit the model well. Thus, my “alternative model” is a tinkering of the original model, not a major overhaul. It would seem to me that getting

No matter what the model, you can always pro- pose an alternative and produce a justification that appeals to intuition or collective common sense. (May your professor be generous in assessing your first efforts as a social theorist!)

On the next page, in summary form, are the steps we have been examining. A full appreciation of these steps takes practice. You must apply them again and again. However, even your initial use will sharpen your mind and reward you with insight.

A final comment: This brief primer on critical read- ing and thinking does not, of course, cover the full spectrum of analytic approaches to studying the world, in general, or the components of society, in particular. Rather, it lays a foundation upon which you can construct subtler and more sophisticated techniques of observation and analysis (whether the analysis involves reading, writing, researching, or just thinking). Indeed, such subtlety and sophis- tication will be apparent in much of the sociologi- cal writing you will read now and in the future.

Doing Sociology with Student CHIP

8

Box 2. Steps Toward Critical Reading (and Thinking)

1. What is the model?

a. What is the dependent variable(s)? (the phenomenon or the phenomena being explained)

b. What are the independent variables? (the reasons for, or causes of, the dependent variable)

c. How are the independent and the dependent variables functionally related? (e.g., as the independent variable changes, does the dependent variable increase or decrease?)

d. Interpret the relationship between each independent and dependent variable. (What’s happening in the world such that we would expect to find these two variables related?)

e. Diagram the model.

2. How is each variable measured?

3. How well do the data fit the model?

4. Propose an alternative model based on your assessments in Steps 1–3.

9

A Primer on Elementary Data Analysis

Dig a well before you are thirsty. —Chinese proverb

Many students are limited in their quantitative abilities, and instructors cannot go into much de- tail without transforming the course into a class on research methods or statistics. However, all college students—regardless of major—should have a grasp of the following elementary con- cepts of data analysis and should be able to put them to practical use:  Measures of central tendency  Measures of dispersion  Basic tabular analysis  Scatterplots and the correlation coefficient  The criteria for establishing causality These basic concepts of data analysis are intrinsic to understanding many of even the simplest re- search efforts in sociology and to applying the protocol for reading critically (see the end of the preceding chapter) to the reporting of such ef- forts. I will illustrate these basic concepts using the information contained in Box 3 on the next page. The data are for all 19-year-olds participat- ing in a recent General Social Survey (GSS). The National Opinion Research Center (NORC) at the University of Chicago conducted the GSS almost every year between 1972 and 1994.1 Since 1994, NORC has conducted the GSS every two years. In the early years, the GSS random sample was around 1,600, by the mid-1990s, it rose to nearly 3,000, and since 2006, it has increased to approximately 4,500 adult Americans. Respon- dents are asked a host of questions on their family backgrounds, personal histories, behav- iors, and attitudes toward a variety of issues. Because NORC uses rigorous scientific sampling strategies and has a high response rate for the GSS, it is safe to assume that the data are of a very high quality.

Measures of Central Tendency

One of the most important purposes of statistical analysis is to summarize and simplify informa- tion. One useful summary is the “typical value”

of a variable. Three ways to calculate the “typical value” are the mean, the median, and the mode.

You probably learned to calculate the mean—or the arithmetic average—of a distribution in the fourth grade: you simply sum the values and di- vide by the number of cases. Which would be easier to keep track of in your mind, the 18 val- ues of Father’s Education listed in Box 3 or their rounded mean of 12.4 (that is, 18 divided into the sum of all of the Father’s Education values, or 224/18)? Prove to yourself that you remember how to calculate the mean by finding it for Moth- er’s Education.2

If Father’s Education for CaseIDs 13 and 15 were 54 instead of 14, then what would the mean for this variable be?3 Although none of these 18 indi- viduals had a father with exactly 12.4 years of education, this mean is in the “ballpark,” that is, these individuals had fathers with educations within a few years (+ or –) of 12 or so. With the change in the values for Father’s Education for CaseIDs 13 and 15, however, the new mean of 16.9 is in no way a “typical value.” The problem, of course, is that we now have “extreme values” (2 fathers with 54 years of schooling in the pres- ence of most of the others being around 12 years). Whenever we encounter a variable with extreme values, the preferred measurement of the “typical value” is the median. The median is defined as the middle value of a distribution after all the values have been ranked (either from low to high, or from high to low; if there is an even number of cases, then the median is the average of the two middle values). For Father’s Educa- tion, the original median of “12” is unaffected when we change the values of Father’s Education for CaseIDs 13 and 15 (it does not matter whether the values are “14” or “54”)—just as any median for any set of values is unaffected by the presence or absence of extreme values. To prove to your- self that you can calculate a median, find it for Family Income.4 Further, prove to yourself that you can do it for an odd number of cases; cal- culate the value of Family Income only for those

Doing Sociology with Student CHIP

10

who report being in “Good” health (CaseIDs 1–5, 8, 9, 12, 13, 16, and 17).5 How does one calculate the “typical value” for a variable such as Health Status in Box 3—that is, for a variable with relatively few values? In this case, we designate the typical value as the one that appears most often. This typical value is cal- led the mode, and for the variable “Health Stat- us” the mode here is “Good.” Show yourself that you can compute a mode: what is it for the Zodi- ac Sign variable?6

Measures of Dispersion Generally speaking, a mean cannot be fully ap- preciated unless we have an idea of how spread out or dispersed the data are. For example, the mean for the following distribution is 50:[0, 50, 100], as is the mean for the distribution [49, 50, 51]. However, in the first case, the mean does not do a very good job of telling us the “typical va- lue,” whereas in the second case it does. In the first distribution, the data are dispersed widely; in the second distribution, they are dispersed narrowly.

Dispersion can be indicated with a numeric sum- mary, such as the range or the standard devia- tion, or it can be captured graphically with a device such as a histogram. The range is comput-

ed by subtracting the lowest value from the high- est. For example, the range for Mother’s Educa- tion in Box 3 is 11 (20–9). Prove to yourself that you can compute a range; do it for Father’s Edu- cation.7

Obviously, the range is a very crude measure- ment of dispersion and does not give us much feel for the data. Most importantly, it gives no idea of the shape of the distribution. For ex- ample, the following two families have equal ranges of education but present dramatically dif- ferent configurations: The Skrang family includes Jim (8 years of schooling), Tim (7), Slim (7), and Sue (16); the Skrug family includes Wilbur (16), Phil (16), Mary (15), and Tom (7). In short, even though the range of education is the same for both families (9 years) the range statistic misses the fact that the Skrangs are bunched at the low end of education, while the Skrugs are clustered at the high end. These differences are highlighted more dramatically in the histograms in Box 4. A histogram is simply a bar graph in which the bars are contiguous and indicate the frequency or the relative frequency (percentage) of the values of a variable. For small numbers of cases, such as the Skrang/Skrug family data or the GSS data in Box 3, configurations in the data are readily observed. However, for larger numbers, the pattern of dis- persion is not always eye-catching and recogniz-

A Primer on Elementary Data Analysis

11

another variable or set of variables.) A second meth- od for assessing whether a change in X is associ- ated with a change in Y is a scatterplot (also cal- led a scattergram). You learned how to plot data in your first math class in junior or senior high school. To refresh your memory, examine the scatterplot of the GSS levels of education of the respondent’s mother and father as presented in Box 5. For each case, note how we proceed along the horizontal or X axis until we meet the value for Father’s Education and then proceed upward until we reach the value for the Mother’s Educa- tion. We could mark this point with any common symbol, such as an asterisk (*), a dot (.), or an x; or, we could be more informative and mark it with the CaseID number.

able. In such a situation, histograms and similar graphical techniques can provide the solution.

Prove to yourself that you can construct a histo- gram to see how data for a variable are dis- persed. Construct one for Health Status.8

Scatterplots and the Correlation Coefficient

Measures of central tendency and dispersion tell us much about particular variables, but they do not get to the heart of what concerns most scien- tists—explaining the world. To do this, we must examine at least two variables simultaneously. (Recall the definition given earlier: To explain is to account for change in one variable with change in

Box 4. Graphical Displays of Dispersion Using Bar Graphs (Histograms)

Education in the Skrang Family (Frequency Histogram)

Level of Education

Number

<High School College 0

1

2

3

4

Education in the Skrug Family (Frequency Histogram)

Level of Education

Number

<High School College 0

1

2

3

4

Education in the Skrang Family (Relative Frequency Histogram)

Level of Education

Number

<High School College 0

20 40 60 80

100

Education in the Skrug Family (Relative Frequency Histogram)

Level of Education

Number

<High School College 0

20 40 60 80

100

Doing Sociology with Student CHIP

12

Box 5. Scatterplot of Mother’s Education and Father’s Education

The scatterplot confirms what you may have sus- pected: The educational backgrounds of a res- pondent’s mother and father are very similar— or, more technically, are strongly positively as- sociated. Prove to yourself that you can construct a scatterplot: plot the data for Mother’s Educa- tion and Family Income in Box 3, putting Moth- er’s Education on the horizontal axis.9

We can summarize the relationship in Box 5 nu- merically with a correlation coefficient (symbol- ized by the letter r), which indicates the strength of the linear relationship between two variables. You can find the arithmetic formula for the cor- relation coefficient in any elementary statistics text. At this point, it is more important for you to have an intuitive feel for r. A correlation coef- ficient may range from –1 to 0 to +1. A value of 0 signifies no linear relationship between two vari- ables. As r tends to –1, the two variables are more strongly negatively associated; as r tends to +1, the two variables are more strongly positively as- sociated.

Real-world correlations never attain a value of –1 or +1: First, anything that happens outside of the laboratory does so for many reasons. To account

for all of the changes we might observe in a par- ticular Y, we need to have many Xs (independent variables). Second, even if we were totally con- vinced that a single X could explain a particular Y, we would never obtain a correlation of + or –1 because of measurement error; that is, we can never measure any variable precisely. The line drawn through the scatterplot points in Box 5 represents the theoretical relationship between the respondent’s Father’s Education (X) and Mother’s Education (Y). The vertical distance be- tween the line and any particular point repre- sents the missing effects of other causes of Y, as well as errors in measuring the variables. The closer the points tend to cluster near this line, the stronger the absolute value of r.

Father’s Education

20181614121086

M ot

he r’s

E du

ca tio

n

22

20

18

16

14

12

10

8

(r=.77)

For a relationship to be considered “statistically significant,” conventional standards require that the probability of it being due to chance is less than 5 percent. Statistical significance depends upon both the strength of the relationship and the size of the sample. For example, if the sample size is 45 or more, a correlation of + or –.25 is statistically significant (i.e., it would be obtained by chance alone less than 5 times in a 100). Although the scatterplot in Box 5 is based on a relatively small number of cases (N=18), the relationship is strong enough to be statistically significant. (An r of .77 is very strong in any soci- ological study; this particular relationship could be expected to appear by chance alone fewer than 1 in 1,000 times.)

Box 6 shows a variety of relationships between a hypothetical X and Y, along with accompanying correlation coefficients. Ideally, a researcher would show you both a scatterplot and its ac- companying correlation coefficient. (At the very least, you could check to make sure that r was appropriate—that is, that the data were related in a straight-line manner.) Unfortunately, to con- serve space in scholarly books and journals, this is rarely done. Nevertheless, you will have a deeper understanding of correlation coefficients now that you know how scatterplots are con- structed and have a mental image of the connec- tion between a particular correlation and what its scatterplot should look like.

A Primer on Elementary Data Analysis

13

Box 6. Scatterplot Examples with Accompanying Correlation Coefficients

Strong Positive Correlation (r = . 98) Strong Negative Correlation (r = –. 98)

INDEPENDENT VARIABLE (X)

3020100

D EP

EN D

EN T

VA R

IA BL

E (Y

)

0

-10

-20

-30

-40

Moderate Positive Correlation (r = . 68) Moderate Negative Correlation (r = –. 68)

Zero Correlation (r = .00) Zero Correlation (r = .00) (because X and Y are not related) (because X,Y relationship is nonlinear)

INDPENDENT VARIABLE (X)

3020100

D EP

EN D

EN T

VA R

IA BL

E (Y

)

30

20

10

0

INDEPENDENT VARIABLE (X)

3020100

D EP

EN D

EN T

VA R

IA BL

E (Y

)

0

-10

-20

-30

INDEPENDENT VARIABLE (X)

3020100

D EP

EN D

EN T

VA R

IA BL

E (Y

)

30

20

10

0

-10

INDEPENDENT VARIABLE (X)

3020100

D EP

EN D

EN T

VA R

IA BL

E (Y

)

12

10

8

6

4

2

0

INDEPENDENT VARIABLE (X)

3020100

D EP

EN D

EN T

VA R

IA BL

E (Y

)

40

30

20

10

0

Doing Sociology with Student CHIP

14

Basic Tabular Analysis A second method for assessing whether a change in X is associated with a change in Y is a special kind of table called a crosstab (short for cross- tabulation). Although easy to construct, a cross- tab can produce powerful insights. Most impor- tantly, like the scatterplot, it allows us to estab- lish whether change in one variable is associated with change in another. Let’s use the table in Box 7 (constructed from the GSS data in Box 3) to detail the elements and the logic of a properly constructed crosstab.

Box 7. Crosstab of Family Income* by Parents’ Education**

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High (> $26,875)

36.4%

(4)

71.4%

(5) Family

Income

Low (< $26,875)

63.6% (7)

28.6% (2)

100%

100% (11) (7)

N=18

Prediction: Parents’ Education and Family Income are positively related. Finding: Strongly confirmatory, i.e., those respondents whose parents possess “high” amounts of education (> 24 years) have a (71.4%–36.4%=) 35.0% greater chance of having a “high” (>$26,875) Family Income

*Family Income has been dichotomized at the median. **Parents’ Education = Mother’s Education + Father’s Educa- tion; the values of 17–24 are at or below the median for this variable, while the values 25–38 fall above the median.

First, note the title. Often the best title is simply the “<dependent variable> by <independent var- iable>”; thus, here it would be “Family Income by Parents’ Education.”

Next, note the footer, that material at the base of the table. Footers vary greatly in content; how- ever, in this workbook we will stress the kind used in Box7—that is, a prediction and a finding. The prediction is simply our hypothesis or best guess about the functional form of the relation-

ship (which variable is independent, which is de- pendent, and the expected relationship between the two—here we expect the two variables to fall and rise together, so we are predicting a “posi- tive” relationship). Using this kind of footer will train you to (a) look for relationships between variables (your “prediction”) and (b) recognize them when they exist (your “finding”).

Next, note the variable labels and the associated value labels. The row variable label here is Family Income, with the values High (>$26,875) and Low (<$26,875). The column variable label is Parents’ Education, with the values Low (≤24 Yrs) and High (>24 Yrs). Note that the many values of both of these variables (e.g., 26 years of parents combined education, 30 years, 34 years, and so on; $21,250, $32,500, and so on) have been col- lapsed into only two key values or categories: for Parents’ Education, less than or equal to 24 years versus more than 24 years (in other words: less than or equal to the median versus greater than the median); for Family Income, less than $26,875 dollars per year and more than $26,875 thousand dollars per year (in other words: less than the median versus greater than the median). Often the categories of a many-valued variable are col- lapsed into fewer categories for simplification and to allow us to begin seeing patterns that may exist in the data. You must be clear whether the values of the inde- pendent variable are placed in the rows or in the columns. (Even if the author does not tell us which of the variables is the independent, you should be able to figure it out theoretically. Here, obviously, Parents’ Education is a possible deter- minant of Family Income—not vice versa.) Most commonly, you will find the independent var- iable placed on the column and the dependent variable on the rows. Moreover, you will find the table constructed as in Box 7—that is, with the values increasing from the first column to the last column (here, the first column is ≤24, and the last column is >24), but with the values decreasing from the first row to the last row (row #1 is >$26,875, and row #2 is <$26,875). The logic of this construction relates to the format of a scatter- plot—the means for finding a relationship be- tween X and Y that you just learned. Notice in

A Primer on Elementary Data Analysis

15

Box 5 how the values increase reading from left to right and from down to up, just like the values in the Box 7 crosstab. Box 8 shows the close con- nection between a properly constructed crosstab and its respective scatterplot, as well as their sim- ilar capacity to describe a relationship between X and Y.

Box 8. A Crosstab Overlaying Its Equivalent Scatterplot, and Crosstab by Itself

Crosstab Overlaying Its Equivalent Scatterplot

High

*

***

*********

Middle

**

********

*

Low

************* *****

****

*

Low Middle High Independent Variable (X)

Crosstab by Itself (Dependent Variable by Independent Variable)

Independent Variable (X) Low Middle High High

5.8% (1)

20.0%

(3)

81.8%

(9)

Middle

11.8%

(2)

53.3%

(8)

9.1% (1)

Low

82.4% (14)

26.7%

(4)

9.1% (1)

100% 100% 100% (17) (15) (11)

N=43

Next, note the direction (row or column) in which cell percentages have been calculated. To achieve our ultimate aim—that is, to see whether

a change in X is associated with a change in Y— the hard-and-fast rule is to calculate cell percent- ages in the direction of the independent variable. Thus, in our Box 7 table, the column sum of 11 was used to calculate the percentages in the cells for the first column (not either of the row sums, nor the total N of 18). Similarly, the sum for the second column, 7, is used to calculate the cell percentages for that column (not either of the row sums, nor the total N of 18).

The rule for reading or interpreting cell percent- ages is to compare them in the opposite direction from that used to calculate them. Let us apply this rule. Seven of those respondents having parents with “low” education (≤24 years) live in homes where family income is “low” (see the Row 2/Column 1 cell); that is, 63.6 percent (7/11) of those with parents having a low amount of education are in families where earnings are less than $26,875 per year. Recall that our ultimate aim (which cannot be repeated too often!) is to discover if a change in X is associated with a change in Y. If we stay on Column 1 (comparing the 63.6% with the 36.4%), we cannot answer this question because we have not changed values in X! Only by com- paring “low” versus “high” parental education (that is, by reading across a row) can we discover whether changes in X lead to a change in Y. Our final conclusion? Indeed, Parent’s Education and Family Income are related: Those respondents having parents with “high” levels of education (combined years of schooling greater than 24) have a 35.0 percent (71.4%–36.4%) greater chance of being part of a family where income exceeds $26,875 per year.

Although there are many ways to determine the strength of an association between two variables in a crosstab, one of the easiest and most power- ful is simply to compare percentage differences: given a change in X, what is the percentage change in being at a particular value of Y? Recall that statistical significance depends upon both the strength of the relationship (here, defined as percentage difference) and the size of the sample Thus, even though we only have a very small sample of 18, the relationship is nearly strong enough to meet conventional standards of signi- ficance—usually defined as the probability of the

Doing Sociology with Student CHIP

16

relationship being due to chance is less than 5 percent. According to the chi-square test for statistical significance, an option under the Table command of Student CHIP, the probability here is only 14 in 100.10 Regardless of statistical sig- nificance, you should be developing an ability to recognize percentage-difference relationships be- tween variables in crosstabs. Obviously, I used a small sample in this chapter so you could prac- tice data analysis by hand. (Whether a sample is “small” or “large” depends on the needs of the particular research project; in general, however, the further below 100 or so, the less likely a sample will yield statistically significant results.)

Prove to yourself that you can construct and read a crosstab. Let X be Family Income as recoded in Box 7. Let Y be Health Status. Crosstab these two variables. Put the two values of Family Income on the columns (let Column 1 = “low” Family In- come, i.e., <$26,875; let Column 2 = “high” Family Income, i.e., >$26,875); put two values of Health Status on the rows (let Row 1 = Excellent; let Row 2 = Good/Fair). Is there a relationship using a percentage-difference comparison? Does it seem weak or strong, especially given the sam- ple size? Focus your analysis on those with Excel- lent health—in other words, does the probability of reporting “excellent health” vary with Family Income?11

Criteria for Establishing Causality in Nonexperimental Situations

The experimental method is the most persuasive means that human beings have developed for establishing that a change in X truly generates a change in Y. The logic of an experiment is simple, yet powerful: the experimenter controls for—that is, holds constant—all possible determinants of Y save one, the suspected causal factor X. Thus, if a change in X is followed by a change in Y, then the only possible cause must be X. One possible confounding factor is the passage of time; per- haps Y changes not in response to X, but simply because of aging. To make sure that the effects of X are not confused with the simple passage of time, the experimenter sets up two groups at the beginning of the study: one that receives X, var- iously called the experimental or treatment or

test group, and one that does not receive X, called the control group. If the two groups differ on Y at the end of the experiment, then the only possible cause must be the presence or absence of X.

Of course, the idea of control or holding every- thing constant implies that the two groups are alike. For example, if we wanted to learn whether the artificial sweetener saccharin causes cancer in rats, we could not put white lab rats in the con- trol group and your basic East Coast wood rat in the experimental group. (From my personal ex- perience, nothing makes these guys sick. They eat aluminum foil, barbed wire, pesticides, and all forms of garbage—without any ill effects!). When dealing with human subjects, experiment- ers employ one of two key methods to ensure the sameness of experimental and control groups. Occasionally, they use matched samples, in which they try to match key characteristics be- tween the two groups. For example, if the experi- mental group is all female, then so is the control group; if the experimental group contains all young adults, then so does the control group.

But where does one draw the line? That is, how does one know that the experimental and the control group have been truly “matched” on everything essential? It is hard, perhaps impos- sible, to determine this. To avoid this problem, experimenters more commonly use the second method to ensure equality between groups. In this procedure, called randomization, they as- sign an equal number of subjects randomly to an experimental group and to an accompanying control group. On any particular trial of an ex- periment, the two groups may not have charac- teristics that are exactly equal. If the experiment is repeated many times, however, and if ran- domization is employed, then the differences be- tween the experimental and the control groups will tend to cancel each other out and any signi- ficant differences between the groups on Y can safely be attributed to differences on X.

Despite the great power of the experimental method to ascertain causation, it is rarely used in sociology (except in the subfield of social psycho- logy). A key reason is that in the real world people’s lives are not influenced by isolated inde-

A Primer on Elementary Data Analysis

17

Church Attendance Infrequent Frequent

Yes

38%

22%

Delinquency

Problem No

62%

78%

100%

pendent variables. We are complex beings who live in a complex world. Our individual and col- lective attitudes and behavior are subjected to a barrage of influences. To understand more clearly the effects of many factors, simultane- ously and relative to one another, sociologists use surveys and censuses, both of which include many items or questions measuring many variables. Given these kinds of data, how does one go about ascertaining that a change in X produces a change in Y? Box 9 contains the an- swer. Take special note of Criterion 3, which draws its logic from the experimental method (getting rid of possible confounding influences on Y by holding them constant).

100%

Prediction: Church Attendance and Delinquency Problem are negatively related.

Finding: Moderately confirmatory, i.e., those teenagers who attend church “frequently” have a (22%–38%=) 14% smaller chance of having a delinquency problem.

Box 9. The Criteria for Establishing Causality in Nonexperimental Situations

1. X and Y are associated. 2. X precedes Y. 3. The X–Y relationship is nonspurious. That is, it is

resistant to controls; it maintains itself when a reasonable competing independent variable that is antecedent to both X & Y is held constant.

4. The X–Y relationship is intuitively pleasing. That is, it fits with our current understanding of how the world functions.

Let’s explicate the criteria in Box 9 with an ex- ample drawn from Rodney Stark’s introductory sociology text.12 Surveying a large sample of teenagers in Richmond, California, Stark and fel- low sociologist Travis Hirschi tested the seeming- ly reasonable model that religiosity and juvenile delinquency are inversely related. We may sketch their bivariate model as follows:

Indeed, the following table (adapted from Stark’s text) seems to support their model:

At his point, we have a strong case for causality. We have met Criteria 1 (the table shows an as- sociation), 2 (most people are introduced to reli- gion before the adolescent years, when delin- quency problems peak), and 4 (it seems plausible that a very religious person would be less delin- quent, as all the major religions forbid stealing, lying, cheating, and similar reprehensible acts). However, perhaps the world looks like the model below; it represents a reasonable competing ex- planation, allowing for a reasonably strong check for the possible spuriousness of the original X–Y relationship:

This model also is backed by reason. Girls are more likely than boys to attend church and are raised to be less aggressive (to wit, consider the prison data cited earlier). Furthermore, if this model is true, then we should not find any signi- ficant association between religiosity and delin- quency after holding sex constant (examining the relationship just for boys, then just for girls). Fi- nally, if this alternative model is true, then we would be forced to look elsewhere for causal fac- tors, perhaps family income or the degree to

Doing Sociology with Student CHIP

18

Partial Table #2: Girls

Church Attendance Infrequent Frequent

Yes

10%

10%

Delinquency Problem

No

90%

90%

100%

100%

Partial Table #1: Boys

Church Attendance Infrequent Frequent

Yes

50%

50% Delinquency

Problem No

50%

50%

100%

which an adolescent’s closest friends are delin- quent.

Indeed, a check for spuriousness reveals the orig- inal relationship between religiosity and delin- quency to be noncausal (i.e., spurious):

100%

The notion of checking for spuriousness is also applicable to nonquantitative research, studies reported in the popular media, and everyday thinking.13 In these instances, the idea is expres- sed as a search for alternatives to the interpreta- tion given to a particular interaction, happening, activity, attitude, sentiment, or pattern in the data.

Prove to yourself that you can analyze data with crosstabs and check findings for spuriousness. For the data in Box 3, operationally define (a) “low” Family Income as below $26,875 and (b) “high” Family Income as greater than $26,875. Transform the summated variable “Parents’ Edu- cation” (Father’s Education + Mother’s Edu- cation; per what was done in Box 7) into a similar

dichotomy, defining (a) “low” as equal to or less than 24 years of schooling, and (b) “high” as greater than 24 years. Then, crosstabulate Family Income by Parents’ Education (putting the values of Family Income on the rows, and writing in a footer with a prediction and finding). Is there an association between Family Income and Parents’ Education? (Box 9, Criterion 1)?14 Can we argue readily that “education” precedes “income” (Box 9, Criterion 2)?15 Can we give a reasonable interpretation for why we should find the data patterned in this way (i.e., that the more educa- ted are more likely to be prosperous—Box 9, Cri- terion 4)?16 Check your findings for spuriousness (Box 9, Criterion 3), using Sex as the control vari- able. Did the original relationship maintain it- self?17

The Art of Reading Partial Tables To check for spuriousness, you must use at least one control variable—as we have discussed above. In the Advanced exercises in this work- book, you will be asked to test many relation- ships for spuriousness, and you will, therefore, be examining many partial tables. Although it takes many years of experience to be able to recognize the many subtleties that can be obtain- ed in partial tables, you should be able to learn some common patterns very quickly. The pat- terns discussed here will be labeled Causality, Spuriousness, Multivariable Model, Intervening Vari- able, and Interaction Effects.

Causality As noted in Box 9, an X–Y relationship can only be considered causal if one can rule out the possibility that it has not been generated by a third variable—often called an antecedent variable because it precedes in time both X and Y. Examine the X–Y relationship in Footnote #14. Perhaps income is not changing in response to changes in education, but rather both are res- ponding to changes in the antecedent variable Sex. That is, perhaps the actual relationships look like this:

A Primer on Elementary Data Analysis

19

In other words, the relationship between Parents’ Education and Family Income exists not because education is promoting income but simply be- cause both variables are related to the antecedent variable Sex. Observe that there is no arrow between Parents’ Education and Family Income, denoting that there is no causal relationship. More generally, whenever we find a relationship between X and Y that we suspect is causal, we must test to see if the following model is true: This model implies a relationship between Z and X, and between Z and Y; but no relationship be- tween X and Y once Z has been controlled for. Remember, in such a case the only reason the ori- ginal X–Y relationship is found is that Z is not held constant (“controlled for”). Allowed to vary, Z imparts its effects on X and Y; if we are un- aware of Z, we may be fooled into thinking that X and Y are related causally—because changes in X are associated with changes in Y. Thus, in short, to discover whether the X–Y relationship is spurious or causal, one must hold Z constant and reexamine the X–Y relationship. If it maintains itself, we can conclude, at least for the moment, that the X–Y relationship is “nonspurious” or “causal.”

In the example at hand, when we control for Z, that is Sex, we find that the original relationship maintains. The large percentage difference in the original table (35.0%) is found to be in the same direction (“high” education means a higher prob- ability of being in the “high” income row) and to still be quite large in both partial tables—even though it increases for females and decreases for males. Note that these crosstabulations are called “partial tables” because they only deal with part of the sample; the existence of a partial table always implies that at least one Z variable is being held constant. The direction of the relation- ship must be the same if we are to conclude nonspuriousness (e.g., if the original relationship

is “positive,” the relationship in each of the par- tial tables must also be “positive”). However, the percentage difference may be stronger or weaker, as long as it does not weaken greatly or even to- tally disappear. Thus, for example, if the original percentage difference of 35.0 had weakened to, say, 18 in one or both of the partial tables, we would still make the conclusion of “nonspuri- ous” or “causal.” This would not be the case, however, if the percentage difference weakened dramatically (say went down to 5), disappeared all together (reduced to zero), or even started heading in the opposite direction (say, from posi- tive to negative); then, we may have an example of spuriousness.

Spuriousness We have already seen an example of what happens in the partial tables if the origin- al relationship is spurious—that is, we saw above that when Sex was held constant, the relationship between Church Attendance and Delinquency did not maintain itself. Let us use our imagina- tions and fill in percentages in the Footnote #17 table shells appearing in the next column that would have shown the original X–Y relationship (between Parents’ Education and Family Income) to be spurious. Concoct percentages for each ta- ble that would support the spuriousness model sketched above; that is, your percentages must show no relationship between education and in- come in each of the partial tables; moreover, your percentages must reveal that Sex is, indeed, the key independent variable here and is determina- tive of income. Note, too, that your percentages must add up to 100 for each column in each table. For this mental exercise, forget about sample size and cell frequencies; just try to come up with the percentages. Pause now, ponder the question marks, and replace them with eight numbers that will demonstrate the desired (non!)relationship between X and Y. There are no right or wrong an- swers here in terms of finding any exact number or numbers (e.g., one of them does not have to be “40”)—what is important are the relationships that they manifest (and do not manifest). Try hard, then check your answer by examining that given in Footnote #18.

Doing Sociology with Student CHIP

20

Family Income by Parents’ Education for Females18

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High

(> $26,875)

?%

?%

Family Income

Low (< $26,875)

?% ?%

100%

100%

Family Income by Parents’ Education for Males18

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High

(> $26,875)

?%

?%

Family Income

Low (< $26,875)

?% ?%

100%

100%

Multivariable Model We do not just pull an ante- cedent Z variable out of thin air; rather, we choose a particular Z because we think that it will fit the Z–to–X and Z–to–Y model as sketched above. That is, we can make reasonable interpretations as to why we think the world looks this way. For example, we could argue that the historically severe discrimination against women resulted in the mothers of the current generation having acquired lower levels of edu- cation. As for the Z–to–Y relationship, it could be argued that women suffer greater discrimination in the workplace and would therefore be more likely to end up in the “low” income column.

Given that we have strong reason to believe that

Z may be a cause of Y, it should come as no sur- prise that many times when checking for spuri- ousness we find not only X maintaining its effect on Y (thus we reject the spuriousness model) but also Z affecting Y. Indeed, this is the case in the partial tables displayed in Footnote #17. For now, forget about sample size (which is intentionally very small so that you could do your work in this chapter by hand). We have already discussed how, regardless of the value of Z (Sex)—that is to say “independent” of Z—the X–Y (parents’ education  family income) relationship main- tains itself. But take a close look. Regardless of X (level of Parents’ Education), Z (Sex) has consis- tent effects on Y (Family Income). For those res- pondents whose parents possess low levels of education (≤24 years of schooling), males have a (57.1%–0%=) 57.1% greater chance of being in the “high” Family Income column (>$26,875). Simi- larly, for those GSS respondents whose parents possess “high” education (>24 years), males have a (75.0%–66.7%=) 8.3% greater chance of being in high income-earning families. Thus, regardless of Parents’ Education—that is to say “independent” of Parents’ Education—males are more likely to have high Family Income. We may sketch the empirical model evident in these tables as fol- lows:

Now, let us do a mental experiment again. Forget about sample sizes and fill in percentages below that would show a multivariable model—that is, both Parents’ Education and Sex having inde- pendent effects on Family Income (of course you cannot use the “true” percentages that are present in the partial tables in Footnote #17!). Then compare your results to those given in Footnote #19 (again, there are no right or wrong answers with regard to any exact number or

A Primer on Elementary Data Analysis

21

Family Income by Parents’ Education for Males19

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High

(> $26,875)

?%

?%

Family Income

Low (< $26,875)

?% ?%

100%

100%

Family Income by Parents’ Education for Females19

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High

(> $26,875)

?%

?%

Family Income

Low (< $26,875)

?% ?%

numbers).

100%

100%

Intervening Variable  It’s time to throw a little “monkey wrench” into your thinking. As of this moment, if someone were to ask you “What does it mean if after controlling for Z we find that the original relationship between X and Y weakens greatly or even disappears?”, what would you reply? Think a minute, write down your reply, then compare it to the answer given in Footnote #20.20 But there is another possible interpretation of the partial tables. More specifically, perhaps the world looks like this:

That is, a change in X produces a change in Z, which, in turn, produces a change in Y. Now think: If this model is true, what will happen to the relationship between X and Y if Z is held con- stant? Pause here to figure this out (stop read-

ing!). If you said “The X–Y relationship will weaken greatly or disappear,” you are correct! If a change in X produces a change in Y via changes in Z, and if Z is not allowed to vary, then X cannot have any effect on Y. At the very least, those effects of X on Y that occur via Z will be absent, and the X–Y relationship will at mini- mum weaken (i.e., reveal smaller percentage dif- ferences in the partial tables).

Empirically, that is to say, with what is revealed in the partial tables, one cannot distinguish between the situation where Z is the antecedent variable (spuriousness) and where Z is the intervening variable. Then how do we know whether the partial tables above for Church Attendance and Delinquency by Sex demonstrate that the X–Y relationship is spurious, as opposed to demonstrating that Z simply intervenes between X and Y, and that the X–Y relationship is causal—with the effects of X on Y being trans- mitted via Z? Again, the partial tables cannot answer this question. What does answer it is simply thinking logically about the world. Which makes more sense: that Church Attendance causes Sex, which in turn causes a Delinquency Problem (X—Z—Y)? or that Sex determines both the odds of one’s going to church as well as the odds of one’s having a delinquency problem (Z– to–X and Z–to–Y, with no relationship between X and Y, i.e., Spuriousness.)? Of course, the first alternative makes no sense (one’s “Sex” cannot be caused by one’s “Church Attendance”), so we would say that our partial tables have revealed that the X–Y relationship is spurious. Similarly, the partial tables you created for Footnote #17 would be indicative of spuriousness, not that Z was an intervening variable between X and Y.

Interaction Effects  Sometimes when checking for spuriousness, the partial tables will reveal that the relationship has maintained itself in at least one partial table and has weakened greatly or disappeared in at least one other partial table. What should we make of this? On the one hand, we have argued that if the relationship maintains itself then we have evidence for causality. On the other hand, we have contended that if the rela- tionship weakens greatly or disappears then we have evidence for spuriousness (or that Z inter-

Doing Sociology with Student CHIP

22

Family Income by Parents’ Education for Males21

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High

(> $26,875)

?%

?%

Family Income

Low (< $26,875)

?% ?%

100%

100%

Family Income by Parents’ Education for Females21

venes between X and Y). In brief, here is the an- swer: We would have evidence of causality but only under certain conditions, that is, only given certain values of Z. Scientists label such a situa- tion “interaction”—in other words, X and Z “interact” in producing their effects on Y. Be- cause our primary focus is on the X–Y relation- ship, in our examination of the partial tables we are especially interested in seeing how the effects of X on Y vary by the value of Z. Let us do one last mental experiment. In the following tables, fill in percentages that would show that the effect of X on Y depends upon the value of Z; more specifically, conjure up percentages that would show Family Income to be sensitive to changes in Parents’ Education for males but not for females. (Again, there are no right or wrong answers in terms of designating any exact percentages— rather the problem is to invent percentages that show interaction in the manner specified.)

Parents’ Education Low

(≤ 24 Yrs)

High (> 24 Yrs)

High

(> $26,875)

?%

?%

Family Income

Low (< $26,875)

?% ?%

100%

100%

The discovery of interaction presents us with special problems of interpretation. You must be able to think up an interpretation that would ac- count for X affecting Y for some values of Z but not for others. Developing such an interpretation is much more than simple mental gymnastics— for this is the process of science, including social science. When anomalous or unexpected findings arise, findings that challenge our views of the world, we must try to interpret them. We then must go out and test our new interpretations, refining them as necessary. Try your hand at the process of science now: come up with a brief interpretation that would account for our finding that Parents’ Education was predictive of Family Income, but only for males.22

Endnotes for A Primer on Elementary Data Analysis 1. James A. Davis, Tom W. Smith, and Peter V. Marsden, General Social Surveys, 1972–2006 (Chicago:

NORC, March 2007). Distributed by Roper Public Opinion Research Center (Storrs, CT). Box 3 contains all 19-year olds surveyed—in the spring of 2000—for whom information was available on the educational levels of their mothers and fathers.

2. Answer: the sum of the Mother’s Education column = 231; this sum divided by 18 equals 12.83.

3. Answer: the new sum of 304 divided by 18 equals 16.89.

4. Answer: The values for Family Income need to be reordered to put them in a low-to-high ranking:

A Primer on Elementary Data Analysis

23

1,500 16,250 55,000

7,500 16,250 82,500

7,500 21,250 100,000

9,000 32,500 100,000

9,000 45,000 55,000

16,250 55,000 82,500

Thus, the average of the middle two values is 26,875 (21,250 + 32,500 = 53,750; divided by 2 = 26,875), which is the median for Family Income.

5. The values for Family Income for CaseIDs 1–5, 8, 9, 12, 13, 16, and 17 need to be rank-ordered:

1,500 21,250

7,500 55,000

7,500 55,000

9,000 100,000

9,000 111,000

16,250

For an odd-numbered distribution, the median is the middle value after rank ordering; thus, it is $16,250 here.

6. Answer: Aries, Leo, Libra, Scorpio and Taurus appear only once. Aquarius, Cancer, Capricorn, Gemini, and Virgo appear twice. Sagittarius appears the most often, three times; therefore, it is the mode.

7. Answer: 18–8 = 10.

0

2

4

6

8

10

12

Fair Good Excellent

Health Status

Fr eq

ue nc

y

8.

Doing Sociology with Student CHIP

24

9. The scatterplot confirms what you may have already presumed: Father’s Education and Family Income are positively associated: those from better-educated backgrounds tend to be in families with higher levels of income. (This relationship is found consistently in national survey data).

Father’s Education

20181614121086

Fa m

ily In

co m

e

120000

100000

80000

60000

40000

20000

0

(r = . 52)

10. As a test of statistical significance for crosstabs, chi-square compares the actual cell frequencies with the expected cell frequencies if there were no relationship between X and Y. The larger the sample size, the larger the table size, and the larger the difference between the percentage of one value of X possessing a particular value of Y and the percentage of another value of X possessing that value of Y, the more likely the value of chi-square will be statistically significant. See any elementary statistics textbook for a more detailed discussion of chi-square (e.g., Chapter 9 in Jack Levin and James Fox, Elementary Statistics in Social Research; Boston: Allyn & Bacon, 2004).

11. Answers: Yes, there is a relationship (a change in X yields a 44.5% difference in being in the “Excellent” Health Status category). Even though the sample size is relatively small (N=18), this relationship meets conventional standards of statistical significance (the probability of it being obtained by chance alone is less than .05). Indeed, the relationship also rings of substantive significance; more specifically, I made the prediction that those from a higher-income background would more likely report being in excellent health because of their greater access to medical care, as well as the likelihood that their jobs involve less physical risk (exposure to machinery, environmental pollutants, inclement weather, and so on.).

A Primer on Elementary Data Analysis

25

Health Status by Family Income

Family Income

Low

(< 26,875) High

(> 26,875)

Health

Family Income by Parents’ Education

Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High (> $26,875)

36.4%

(4)

71.4%

(5) Family

Income

Low (< $26,875)

63.6% (7)

28.6% (2)

100%

100% (11) (7) N=18

Prediction: Parents’ Education and Family Income are positively related. Finding: Moderately confirmatory, i.e., those respondents whose parents possess a “high” amount of education (>24 yrs) have a (71.4%–36.4%=) 35.0% greater chance of having a “high” (> $26,875) Family Income.

12. Rodney Stark, Sociology 3rd ed. (Belmont, CA: Wadsworth, 1990), pp. 93–95.

13. Despite the objections of some nonquantitative sociologists, “the moral for qualitative researchers is the same. When two variables look correlated, especially when you think they are causally associated, wait a beat, and consider whether some third variable might be underlying/influencing/causing them both” (Matthew B. Miles and A. Michael Huberman, Qualitative Data Analysis. Beverly Hills, CA: Sage, 1984, p. 239).

14. Answer: Yes, there is a moderately strong association between the dichotomized Father’s Education variable and Family Income. Individuals whose fathers have 13 or more years of schooling have a 40% greater chance of being in the “high” Family Income (>$56,250) category. Not surprisingly, even when larger samples are analyzed (e.g., all GSS respondents), the relationship remains highly significant.

Original Table (has no controls; other variables therefore may be confounding the relationship, per- haps even to the point of making it spurious):

15. Answer: Yes, it is highly plausible to argue that “education” precedes “income.” Most individuals com- plete their schooling before embarking on full-time employment.

Excellent

11.1%

(1)

55.6%

(5) Status

Fair/Good

88.9% (8)

44.4%

(4) 100%

(9)

100% (9) N=18

Prediction: Family Income and the Health Status are positively related. Finding: Moderately confirmatory, e.g., those respondents with “high” Family Income (>$26,875) have a (55.6%–11.1%=) 44.5% greater chance of reporting themselves to be in “Excellent” health.

Doing Sociology with Student CHIP

26

16. Answer: Yes, the relationship is plausible. One of the more powerful interpretations: Better educated individuals can compete better in the job market. Being a high school graduate and having gone on to college demonstrate to potential employers the individual’s willingness to learn and to persevere— qualities highly predictive of productivity and therefore highly valued by employers. Contrast this rela- tionhip with one that might appear between Zodiac Sign and Family Income. Even if we were to find a significant relationship (say, individuals born under summer signs having a 75% greater chance of being in the “high” income category compared to individuals born under winter signs), would we be able to meet Criterion 4 in Box 9? We could not do so without a revolution in our current thinking about the rationales underlying pay scales and the value of labor.

17. Partial Tables #1 and #2 (controlling for, or holding constant, Sex; if the relationship in the original table maintains itself, we will have strong evidence for nonspuriousness—i.e., for a causal relationship between Parents’ Education and Family Income):

Family Income by Parents’ Education Family Income by Parents’ Education for Males for Females

Parents’ Education Parents’ Education Low High

(≤ 24 Yrs) (> 24 Yrs)

High

(> $26,875)

57.1%

(4)

75.0%

(3) Family

Income

Low (< $26,875)

42.9% (3)

25.0% (1)

100%

100% (7) (4) N=11

Low High (≤ 24 Yrs) (> 24 Yrs)

0% 66.7% High

(> $26,875) (0) (2)

Family Income

Low 100.0% 33.3% (4) (1) (< $26,875)

100% 100% (4) (3) N=7 Finding: The partial tables reveal that the direction of the original relationship between Parents’ Education and Family Income has maintained itself. Even though the relationship has strengthened slightly for females and reduced moderately for males, we can still conclude that the Parents’ EducationFamily Income relationship is nonspurious or causal.

Family Income by Parents’ Education Family Income by Parents’ Education for Males for Females

Parents’ Education Parents’ Education Low High Low High

(≤ 24 Yrs) (> 24 Yrs) (≤ 24 Yrs) (> 24 Yrs)

High

(> $26,875)

70%

70%

Family Income

Low (< $26,875)

30%

30%

100%

100%

High 60% 60%

(> $26,875) Family

Income Low

(< $26,875) 40%

40%

100%

100%

18. There are no “right” and “wrong” answers here; the key is that you have reduced the original percent- age difference from 35.0 to zero. Note very carefully the wording of the Finding below:

Finding: The partial tables reveal that the original positive relationship between Parents’ Education and Family Income has not maintained itself; we may therefore conclude that this relationship was spurious or non-causal.

A Primer on Elementary Data Analysis

27

19. Again, there are no “right” and “wrong” answers here; the key is that you have maintained both the direction (positive!) and approximate strength of the original percentage difference (which was 35.0). Note very carefully the wording of the Finding below.

Family Income by Parents’ Education Family Income by Parents’ Education for Males for Females

Parents’ Education Parents’ Education Low High (≤ 24 Yrs) (> 24 Yrs)

High

(> $26,875)

40%

90%

Family Income

Low (< $26,875)

60%

10%

100%

100%

Low High (≤ 24 Yrs) (> 24 Yrs)

High 20% 60%

(> $26,875) Family

Income Low 80% 40%

(< $26,875)

100% 100%

Finding: The partial tables reveal that the original positive relationship between Parents’ Education and Family Income has maintained itself in both tables; we may therefore conclude that this relationship is nonspurious or causal. Moreover, the tables reveal the independent-variable effects of Sex on Family Income: for those whose parents possess less than 24 years of education, males have a (40%–20%=) 20% greater chance of being in the “high” (>$26,875) income row; similarly, for those having parents with >24 years of education, males have a (90%–60%=) 30% greater chance of being in the “high” income row. We may sketch the empirical model evident in these tables as follows:

(Obviously, the sketch for the actual results in the footnote #17 would be the same as this one.)

20. As of this moment, you would answer: “If after controlling for Z we find that the original relationship between X and Y weakens greatly or even disappears, then we must conclude that this relationship was spurious or noncausal.

21. Again, there are no “right” and “wrong” answers here; the key is that you have reduced the original percentage-difference from 35.0 to zero or some small number in one of the partial tables, while main- taining both the direction (positive!) and approximate strength of the original percentage-difference (41.7). Note very carefully the wording of the Finding on the next page.

Doing Sociology with Student CHIP

28

Family Income by Parents’ Education for Males

Family Income by Parents’ Education for Females

Parents’ Education Parents’ Education Low

(≤ 24 Yrs) High

(> 24 Yrs)

High

(> $26,875)

40%

90%

Family Income

Low (< $26,875)

60%

10%

100%

100%

Low High (≤ 24 Yrs) (> 24 Yrs)

High 40% 40%

(> $26,875) Family Income

Low 60% 60% (< $26,875)

100% 100%

Finding: The partial tables reveal that the original positive relationship between Parents’ Education and Family Income has maintained itself, but only for males; we may therefore conclude that there is an interaction effect between Parents’ Education and Sex in determining Family Income. We may sketch the empirical model evident in these tables as follows:

22. Perhaps males are evaluated more on the basis of their educations, while females suffer discrimination and are assumed to have lower levels of productivity, regardless of educational attainment. Of course, this is only conjecture that must be rigorously assessed empirically before it is accepted. You need not have made the same conjecture; however, your speculation must be reasonable and must give an accounting of why education would be predictive of family income for males, but not for females.

29

Chapter 1. The Problem of Social Order In the face of seemingly endless social change, what keeps society together and orderly? Major instigators of change include advances in technology, immigration, urbanization, industrialization, economic booms and busts, baby booms and busts, war, and civil unrest. Why do some societies, some groups, and some individ- uals thrive under shifting conditions, while others do not?

How individuals, groups, and societies achieve order in the face of social change was the central question that motivated classical sociology. French scholar Émile Durkheim (1858–1917) argued that order was root- ed in the quality and quantity of social ties among individuals and groups. Along with Germans Karl Marx (1818–1883) and Max Weber (1864–1920), Durkheim was one of the most influential founders of modern sociology. All three had special interest in the political, social, and individual upheavals accompanying the industrialization of Europe in the 18th and 19th centuries. Durkheim found one consequence of the weaken- ing of social ties generated by industrialization—suicide.

Karl Marx viewed coercion and force as central to the maintenance of social order, with the lower classes be- ing compelled to accept the dictates of the upper classes (more specifically, those who control the means of economic production). Unfortunately, many people only associate Marx with the Communist Manifesto and communism; however, he was also a dedicated and distinguished scholar. Das Kapital, his most rigorously researched work, was published in three volumes after his death by his colleagues Frederich Engels and Karl Kautsky. Marx and his colleagues demonstrated how the rich and powerful have used force to reshape society to meet their economic interests at the expense of the less affluent.

Durkheim, Marx, Weber, and most of the other nineteenth century founding fathers of sociology were primarily concerned with the great disruptions that industrialization wrought on European and U.S. society. They showed how social order based on common values and face-to-face interactions was replaced by social order based on interdependence (due to the division of labor), bureaucracy, coercion (e.g., the rise of the legal system, the police, the courts, prisons, and other forms of institutionalized authority), and eventually a new system of values and norms that regulated individuals and gave order to society. The reconstituted social order lasted until the mid-1960s, when—according to Francis Fukuyama—it began to crumble.1 This crumbling was indicated by soaring rates of crime, divorce, and out-of-wedlock births, and by declining levels of trust of major institutions, like government, and of people, in general. The causes, Fukuyama contends, could be found in the rise of the information age, which—by way of a complex causal skein that he details—resulted in the unintentional spread of unbridled individualism and its consequent encouraging of rule-breaking regarding marriage, crime, drug use, and so on. Such behavior destroys a society’s social capital, that is, its shared values—which are, in turn, the prerequisite of all forms of group enterprise (“from running a grocery store to lobbying Congress to raising children”). Group enterprises are the foundation for communities, which both reflect and give orderliness to a society. Fukuyama’s thesis coheres with that of social scientist Robert D. Putnam, whose provocative research includes data supporting the notion that immigration—and the subsequent social diversity it spawns—creates an additional sap on a community’s social capital.2 This is an especially important finding to ponder for those living in the contemporary U.S., Canada, and Western Europe, where immigration is fueling the economy but straining the social fabric.

The rule-breakings that Fukuyama identifies as most destructive concern sex (promiscuity becoming rampant) and the family (especially men deserting their families or their children born out of wedlock);

1 Francis Fukuyama, The Great Disruption (New York: Free Press, 1999). 2 Robert D. Putnam, “E Pluribus Unum: Diversity and Community in the Twenty-First Century,” Scan- dinavian Political Studies, 30:3 (June 2007), pp. 137–174; portions of which are reprinted in Part 1 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010).

Doing Sociology with Student CHIP

30

both of these are rooted in the large-scale movement of women into the paid labor force that began in the 1960s. Women, reasoned many men, no longer needed the traditional male “provider,” as they could earn enough to take care of both themselves and their children.

Fukuyama argues that people are, at the core, rational—they seek to maximize their gains and minimize their pains. As such, a disrupted social order will not be tolerated for long, and he contends that during the past half decade the populaces of the United States and its peer nations have begun to reconstitute the social order, which is reflected in the recent stabilization—and even falling—of divorce, crime, and out- of-wedlock birth rates.

Contemporary sociologists answer the question of order, in part, by using Durkheim’s concept of the quali- ty and quantity of social ties; they also emphasize Marx’s ideas on force and coercion (other answers to the question of social order—including empathy, shared meanings, and societal expectations for proper con- duct—are not discussed here). Durkheim’s ideas are empirically tested in this chapter, while some of the implications of Marx’s thinking are examined in later chapters. We will also explore some of Fukuyama’s fundamental ideas on the increase in social disruption in the United States, the accompanying loss of trust, and whether these have been reversed as we have moved into the 21st century. Finally, we will examine one implication of Putnam’s research—whether immigrant status and neighborhood diversity are related to mistrust.

1. Social Order and Control via Close Social Ties: The Example of Suicide

In his landmark study of suicide,1 Durkheim demonstrated the fundamental precept of social science—that individual behavior can only be fully understood by examining social context, that is, the web of affiliations an individual has with other people. In the following computer exercises, suicide rates are correlated with several of the independent variables emphasized by Durkheim and those who have followed in his theo- retical footsteps.2 “Net migration,” “divorce,” and having “no religion” are conceptualized as measure- ments or indicators of the key explanatory concept of the “quality of social ties.” Areas with high popu- lation turnover (a high net migration rate) contain many individuals who have left friends and family behind; when life becomes stressful, these individuals suffer from the loss of social support and come to have a greater risk of suicide. Similarly, areas with high divorce rates contain many individuals who have had their social relationships torn asunder and have, therefore, lost some of the social support that would see them through emotionally low and stressful periods of their lives. Finally, having a religion, especially when it is accompanied by belonging to a church, synagogue, mosque, or other religious organization can provide individuals with social support during times of emotional need and thereby reduce the risk of the ultimate act of deviance—suicide. File: States2006 (State-level data, U.S. circa 2006)

Info: RegionPopUrbanNetMigAge65College PovertyNoReligVCrimeDivorceSuicide (4) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) Notes: All variables except Region are dichotomized at the median. Unless otherwise noted, all data come

from the Census Bureau or the National Center for Health Statistics.

1 Suicide: A Study in Sociology (New York: Free Press, 1951 [org. 1897]); key portions of which are reprinted in Part 1 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010).

2 For example, see K. D. Breault, Lynn Hampton, and Dustin Brown, “Replicating Suicide in America: Durkheim’s 1897 Theory,” in Part 1 of Carter, ibid.

Chapter 1: The Problem of Social Order

31

Region Northeast (CT, ME, MA, NH, NJ, NY, PA, RI, and VT) Midwest (IL, IN, IA, KS, MI, MN, MO, NE, ND, OH, SD, and WI)

South (AL, AR, DE, FL, GA, KY, LS, MD, MS, NC, OK, SC, TN, TX, VA, and WV) West (AK, AZ, CA, CO, HI, ID, MT, NV, NM, OR, UT, WA, and WY)

Pop (Total population) <4.21 million, >4.21 million

Urban (Percentage of population living in a metropolitan area) <70.2%, >70.2%

NetMig (Percentage change in total population in past decade due to migration) <7.4% ,>7.4%

Age65 (Percentage of the population age 65 years and over) <12.7%, >12.7%

College (Percentage of population, age 25 and over, with a college degree,) <25.5%, >25.5%

Poverty (Poverty rate for families) <9.1%, >9.1%

NoRelig1 (Percentage of population reporting no religion) <16.1%, >16.1%

VCrime2 (Violent crimes—robberies, rapes, assaults, and homicides—per 100,000 pop.) <364.6, >364.6

Divorce (Percentage of population age 15 and over who are divorced) <10.7%, >10.7%

Suicide (Suicides per 100,000 population) <11.9, >11.9 1Barry A. Kosmin, Egon Mayer, and Ariela Keysar, American Religious Identification Survey; available on-line at: http:// www.gc.cuny.edu/faculty/research_briefs/aris.pdf. Note that the percentages for Arkansas and Hawaii were estimat- ed using multiple regression (NoRelig = 2.35*Divorce–.227*Suicide–.222*Blacks+.449*College–15.503, R=.78; R2=.60).

2Uniform Crime Reports (FBI; http://www.fbi.gov/ucr/ucr.htm).

State Region Pop Urban NetMig Age College Poverty NoRelig VCrime Divorce Suicide Alabama South 4,557,808 69.9 8.2 13.2 22.3 13.1 7 426.6 10.9 11.4 Alaska West 663,661 41.5 12.6 6.4 25.5 5.5 17 634.5 11.4 21.0 Arizona West 5,939,292 88.2 30.4 12.7 28.0 10.9 19 504.1 10.8 16.5 Arkansas South 2,779,154 49.4 8.5 13.8 18.8 14.1 14 499.1 11.3 14.0 California West 36,132,15 96.7 11.2 10.7 31.7 10.5 21 551.8 9.6 9.6 Colorado West 4,665,177 83.9 23.1 9.8 35.5 8.6 23 373.5 11.2 16.2 Connecticut Northeast 3,510,297 95.6 -0.2 13.5 34.5 6.2 14 286.3 9.4 7.4 Delaware South 843,524 80.0 13.1 13.1 26.9 7.6 20 568.4 10.6 9.0 Florida South 17,789,86 92.8 16.8 16.8 26.0 9.1 14 711.3 11.9 13.4 Georgia South 9,072,576 69.2 20.2 9.6 27.6 12.0 13 455.5 10.8 11.0 Hawaii West 1,275,194 72.3 7.0 13.6 26.6 7.9 16 254.4 9.3 9.5 Idaho West 1,429,096 39.3 24.3 11.4 23.8 11.1 21 244.9 11.7 15.5 Illinois Midwest 12,763,37 84.9 6.1 12.0 27.4 9.0 16 542.9 9.6 9.1 Indiana Midwest 6,271,973 72.2 7.2 12.4 21.1 7.9 17 325.4 10.6 12.1 Iowa Midwest 2,966,334 45.3 3.3 14.7 24.3 6.8 14 270.9 9.3 10.5 Kansas Midwest 2,744,687 56.6 7.1 13.0 30.0 6.8 17 374.5 10.2 12.6 Kentucky South 4,173,405 48.8 7.4 12.5 21.0 13.9 15 244.9 10.6 12.8 Louisiana South 4,523,628 75.4 3.6 11.7 22.4 14.9 10 638.7 10.6 11.2 Maine Northeast 1,321,505 36.6 2.0 14.4 24.2 9.5 17 103.5 11.1 12.3 Maryland South 5,600,388 92.7 8.2 11.4 35.2 5.8 15 700.5 9.4 8.7 Massachusetts Northeast 6,398,743 96.1 2.6 13.3 36.7 7.1 18 458.8 8.2 6.5 Michigan Midwest 10,120,86 82.2 6.1 12.3 24.4 9.0 17 490.2 10.6 11.0 Minnesota Midwest 5,132,799 70.4 9.1 12.1 32.5 5.3 15 269.6 8.5 9.7 Mississippi South 2,921,088 36.0 7.5 12.2 20.1 17.6 7 295.1 9.9 12.1 Missouri Midwest 5,800,310 67.8 6.9 13.3 28.1 8.7 17 490.5 11.2 12.1 Montana West 935,670 33.9 10.5 13.7 25.5 10.5 19 293.8 10.7 19.9 Nebraska Midwest 1,758,787 52.6 5.6 13.3 24.8 7.7 10 308.7 9.4 11.7 Nevada West 2,414,807 87.5 50.6 11.2 24.5 10.0 21 615.9 14.2 19.8 New Northeast 1,309,940 59.9 8.3 12.1 35.4 5.5 19 167.0 11.1 10.2 New Jersey Northeast 8,717,925 100.0 5.1 12.9 34.6 6.5 16 355.7 7.8 6.3 New Mexico West 1,928,384 56.9 14.8 12.1 25.1 15.9 18 687.3 11.9 19.1 New York Northeast 19,254,63 92.1 1.1 13.0 30.6 11.1 15 441.6 7.9 6.3 North Carolina South 8,683,242 67.5 15.4 12.1 23.4 12.1 11 447.8 9.0 11.8

Doing Sociology with Student CHIP

32

State Region Pop Urban NetMig Age College Poverty NoRelig VCrime Divorce Suicide North Dakota Midwest 636,677 44.2 -0.8 14.7 25.2 8.0 4 79.4 8.4 14.2 Ohio Midwest 11,464,04 81.2 3.8 13.3 24.6 10.0 17 341.8 10.7 11.2 Oklahoma South 3,547,884 60.8 6.8 13.2 22.9 12.0 15 500.5 11.8 14.3 Oregon West 3,641,056 73.1 16.7 12.8 25.9 10.4 28 298.3 12.3 14.4 Pennsylvania Northeast 12,429,61 84.6 0.9 15.3 25.3 8.9 13 411.1 8.7 10.7 Rhode Island Northeast 1,076,189 94.1 -1.3 13.9 27.2 10.4 17 247.4 9.5 7.9 South Carolina South 4,255,083 70.0 11.5 12.4 24.9 12.5 8 784.2 9.8 10.6 South Dakota Midwest 775,933 34.6 5.3 14.2 25.5 8.4 8 171.5 8.9 12.2 Tennessee South 5,962,959 67.9 12.4 12.5 24.3 11.6 9 695.2 11.6 13.2 Texas South 22,859,96 84.8 18.0 9.9 24.5 13.5 12 540.5 10.1 11.0 Utah West 2,469,585 76.5 23.6 8.7 30.8 8.2 18 236.0 8.4 16.1 Vermont Northeast 623,050 27.8 5.5 13.0 34.2 5.6 24 112.0 11.4 14.1 Virginia South 7,567,465 78.1 11.0 11.4 33.1 7.2 13 275.6 9.2 10.9 Washington West 6,287,759 83.1 18.3 11.3 29.9 9.4 27 343.8 11.7 13.3 West Virginia South 1,816,856 42.3 0.7 15.3 15.3 14.1 15 271.2 10.4 14.8 Wisconsin Midwest 5,536,201 67.9 7.3 13.0 25.6 7.6 15 209.6 9.3 11.3 Wyoming West 509,294 30.0 5.7 12.1 22.5 7.7 21 229.6 12.6 20.7

Basic

1. According to Durkheim, what should we find if we crosstabulated Suicide (Y) by Divorce (X)? 2. Open the States2006 file and test your expectation using both a crosstab1 and a plot (after doing the

Crosstab, select the Line Chart option under the Table command and highlight the “>11.9” category). Is your finding confirmatory?

1 Unless otherwise noted, you should put Y on the rows, X on the columns, and use the Percent Down option for all tables in this workbook. Given this arrangement, percentages should be compared by row to assess the strength of the X–Y relationship. For a 2(row)–by–2(column) table, only one row of percentages needs to be examined. Although more comparisons can be made, for tables of larger dimensions (e.g., 3×4), you usually will need to examine only one row (e.g., the “high” value of Y) and calculate only one percentage-difference (e.g., the “high” versus the “low” value of X).

Chapter 1: The Problem of Social Order

33

Crosstab: Suicide / Divorce

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

<11.9

100% 100% N=

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = ) Prediction:

Finding:

Answer/Discussion

Doing Sociology with Student CHIP

34

3. According to Durkheim, what should we find if we crosstabulated Suicide (Y) by NoRelig (X)? 4. Test your expectation using both a crosstab and a plot (after doing the Crosstab, select the Line Chart

option under the Table command and highlight the “>11.9” category). Is your finding confirmatory?

Crosstab: Suicide / NoRelig

(Table: Percent Down)

<16.1%

>16.1%

Total

>11.9

<11.9

100% 100% N=

(Percentage difference between >16.1% and <16.1% for the “>11.9” row = ) Prediction:

Finding:

Chapter 1: The Problem of Social Order

35

Answer/Discussion 5. According to Durkheim, what should we find if we crosstabulated Suicide (Y) by NetMig (X)? 6. Test your expectation using both a crosstab and a plot (after doing the Crosstab, select the Line Chart

option under the Table command and highlight the “>11.9” category). Is your finding confirmatory?

Crosstab: Suicide / NetMig

(Table: Percent Down)

<7.4%

>7.4%

Total

>11.9

<11.9

100% 100% N=

(Percentage difference between >7.4% and <7.4% for the “>11.9” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

36

Answer/Discussion 7. Durkheim developed the concept of “anomie” to help explain suicide rates. Anomie means being with-

out norms or in a state of normative confusion (“norms” are the everyday rules of living that guide our interactions with others). Anomie can be used to characterize individuals, groups, or societies. Durkheim believed that it was caused by the weakening of the quality and quantity of social ties—such weakening being created by rapid social change, divorce, and other threats to the solidarity of groups to which individuals belong. Thus, if you think about it, not only should rates of divorce, having no religion, and population turnover be predictive of suicide, but also of crime and other forms of social “deviance.” What should we find if we crosstabulate:

(a) VCrime (Y) by Divorce (X)?

(b) VCrime (Y) by NoRelig (X)?

(c) VCrime (Y) by NetMig (X)?

Chapter 1: The Problem of Social Order

37

8. Test your expectations using both crosstabs and plots (after doing each Crosstab, select the Line Chart option under the Table command and highlight the “>364.6” category). Are your findings confirmatory?

(a) Crosstab: VCrime / Divorce

(Table: Percent Down)

<10.7%

>10.7%

Total

>364.6

<364.6

100% 100% N=

(Percentage difference between >10.7% and <10.7% for the “>364.6” row = ) Prediction:

Finding:

Answer/Discussion

Doing Sociology with Student CHIP

38

(b) Crosstab: VCrime / NoRelig

(Table: Percent Down)

<16.1%

>16.1%

Total

>364.6

<364.6

100% 100% N=

(Percentage difference between >16.1% and <16.1% for the “>364.6” row = ) Prediction:

Finding:

Answer/Discussion

Chapter 1: The Problem of Social Order

39

(c) Crosstab: VCrime / NetMig

(Table: Percent Down)

<7.4%

>7.4%

Total

>364.6

<364.6

100% 100% N=

(Percentage difference between >7.4% and <7.4% for the “>364.6” row = ) Prediction:

Finding:

Answer/Discussion

Doing Sociology with Student CHIP

40

Advanced

9. Perhaps the relationship between divorce and suicide uncovered in question #1 above is spurious. More specifically, perhaps NetMig is determining both Divorce and Suicide. If this is so, controlling for NetMig might make the DivorceSuicide association disappear. Sketch this alternative model—that is, that Divorce and Suicide are only related because of their common association with NetMig.

10. Test the alternative model in #9 by crosstabulating Suicide (Y) by Divorce (X) and controlling for Net-

Mig (Z). Was the original Suicide/Divorce relationship spurious? Is a multivariable model apparent— that is, do both Divorce and NetMig have independent effects on the Suicide rate? Refer to the correct percentage differences to support your answer.

Original Table Crosstab: Suicide / Divorce

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

<11.9

100% 100% N=

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = ) Prediction:

Finding:

Chapter 1: The Problem of Social Order

41

Partial Table #1 Crosstab: Suicide / Divorce

Control: NetMig (<7.4%)

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

<11.9

100% 100% N=

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = )

Partial Table #2 Crosstab: Suicide / Divorce

Control: NetMig (>7.4%)

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

<11.9

100% 100% N=

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = )

Summary of findings for the partial tables:

Doing Sociology with Student CHIP

42

Chapter 1: The Problem of Social Order

43

2. Social Characteristics of Happy Individuals

Durkheim’s thinking implies that married people—contrary to popular stereotypes and jokes—may be hap- pier, on average, than nonmarried individuals. A spouse and children provide sources of emotional support during times of stress and can magnify the pleasure of more joyous times. File: Happy2006 (Social Context of Happiness. Source: 2004 & 2006 GSS)

Info: RaceHispanicSexAgeSocClassHowReligMarriedHealthHappy (3) (2) (2) (4) (3) (3) (4) (3) (3)

Race White, Black, Other

Hispanic No, Yes

Sex Male, Female

Age 18–39, 40–64, 65+

SocClass (Respondent’s socioeconomic status, recoded as:1) Lower (17.1–33.0), Lower-Middle (33.1–44.2), Upper-Middle (44.3–64.1), Upper (64.2–97.2)

HowRelig (How religious is respondent?) Not Very, Somewhat, Strongly

Married Never Married, Div–Sep, Widowed, Married

Health (Self-reported level of personal health) Poor–Fair, Good, Excellent

Happy (Self-reported level of personal happiness) Not Too Happy, Pretty Happy, Very Happy

Basic

1. Crosstab and plot happiness by marital status (after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Very Happy” category). Do these variables correlate in the predicted direction? That is, do married individuals tend to be happier? (For this and the questions that follow, concentrate your discussion on the ”Very Happy” row.)

1 Socioeconomic index value based on respondent’s education, income, and occupational prestige; scores range from a minimum value of 17.1 to a maximum value of 97.2; see General Social Surveys, 1972–2006: Cumulative Codebook (Storrs, CT: Roper Center for Public Opinion Research, University of Connecticut, March 2007), p. 2235.

Doing Sociology with Student CHIP

44

Crosstab: Happy / Married

(Table: Percent Down)

Never Married

Div–Sep

Widowed

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% 100% 100% N=

(Percentage difference between Married and Never Married on “Very Happy” = )

Prediction:

Finding:

Chapter 1: The Problem of Social Order

45

Answer/Discussion

Doing Sociology with Student CHIP

46

Advanced

2. Sketch the model that would show the relationship between marital status and happiness to be spuri- ous, using social class as your antecedent variable. Defend this alternative model.

(a) Model sketch:

(b) Interpretation of the SocClass–Married relationship:

(c) Interpretation of the SocClass–Happy relationship: 3. Let’s test the model you have sketched and defended above. To simplify your analyses, do the follow-

ing modifications to your Happy2006 data file: (1) Click on Modify, then Combine, and combine the NeverMarried, Div-Sep, and Widowed categories of the variable Married; label this new combined value “Not Married.” (2) Click on Modify again, but this time Combine the lower two categories of SocClass (Lower; Lower-Middle) and label these combined categories “Less Prosperous”; repeat this same process, labeling the upper two categories of SocClass (Upper-Middle; Upper) as “More Pros- perous.” Note that on some computers these data modifications might take several seconds and your computer might hang up for a very short while. Verify that you have done these modifications correctly by clicking on Info; you should have “2” values for SocClass, as well as “2” values for Married (note that you had 4 values for each of your original variables). Now, crosstab Happy (Y) by Married (X), con- trolling for SocClass (Z). Does the relationship between marital status and happiness maintain itself when controlling for social class? Is a multivariable model evident—that is, do both Married and SocClass have independent effects on Happy? Refer to the correct percentage-differences to support your answer.

Chapter 1: The Problem of Social Order

47

Original Table

Crosstab: Happy / Married

(Table: Percent Down)

Not Married

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% N=

(Percentage difference between Married and Not Married on “Very Happy” = )

Prediction:

Finding:

Partial Table #1

Crosstab: Happy / Married

Control: SocClass (Less Prosperous)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% N=

(Percentage difference between Married and Not Married on “Very Happy” = )

Doing Sociology with Student CHIP

48

Partial Table #2

Crosstab: Happy / Married

Control: SocClass (More Prosperous)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% N=

(Percentage difference between Married and Not Married on “Very Happy” = ) Summary of findings for the partial tables:

Chapter 1: The Problem of Social Order

49

4. Let’s explore the marital statushappiness relationship more deeply. Perhaps the relationship be- tween these two variables can be explained by the intervening variable of personal health; more specifically, perhaps married people are more likely to have better health, and having better health, in turn, increases ones chances of being happy. Sketch the model that would show that the relationship of Married and Happy can be accounted for, in part, by Health.

5. If the model you sketched in #4 is valid, then what should happen to the Happy/Married relation-

ship after we control for Health? (Hint: you may want to go back to the chapter A Primer on Elemen- tary Data Analysis and review the section entitled “The Art of Reading Partial Tables.”) Realistically, you can assume that personal health would be only one of the intervening variables (causal mechan- isms) connecting marital status with happiness, so holding it constant should not be devastating; but what do you think doing so should cause to happen to the original relationship between Happy and Married?

6. Test your expectation in #5 by crosstabulating Happy by Married and controlling for Health. Use the

transformed version of your Married variable. How strongly was your expectation confirmed?

Partial Table #1

Crosstab: Happy / Married

Control: Health (Poor–Fair)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% N=

(Percentage difference between Married and Not Married on “Very Happy” = )

Doing Sociology with Student CHIP

50

Partial Table #2

Crosstab: Happy / Married

Control: Health (Good)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% N=

(Percentage difference between Married and Not Married on “Very Happy” = )

Partial Table #3

Crosstab: Happy / Married

Control: Health (Excellent)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

Pretty Happy

Not too Happy

100% 100% N=

(Percentage difference between Married and Not Married on “Very Happy” = )

Answer/Discussion

Chapter 1: The Problem of Social Order

59

Exploratory

I. Using any of the CHIP data files for this chapter (States2006, Happy2006, or Trust2006), state a hypoth- esis relating an X and a Y variable that have not already been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy-

pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 4×4 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

60

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic-

tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabulating Y by X and controlling for Z—using the

appended table shells.

Chapter 1: The Problem of Social Order

61

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z—using the

appended table shells. (Note: you may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

Chapter 1: The Problem of Social Order

63

ANSWERS FOR SELECTED CHAPTER 1 EXERCISES

It is expected that you will learn fairly quickly how to use Student CHIP and have the ready capacity to do crosstabulations and plots (transcribing your findings from your computer screen to your workbook). This is not the true test, however, of what you are learning about sociology. The true test comes in your ability to concisely and incisively fill in the “white space” provided for your written answers. To give you an idea of what is expected of you, selected questions from above are answered in this section. These answers will pro- vide “exemplars” that you can refer to from time to time as your “white-space” abilities blossom. Please note that this section is not repeated in the remaining chapters of this workbook.

Social Order and Control via Close Social Ties: The Example of Suicide

Basic

1. According to Durkheim, what should we find if we crosstabulated Suicide (Y) by Divorce (X)?

As noted in the introduction to this chapter, “divorce” is conceptualized as a measurement or indicator of the key explanatory concept of the “quality of social ties.” Areas with high divorce rates contain many individuals who have had their social relationships torn asunder and have, therefore, lost some of the social support that would see them through emotionally low and stressful periods of their lives.” Thus, we would expect to find a positive relationship between Suicide and Divorce (i.e., those states with “high” divorce rates should also have “high” suicide rates).

2. Open the States2006 file and test your expectation using both a crosstab and a plot (after doing the Cross-

Tab, select the Line Chart option under the Table command and highlight the “>11.9” category.) Is your finding confirmatory?

Crosstab: Suicide / Divorce

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

27.6

81.0

50.0 <11.9

72.4

19.0

50.0

100% 100% N= 50

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = 53.4) Prediction: Divorce and Suicide are positively related.

Finding: Strongly confirmatory, i.e., those states with “high” Divorce rates (>10.7%) have a (81.0%–27.6.0%=) 53.4% greater chance of having a “high” Suicide rate (>11.9).

In general, you will make a prediction that X is either positively related to Y, or that X is negatively related to Y. Remember, “positive” implies that the two variables increase and decrease together (as X increases, so does Y; as X decreases, so does Y), and “negative” implies that the two variables change in opposite directions (increases in one are associated with decreases in the other). If X is a nominal variable, then one must specify a category of X when making the prediction (see p. 7 in the preceding chapter entitled “A Primer on Critical Reading”). Recall that a “nominal” variable is one whose values cannot be rank-ordered “more than” or “less than.” Thus, in the General Social Survey, “Sex” would be a nominal variable (“female” is not more than “male,” or vice versa), while “Income” is not (“$20,000” is less than ”$30,000” is less than “$40,000”). Variables having values that can be rank ordered, that is, that have gradients or weight, are called “quantitative” variables. Thus, if our

Doing Sociology with Student CHIP

64

prediction had involved the X variable “Sex” and the Y variable “Income” we would not say “Sex is positively related to Income” (this does not make any sense); rather, we would have to specify a value of Sex and say something like: “Males are more likely to be in the ‘high’ row of Income” (or some such phrasing, e.g., “Men are more likely to be ‘high’ Income earners”).

Your finding should be stated in the form:

 MODIFIER (Weakly, Moderately, or Strongly), followed by the word  CONFIRMATORY (or NONCONFIRMATORY), followed by either  E.G., or I.E., followed by a  PERCENTAGE DIFFERENCE (a change in X produces how much of a

change in Y; do the subtraction from the right column to the left column; if this difference is a positive number, then you have evidence of a positive relationship; if the difference is negative, you have evidence of a negative relationship)

Thus, say that the percentage difference had been “3.4” instead of 53.4, then our finding would read:

Weakly confirmatory, i.e., those states with “high” Divorce rates (>10.7%) have a 3.4% greater chance of having a “high” Suicide rate (>11.9).

Although you can assess the degree of confirmation by calculating contingency table statistics such Chi Square, you should be developing your own internal standards of assessing whether a percentage difference is confirmatory or nonconfirmatory and the degree to which it is one of these (e.g., weakly, moderately, or strongly). As sample size increases and as the percentage-difference becomes greater, the degree of confirmation becomes stronger. Thus, for example, given a relatively small sample of, say, 50, and given a relatively small percentage-difference of, say 7, we would give the assessment “weakly” confirmatory (or nonconfirmatory, if the relationship is the wrong direction—say negative when it was predicted to be positive). If the relationship is in the expected direction, then it is confirmatory. Thus, for example, if we hypothesize that it is positive and it turns out to be positive, then the finding is confirmatory. Again, the degree to which it is confirmatory will depend upon sample size and the strength of the relationship (percentage-difference). When we use the General Social Survey, sample size is often in the thousands, thus a relatively small percentage-difference will be both statistically and substantively significant (at least “moderately” confirmatory)—that is, any percentage-difference above 9 or 10 should be considered important. When we use U.S. states (where N=50), then we need larger percentage-differences to give an assessment of “moderately” or “strongly” confirmatory—well into double digits (15, 20, 25 percentage points).

Whether you use the connecting abbreviation “i.e.” versus “e.g.” depends upon the dimensions of your crosstab. Two-by-two (2 row / 2 column) crosstabs have only one correct percentage-difference (compar- ing columns on either row yields the same percentage-difference), and thus you use “i.e.” (“that is”). For crosstabs of larger dimensions (anything greater than 2×2, e.g., 3×2 or, say, 3×3), you would use “e.g.” (“for example”) because there is more than one possibility for comparing percentages. In these situations, you are instructed which percentage-difference to compute; in general, it will be the difference between the “highest” and “lowest” values of X on the “highest” value Y.

For the second part of this question involving the plot, note that you should have hand-drawn the straight line between the percentage value for Suicide “>11.9” for the Divorce value of <10.7% and the percentage value for Suicide “>11.9” for the Divorce value of >10.7%:

Chapter 1: The Problem of Social Order

65

An

I o

T wor “

( a we ha finding—get d A

I o

T wor “

( a we ha finding—get d A

swer/Discussion swer/Discussion

f you acquire the ability to apply this protocol whenever you are confronted by data, you will increase the dds of your success in life, in general, and in the work world, in particular. f you acquire the ability to apply this protocol whenever you are confronted by data, you will increase the dds of your success in life, in general, and in the work world, in particular.

his protocol will usually (but not always) produce success for your Answer/Discussion sections in this kbook. The amount of “white space” allotted your writing is small, so you must learn to be concise and

cut to the chase.” Let’s apply it now to our findings on Suicide and Divorce:

his protocol will usually (but not always) produce success for your Answer/Discussion sections in this kbook. The amount of “white space” allotted your writing is small, so you must learn to be concise and

cut to the chase.” Let’s apply it now to our findings on Suicide and Divorce:

It was hypothesized that Divorce and Suicide would be positively associated; indeed, the crosstab and plot both reveal a strong positive relationship. This was expected because “areas with high divorce rates contain many individuals who have had their social relationships torn asunder and have, therefore, lost some of the social support that would see them through emotionally low and stressful periods of their lives.”

It was hypothesized that Divorce and Suicide would be positively associated; indeed, the crosstab and plot both reveal a strong positive relationship. This was expected because “areas with high divorce rates contain many individuals who have had their social relationships torn asunder and have, therefore, lost some of the social support that would see them through emotionally low and stressful periods of their lives.”

Note how all three points of the protocol are incorporated in the paragraph; note, too, that you will often be ble to pull your interpretation out of the introductory comments preceding each major set of exercises, as

ve done just now; finally, note that the independent variable is stated first in a prediction and in a into the habit of doing this, as you will be less likely to misunderstand or misstate the “causal

irection,” which always flows from X to Y.)

Note how all three points of the protocol are incorporated in the paragraph; note, too, that you will often be ble to pull your interpretation out of the introductory comments preceding each major set of exercises, as

ve done just now; finally, note that the independent variable is stated first in a prediction and in a into the habit of doing this, as you will be less likely to misunderstand or misstate the “causal

irection,” which always flows from X to Y.)

dvanced dvanced 9. Perhaps the relationship between divorce and suicide uncovered in question #1 above is spurious.

More specifically, perhaps NetMig is determining both Divorce and Suicide. If this is so, controlling for NetMig might make the DivorceSuicide association disappear. Sketch this alternative model—that is, that Divorce and Suicide are only related because of their common association with NetMig.

9. Perhaps the relationship between divorce and suicide uncovered in question #1 above is spurious. More specifically, perhaps NetMig is determining both Divorce and Suicide. If this is so, controlling for NetMig might make the DivorceSuicide association disappear. Sketch this alternative model—that is, that Divorce and Suicide are only related because of their common association with NetMig.

A Protocol for Successfully Writing-Up Your Answer/Discussion Section

Whenever you see a table or graphic intended to reveal the relationship between or among variables (whether in this workbook, in other courses, or in the occupational world), you should have the ability to:

(1) state the underlying hypothesis that motivated the particular analysis at hand (e.g., table or graph);

(2) state the finding—and in doing so, assess the degree to which it is in harmony with the hypothesis; and

(3) generate a brief interpretation that makes sense out of the hypothesis and finding.

Doing Sociology with Student CHIP

66

This type of question is asked many times in Doing Sociology with Student CHIP. Note that the model sketch will always look the same: Z–to–X, Z–to–Y, and no link between X and Y.

10. Test the alternative model in #9 by Suicide (Y) by Divorce (X) and controlling for NetMig (Z). Was the original Suicide/Divorce relationship spurious? Is a multivariable model apparent—that is, do both Divorce and NetMig have independent effects on the Suicide rate? Refer to the correct percentage differences to support your answer.

Original Table

Crosstab: Suicide / Divorce

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

27.6

81.0

50.0 <11.9

72.4

19.0

50.0 100% 100% N= 50

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = 53.4) Prediction: Divorce and Suicide are positively related.

Finding: Strongly confirmatory, i.e., those states with “high” Divorce rates (>10.7) have a (81.0%–27.6.0%=) 53.4% greater chance of having a “high” Suicide rate (>11.9).

Partial Table #1 Crosstab: Suicide / Divorce

Control: NetMig (<7.4%)

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

31.6

83.3

44.0 <11.9

68.4

16.7

56.0

100% 100% N= 25

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = 51.7)

Partial Table #2 Crosstab: Suicide / Divorce

Control: NetMig (>7.4%)

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

20.0

80.0

56.0 <11.9

80.0

20.0

44.0

100% 100% N= 25

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = 60.0)

Chapter 1: The Problem of Social Order

67

Summary of findings for the partial tables:

Although the relationship has weakened slightly in the first partial table (the percentage-difference decreasing from 53.4 to 51.7), and has strengthened somewhat in the second partial table (increasing to a percentage-difference of 60.0), it is still strong and in the same direction; we may therefore conclude that it is nonspurious or causal. Moreover, the tables reveal the independent-variable effects of NetMig on Suicide: For “low-divorce” states (<10.7%), “high” NetMig states (>7.4%) have a (20.0%– 31.6%=) 11.6% smaller chance of having a “high” (>11.9) Suicide rate; similarly, for “high-divorce” states (>10.7%), “high” NetMig states have a (80.0%–83.3%=) 3.3% smaller chance of having a “high” (>11.9) Suicide rate; thus, regardless of divorce rate—that is to say, “independent” of divorce rate—NetMig has a negative effect on Suicide. We may sketch the causal model evident in these partial tables as follows:

Let us pause here and occupy ourselves with these tables a little more. More specifically, let us say that the percentages in the second column of the second partial table had come out as below. What would our conclusions be then?

Partial Table #2 Crosstab: Suicide / Divorce

Control: NetMig (>7.4%)

(Table: Percent Down)

<10.7%

>10.7%

Total

>11.9

20.0

93.3

64.0 <11.9

80.0

6.7

36.0

100% 100% N= 25

(Percentage difference between >10.7% and <10.7% for the “>11.9” row = 73.3)

Although the relationship has weakened slightly in the first partial table (the percentage-difference decreasing from 53.4 to 51.7), and has strengthened somewhat in the second partial table (increasing to a percentage-difference of 73.3), it is still strong and in the same direction; we may therefore con- clude that it is nonspurious or causal.

Notice that there is no independent-variable effect of NetMig on Suicide—that is, the effect depends upon the value of Divorce (X); and we thus do not have the extended write-up that we had above for the actual partial table findings. That is, we do not include that part of the finding beginning with “Moreover.” (Now is a good time to go back to pages 18–22 and review the section entitled “The Art of Reading Partial Tables.”)

Social Characteristics of Happy Individuals

4. Let’s explore the marital statushappiness relationship more deeply. Perhaps the relationship between the two variables can be explained by the intervening variable of personal health; more specifically, perhaps married people are more likely to have better health, and having better health, in turn, increases ones chances of being happy. Sketch the model that would show that the relationship of Married and Happy can be accounted for, in part, by Health.

Doing Sociology with Student CHIP

68

5. If the model you sketched in #4 is valid, then what should happen to the Happy/Married relation-

ship after we control for Health? (Hint: you may want to go back to the chapter A Primer on Elemen- tary Data Analysis and review the section entitled “The Art of Reading Partial Tables.”) Realistically, you can assume that personal health would be only one of the intervening variables (causal mechan- isms) connecting marital status with happiness, so holding it constant should not be devastating; but what do you think doing so should cause to happen to the original relationship between Happy and Married?

If the model sketched in #4 were completely accurate, then controlling for Health would wipe out the relationship between Married and Happy (that is, percentage difference would be zero) because the effects of Married on Happy are being transmitted via Health—and this can occur only if Health is allowed to vary. Realistically, however, there are many intervening variables (e.g., greater “social support” for married individuals) and thus we would only expect a reduction in the relationship—not its complete disappearance. Why should it reduce? It should do so because we have removed one of the causal pathways by which Married incurs its effects on Happy!

6. Test your expectation in #5 by crosstabulating Happy by Married and controlling for Health. Use the

transformed version of your Married variable. How strongly was your expectation confirmed?

Partial Table #1

Crosstab: Happy / Married

Control: Health (Poor–Fair)

(Table: Percent Down)

Not Married

Married

Total Very Happy

11.8

30.3

19.7 Pretty Happy

52.4

56.0

53.9 Not too Happy

35.8

6.7

26.3

100% 100% N= 649

(Percentage difference between Married and Not Married on “Very Happy” = 18.5)

Partial Table #2

Crosstab: Happy / Married

Control: Health (Good)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

19.3

40.1

29.4 Pretty Happy

64.6

53.6

59.3 Not too Happy

16.1

6.3

11.4

100% 100% N= 1311

(Percentage difference between Married and Not Married on “Very Happy” = 20.8)

Chapter 1: The Problem of Social Order

69

Partial Table #3

Crosstab: Happy / Married

Control: Health (Excellent)

(Table: Percent Down)

Not Married

Married

Total

Very Happy

36.8

57.2

48.0 Pretty Happy

56.1

39.2

46.8 Not too Happy

7.0

3.6

5.1

100% 100% N= 758

(Percentage difference between Married and Not Married on “Very Happy” = 20.4)

Answer/Discussion

With no controls, the percentage difference between those who are “Married” and those who are “Not Married” is 21.8. We hypothesized that part of the effect of marital status on happiness operates through personal health—more specifically, that married individuals are more likely to have better personal health than nonmarried individuals, and that, in turn, better health produces a generally happier life. If this scenario is true, then controlling for personal health should weaken the original relationship that we found between Married and Happy. Indeed, in all three tables the relationship weakened: For those with “Poor– Fair” Health, the percentage difference between Married and Not Married reduced from 21.8 in the Original Table (with no controls) to 18.5; where personal health is “Good” (Partial Table #2), this difference fell slightly to 20.8; and for those reporting “Excellent” Health, there was also a slight reduc- tion—to a 20.4 percentage difference. For the three partial tables, the average percentage-difference on being “Very Happy” for Married versus Not Married individuals is 19.9. This is, of course, less than 21.8 and therefore what we expected to find (a reduction in percentage-difference, but not to the point of reaching “0”). However, given that the reduction was so small, the overall conclusion is that there is only very weak evidence to support the intervening-variable model Married  Health  Happy. Thus, for ex- ample, had the reduction been more like 20 or 30 percent (from 21.8 to, say, a partial-table-percentage- difference average of 17 or 15), we would be much more impressed by this model.

Exploratory

I. Using any of the CHIP data files for this chapter (States2006, Happy2006, or Trust2006), state a hypoth- esis relating an X and a Y variable that have not already been analyzed together.

Examining the Happy2006 data file, we might reasonably hypothesize that social class would be predic- tive of personal health, that is, SocClass and Health are positively related.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that we would expect to find data patterned in the way in which you have predicted.

Some interpretations would include: The social and environmental contexts of the lives of low-income individuals make them more susceptible to illness and disease. They live under less sanitary condit- ions, have less nutritious diets, and are less likely to take preventive health actions such as obtaining routine physical examinations. Poor women are less likely to have prenatal checkups and more likely to have poor diets; the result is an increased probability of having low birth-weight infants, who are at greater risk of dying or surviving with mental or physical imperfections. Moreover, “despite the considerable advances brought about by Medicare and Medicaid, many poor people are not covered by these programs, and some health care still depends on out-of-pocket costs, which the poor usually cannot afford.” Finally, the care that poorer individuals receive is more likely to be of a lower quality: They tend to be treated in hospital outpatient clinics or emergency rooms, where the “continuity of

Doing Sociology with Student CHIP

70

care, follow-up treatment, and patient education are less common than in a physician’s office” (see Thomas J. Sullivan and Kenrick S. Thompson, Introduction Social Problems, 3rd ed., New York: Macmillan, 1993, pp. 126–127; note that any one of these interpretations would suffice given the limited amount of “white space” you have to work with).

IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy- pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 4×4 table shell is simply a starting point.)

Crosstab: Health / SocClass

(Table: Percent Down)

Lower

Lower Middle

Upper Middle

Upper

Total

Excellent

18.9

23.9

31.9

36.7

27.9 Good

46.9

48.4

49.9

47.7

48.2 Poor–Fair

34.2

27.7

18.2

15.7

23.9

100% 100% 100% 100% N= 2,718

(Percentage difference between Upper and Lower for the “Excellent” row = 17.8) Prediction: SocClass and Health are positively related. Finding: Moderately confirmatory, e.g., those individuals with from the Upper social class have a (36.7%–18.9%=) 17.8% greater chance of being in the Excellent row compared to those individuals from the Lower social class.

(Note that in the tables for this set of exploratory exercises, only 3 rows are needed—and you would, of course, not use the 4th row in the table shells that are provided).

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Chapter 1: The Problem of Social Order

71

V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic- tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

For a variety of reasons we might expect whites to have an educational advantage over blacks and many other people of color, especially when we are examining the entire adult population. For example, due to segregation and lack of choice, many African Americans have been forced to go to inner-city schools in which much of the student body is poor. Role models for a student considering going on to college are relatively few, while models for a student dropping out of high school are disproportionately high. In addition, education begets itself— with the sons and daughters of educated parents being more likely to become educated themselves (more opportunity; more encouragement). Because of historical discrimination, African Americans and many other people of color are less likely to have educated parents and thus lack this advantage.

(b) Interpretation of the Z–Y relationship:

We might expect whites to have a higher chance of having excellent health because they have a lower probability of getting Type II diabetes, which is currently epidemic in the U.S., and if they do get the disease, they have fewer complications. Current research indicates that African Americans have a genetic predisposition to acquire the disease (e.g., see http://www.blackhealthcare.com/BHC/ Diabetes/Description.asp).

NOTE that in neither (a) nor (b) was there a direct or causal connection ever made between social class and health. Indeed, we are playing the “devil’s advocate” here and are trying to see whether it is possible to make a persuasive argument for a spurious association between X and Y, not a causal one. Thus, we are conjecturing that the reason we might have found an association between SocClass and Health in the cross- tab in #IV above is that both of these variables happen to vary with Race, not because X (SocClass) is causally related to Y (Health). VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Doing Sociology with Student CHIP

72

Partial Table #1

Crosstab: Health / SocClass

Control: Race (White)

(Table: Percent Down)

Lower

Lower Middle

Upper Middle

Upper

Total

Excellent

20.1

25.1

32.8

38.6

29.8 Good

47.7

48.8

49.2

46.2

47.9 Poor–Fair

32.2

26.1

18.0

15.2

22.3

100% 100% 100% 100% N= 2069

(Percentage difference between Upper and Lower for the “Excellent” row = 18.5) Partial Table #2

Crosstab: Health / SocClass

Control: Race (Black)

(Table: Percent Down)

Lower

Lower Middle

Upper Middle

Upper

Total

Excellent

18.0

20.2

25.0

23.5

21.0 Good

44.3

40.4

55.9

60.3

48.5 Poor–Fair

37.7

39.4

19.1

16.2

30.5

100% 100% 100% 100% 100% N= 357

(Percentage difference between Upper and Lower for the “Excellent” row = 5.5) Partial Table #3

Crosstab: Health / SocClass

Control: Race (Other)

(Table: Percent Down)

Lower

Lower Middle

Upper Middle

Upper

Total

Excellent

14.6

19.4

31.7

33.3

22.9 Good

46.6

56.9

50.0

47.4

50.0 Poor–Fair

38.8

23.6

18.3

19.3

27.1

100% 100% 100% 100% 100% N= 292

(Percentage difference between Upper and Lower for the “Excellent” row = 8.3) (Note that we needed to use only three for the four partial-table shells provided.) VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious

(i.e., causal)? Is a multivariable model evident?

Inspection of the partial tables reveals that the original positive relationship between SocClass and Health has maintained itself; we may therefore conclude that this relationship is nonspurious or causal.

(Note that the relationship has maintained itself almost perfectly for whites—increasing slightly from a 17.8 to an 18.5 percentage-difference; but that it has decreased substantially—though it is still there—for blacks

Chapter 1: The Problem of Social Order

73

and for those designated as representing an “other” race—to 5.5 and 8.3 respectively. However, if we look at the other extreme value of Health, the Poor–Fair row, we see that the relationship holds steadfast for all categories of the race variable.) A special note on multivariable effects: The percentage-differences for the partial tables above support the notion of a multivariable model: Whites, regardless of the value of SocClass, are more likely to be in the Excellent row for Health (e.g., compared to their black counterparts, whites in the Lower class have 2.1% greater chance of reporting Excellent Health; in the Lower Middle column, whites have a 4.9% advantage; in the Upper Middle, a 7.8% advantage, and in the Upper column, a 15.1% advantage. We thus need to continue our conclusion as follows:

Moreover, the tables reveal the independent-variable effects of Race on Health—that is, regardless of the value of SocClass, whites are more likely to be in the “Excellent” row of Health. For example, for those in the Lower social class, whites have a (20.1%–18.0%=) 2.1% greater chance of having “Excellent” health; similarly, for those in the Lower Middle class, whites have a 4.9% greater chance of having Excellent health, for the Upper Middle class, whites have 7.8% advantage; and for those in the Upper class, whites have a 15.1% advantage. We may sketch the multivariable model evident in these tables as follows:

As we have already anticipated, the effects of Z (in this case “Race”) are interpretable (whites suffer lower chances of getting diabetes—as well as the many diseases that stem from it—and thus should be more likely to have excellent health).

Doing Sociology with Student CHIP

74

75

Chapter 2. Issues in Sociological Research

4. Attitudes vs. Actions: Do Religiosity and Church Attendance Go Hand-in-Hand?

For some, the “great divide” in the sociological perspective is both a cause and a consequence of the two basic research orientations in the discipline. The orientations can be variously described: qualitative versus quantitative, subjective versus objective, observing and empathizing versus counting and measuring. Rich- ard T. LaPiere’s classic study of discrimination supports the qualitative-subjective-empathizing orientation.1 Quantitative studies on discrimination reveal an image of the world that is very different from the image qualitative research offers us. In the former, people fill out questionnaires—that is, answer survey ques- tions—to indicate what they do or think, while in the latter the sociologist actually observes their behaviors and has discussions with them to uncover their thoughts and motivations. Using survey research, LaPiere found that many owners of restaurants and motor inns said that they would discriminate against Asians— that is, they would not serve an Asian individual (the research was conducted in the 1930s!). However, some months earlier, when a Chinese couple working with LaPiere had attempted to eat or sleep at those establishments where the owners said they had a policy of discrimination, the couple had almost never been denied service. LaPiere’s research also highlights a very important caveat to keep in mind when conducting or evaluating sociological research: There is a gap—sometimes a very large one—between people’s attitudes and their behaviors, what they say they do or will do versus what they actually do.

The following computer exercises begin with an investigation of the gap between attitudes and behaviors. Students hearing about or studying LaPiere’s research can come away with the notion that this gap is always enormous. However, the CHIP exercises demonstrate that—although a gap exists—there is quite often still a connection between people’s attitudes and actions; and between subjective perception and objective fact. Indeed, it would be a crazy world to navigate if there weren’t such connections.

File: Relig2006 (Who’s religious? Source: 2004 & 2006 GSS—Christians only)

Info: RaceHispanicSexAgeSocClass Kids DenominationHowReligAttend (3) (2) (2) (3) (4) (2) (4) (3) (4)

Race White, Black, Other

Hispanic No, Yes

Sex Male, Female

Age 18–39, 40–64, 65+

SocClass (Respondent’s socioeconomic status, recoded as:2) Lower (17.1–33.0), Lower-Middle (33.1–44.2), Upper-Middle (44.3–64.1), Upper (64.2–97.2)

1 Richard T. LaPiere, “Attitudes vs. Actions,” Social Forces, 13 (December 1934), pp. 230–237; reprinted in Part 2 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010). 2 Socioeconomic index value based on respondent’s education, income, and occupational prestige; scores range from a minimum value of 17.1 to a maximum value of 97.2; see General Social Surveys, 1972–2006: Cumulative Codebook (Storrs, CT: Roper Center for Public Opinion Research, University of Connecticut, March 2007), p. 2235.

Doing Sociology with Student CHIP

76

Kids (Has respondent ever had a child?) No, Yes

Denomination Catholic, LibProt (Liberal Protestant), ModProt (Moderate Protestant), FundProt (Funda- mentalist Protestant)1

HowRelig (Would you call yourself a strong [respondent’s religion], or a not very strong [respondent’s religion]) Not Very, Somewhat, Strongly

Attend (How often do you attend religious services? [respondents were given a 9-point scale ranging from: (0) Never . . . to (8) Several Times a Week; these were recoded as follows:]) Never, Yearly, Monthly, Weekly

1. LaPiere’s research alerts us to the problem of the gap between an attitude or belief and what the expect- ed behavior associated with it would be. Assess this gap for religiosity (HowRelig) and church atten- dance by crosstabulating these two variables; also generate the associated plot (after doing the Cross- tab, select the Line Chart option under the Table command and highlight the “Weekly” category). Dis- cuss your findings, making sure that you assess the degree to which “behavior” (Attend) and attitude (HowRelig) correlate. Do you think LaPiere would be surprised—or not surprised—by what you found? (Give your answer in the Answer/Discussion following the plot of this relationship.)

1 Whether a Protestant denomination falls into a “fundamentalist,” “moderate,” or “liberal” category is based on the conservatism of its theology. With only a few key exceptions, conservative is tantamount to fundamentalist according to the GSS’s coding scheme. Fundamentalists are generally opposed to the growth of secular influence in society. They believe in the inerrancy of the Bible, personal salvation by accepting Christ as their savior (and being “born again”), the imminent return of Christ, evangelism/revivalism to reach out to save and convert others, the Holy Trinity, and the Virgin birth. Liberal denominations tend to emphasize concerns about this world more than salvation in the next, which leads them to support social action and progressive reform. They also tend to accept secular change and sci- ence, and do not accept the literal message of the Bible (seeing, for example, Biblical miracles as metaphorical in na- ture and not as historical facts). Denominations that fall between these extremes have been deemed moderate.” Most denominations can readily be placed in one of these three categories; for example, the Assembly of God and the Church of Christ are clearly fundamentalist, while Congregationalists and Unitarians are clearly liberal. Other funda- mentalist denominations include all Baptists (except the Northern Baptists), Missouri Synod Lutherans, and Jeho- vah’s Witnesses, while other liberal denominations include the United Methodist Church and the United Pres- byterian Church in the USA. Two denominations, that pose difficulty are the Mormons (Church of Jesus Christ of Latterday Saints) and the Christian Scientists. Based on a variety of historical and sociological facts, the GSS even- tually placed both in the fundamentalist category. For the details of this decision and for an in-depth analysis of the GSS coding scheme, see Tom W. Smith, “Classifying Protestant Denominations,” GSS Methodological Report No. 43, July, 1987 (available from the National Opinion Research Center, University of Chicago; also downloadable from the Internet: http://www.adherents.com/largecom/prot_classify.html); a revised edition of this report appears in the Review of Religious Research 31 (March 1990), pp. 225–245.

Chapter 2: Issues in Sociological Research

77

Crosstab: Attend / HowRelig

(Table: Percent Down)

Not Very

Somewhat

Strongly

Total

Weekly

Monthly

Yearly

Never

100% 100% 100% N=

(Percentage difference between Strongly and Not Very on “Weekly” = ) Prediction:

Finding:

Doing Sociology with Student CHIP

78

Answer/Discussion 2. Perhaps the gap between attitude and action is partially determined by one’s social characteristics. More

specifically, here, perhaps the gap narrows with age with regard to religiosity and church attendance— that is, as people grow older, those who say they are religious are more likely to attend church.

(a) Why do you suppose this might be so—that is, that older individuals who say they are reli- gious are more likely to go to church than younger people who also say they are religious?

(b) Test your interpretation in (a) by crosstabulating Attend (Y) by HowRelig (X) and control for Age. Focus your discussion on those individuals who are Strongly religious and go to church at least once a week (row 1, column 3 of each partial table). Does this percentage vary as we have hypothesized? (Be sure to answer this question and show the relevant percentages under the Answer/ Discussion section that follows your tables.)

Chapter 2: Issues in Sociological Research

79

Partial Table #1 Crosstab: Attend / HowRelig

Control: Age (18–39)

(Table: Percent Down)

Not Very

Somewhat

Strongly

Total

Weekly

Monthly

Yearly

Never

100% 100% 100% N=

(Percentage difference between Strongly and Not Very on “Weekly” = )

Partial Table #2 Crosstab: Attend / HowRelig

Control: Age (40–64)

(Table: Percent Down)

Not Very

Somewhat

Strongly

Total

Weekly

Monthly

Yearly

Never

100% 100% 100% N=

(Percentage difference between Strongly and Not Very on “Weekly” = )

Doing Sociology with Student CHIP

80

Partial Table #3 Crosstab: Attend / HowRelig

Control: Age (65+)

(Table: Percent Down)

Not Very

Somewhat

Strongly

Total

Weekly

Monthly

Yearly

Never

100% 100% 100% N=

(Percentage difference between Strongly and Not Very on “Weekly” = ) Answer/Discussion

Chapter 2: Issues in Sociological Research

81

5. The Idea of Contextual Effects: The Social Context of Working Full-Time for Women

In an article published in Teaching Sociology,1 I tried to show how quantitative methods can reveal a core concept in sociology—that individuals with the same personal characteristics behave and think differently when exposed to different social situations. The article also demonstrated how quantitative sociologists try to incorporate the subjective orientation into their research. The computer exercises that follow demonstrate the logic and truth of “contextual effects,” illustrating how the relationship between working full-time and education for women depends upon the presence or absence of children, as well as upon marital situation.

File: Work2006 (Social correlates of working. Source: 2004 & 2006 GSS)

Info: RaceHispanicImmigrantSexAge EdMarried KidsWorking (3) (2) (2) (2) (3) (4) (4) (2) (3)

Race White, Black, Other

Hispanic No, Yes

Immigrant No, Yes

Sex Male, Female

Age 18–39, 40–64, 65+

Ed (Years of schooling) <12yrs, 12yrs, 13–15yrs, 16+yrs

Married Never Married, Div–Sep, Widowed, Married

Kids (Has respondent ever had a child?) No, Yes

Working Retired, Other (working part-time, housewife, student, unemployed, on leave), Fulltime 1. We might expect that college-educated women would be more likely to work full-time than their less

educated counterparts. We would expect this because (a) they have invested a good deal of time and effort in their educations, and though there are many reasons to become educated, a primary one for almost everyone is to find a good job—one whose rewards make an individual want to work full-time; and, (b) in modern society, more and more jobs require a good deal of education—and thus we would expect that women with a college education would fare better in the job market than their less educated counterparts. Thus, we would hypothesize that education and working full-time are positively related. However, in my Teaching Sociology article, I argue that individuals with the same personal characteris- tics (e.g., being “college-educated”) behave and think differently when exposed to different social situa- tions (e.g., having children versus not having them). Does the relationship between education and working full-time depend upon the context of having or not having children? Further, does the relationship vary by marital context? To answer these questions, do the following:

(a) Use Modify to Omit the “Retired” category of Working (here, as always when using this command, give CHIP some extra time to do the modification; your screen may freeze for a brief while).

(b) Use Modify to Omit the “Male” category of Sex.

(c) Use Modify to Combine the “Div–Sep,” “Widowed,” and “Never Married” categories of Married; label this new category “Not Married.”

1 “Teaching the Idea of Contextual Effects,” Teaching Sociology 19:4 (October 1991), pp. 526–531; reprinted in Part 2 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010).

Doing Sociology with Student CHIP

82

(d) Use Modify to Combine the “12yrs” and “13–15yrs”categories of Ed; label this new category “12–15yrs.”

(e) Use Modify to Omit the “40–64” and “65+” categories of Age.

(f) Verify that you have executed your Modify commands correctly by doing an Info: you should now have 1,153 observations, and (1) Sex should have just one category; (2) Age should have just one category; and (3) Working and Married should each have only two categories.

(g) Crosstab Working (Y) by Education (X). Is the relationship in the expected direction?

Crosstab: Working / Ed

(Table: Percent Down)

<12yrs

12–15yrs

16+yrs

Total

Full-time

Other

100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Full-time” row = ) Prediction: Finding:

Control for both Kids and Married, then address the questions that follow your partial tables.

Partial Table #1

Crosstab: Working / Ed

Controls: Married (Not Married); Kids (No)

(Table: Percent Down)

<12yrs

12–15yrs

16+yrs

Total

Full-time

Other

100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Full-time” row = )

Chapter 2: Issues in Sociological Research

83

Partial Table #2

Crosstab: Working / Ed

Controls: Married (Not Married); Kids (Yes)

(Table: Percent Down)

<12yrs

12–15yrs

16+yrs

Total

Full-time

Other

100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Full-time” row = )

Partial Table #3

Crosstab: Working / Ed

Controls: Married (Married); Kids (No)

(Table: Percent Down)

<12yrs

12–15yrs

16+yrs

Total

Full-time

Other

100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Full-time” row = )

Partial Table #4

Crosstab: Working / Ed

Controls: Married (Married); Kids (Yes)

(Table: Percent Down)

<12yrs

12–15yrs

16+yrs

Total

Full-time

Other

100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Full-time” row = )

Doing Sociology with Student CHIP

84

(h) For younger (<40 years old), college-educated (16+ years of schooling) women, who is most likely to be working “Full-time?” (Hint: compare across the four partial tables, looking only at college- educated individuals.)

(i) For younger, college-educated women, who is least likely to be working “Full-time?” (j) In sum, does the effect of their education on working have a tendency to depend upon their

marital and children situations? (j) Why do you think this is so—that is, why do you think the data turned out as they did? Focus

your remarks on your answer in (i).

Chapter 2: Issues in Sociological Research

85

Exploratory

I. Using either of the CHIP data files for this chapter (Relig2006 or Work2006), state a hypothesis relating an X and a Y variable that have not already been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy-

pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 4×4 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

86

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic-

tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Chapter 2: Issues in Sociological Research

87

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z—using the

appended table shells. (Note: you may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

89

Chapter 3. Culture 6. The Problem of Ethnocentrism

In his classic sociological study entitled Folkways,1 William Graham Sumner develops the idea of ethno- centrism—the tendency for every group to think that its ways of doing things are the best. He shows its many faces and the many ways in which it structures our thinking about ourselves and others.

It should be hoped that one outcome of going to college is a greater appreciation for the cultural diversity of human societies on this planet. If this is true, then we should expect college-educated individuals to be less likely to hold negative racial stereotypes.

We might also expect that having opportunities for interracial contact, e.g., at work or in school or in one’s neighborhood, should reduce the likelihood of developing or maintaining negative racial stereotypes. Indeed, this expectation is supported by what social scientists call contact theory. Social science research has consistently found that interracial contacts reduce prejudice, especially where there is equal status between or among the groups involved; where norms encourage equality; and where there are tasks that require cooperation among individuals to achieve the goal at hand. Classic situations in which all three of these conditions often flourish are in the military and in school.2

The following computer exercises allow you to test these expectations, as well as others involved with the concept of ethnocentrism.

File: Stereo2006 (Stereotyping of racial groups; Source: 2004 & 2006 GSS)

Info: RegionRace2Age SexEdImmigrantOtherRaceStereoWhtsStereoBlks (4) (2) (3) (2) (4) (2) (2) (3) (3) Region Northeast (CT, ME, MA, NH, NJ, NY, PA, RI, and VT) Midwest (IL, IN, IA, KS, MI, MN, MO, NE, ND, OH, SD, and WI) South (AL, AR, DE, FL, GA, KY, LS, MD, MS, NC, OK, SC, TN, TX, VA, and WV) West (AK, AZ, CA, CO, HI, ID, MT, NV, NM, OR, UT, WA, and WY)

Race2 White, Black

Sex Male, Female

Age 18–39, 40–64, 65+

Ed (Years of schooling) <12yrs, 12yrs, 13–15yrs, 16+yrs

Immigrant No, Yes

OtherRace (If you are white, are there any blacks living in your neighborhood? If you are black, are there any whites living in your neighborhood?) No, Yes

1 Folkways: A Study in the Sociological Importance of Usages, Manners, Customs, Mores, and Morals (New York: Ginn and Company, 1906); key portions of which are reprinted in Part 3 of Gregg Lee Carter (ed.), Empirical Approaches to Soci- ology, 5th edition (Boston: Allyn & Bacon, 2010).

2 See, for example, Linda R. Tropp and Thomas F. Pettigrew, “Reducing Ethnocentrism and Prejudice through Inter- group Contact: Summarizing Results from a Meta-Analytic Review,” in Part 3 of Carter, ibid.

Doing Sociology with Student CHIP

90

StereoWhts (Stereotype of the work habits of whites based on the item: “Do [whites] . . . tend to be hard working or do they tend to be lazy:” 7-point scale from [1] Work Hard . . . to . . .[7] Lazy. Recoded as:) Work-Hard (1–3), So-so (4), Lazy (5–7)

StereoBlks (Stereotype of the work habits of blacks based on the item: “Do [blacks] . . . tend to be hard working or do they tend to be lazy:” 7-point scale from [1] Work Hard . . . to . . .[7] Lazy. Recoded as:) Work-Hard (1–3), So-so (4), Lazy (5–7)

1. As Sumner notes, ethnocentrism has many faces and is universal. We would therefore expect whites to have a tendency to extol their own work ethic and to disparage that of blacks; and, likewise, we would expect the same tendency to exist among blacks—that is, that they should be more likely to extol their own work ethic and to belittle that of whites. Test these expectations by crosstabulating StereoBlks (Y) by Race2 (X); and StereoWhts (Y) by Race2 (X);. Were the expectations realized? (Focus your discussion on the “Lazy” row of StereoBlks and of StereoWhts throughout this chapter.)

Crosstab: StereoBlks / Race2

(Table: Percent Down)

White

Black

Total

Lazy

So-so

Work-Hard

100% 100% N=

(Percentage difference between Black and White for the “Lazy” row = ) Prediction:

Finding:

Chapter 3: Culture

91

Crosstab: StereoWhts / Race2

(Table: Percent Down)

White

Black

Total

Lazy

So-so

Work-Hard

100% 100% N=

(Percentage difference between Black and White for the “Lazy” row = ) Prediction:

Finding:

Answer/Discussion

Doing Sociology with Student CHIP

92

2. Even though Sumner observes that ethnocentrism has been with humankind since the beginning, he also points out that it has its roots in unfamiliarity with other cultures and peoples. It would therefore seem arguable that individuals who have graduated from college—in the present data set, they would be those with 16+ years of schooling—would be less likely to hold other groups and subcultures in disdain. To test this hypothesis regarding one face of ethnocentrism, having negative stereotypes of another racial group, do the following:

(a) Use the Modify command to Omit “Blacks” from the variable Race2; under Commands, do

an Info to confirm that this modification was successful—you should now have only 2,016 cases (just the white respondents). Here, as always when using this command, give CHIP some extra time to do the modification; your screen may freeze for a brief time.

(b) Then Crosstab StereoBlks (Y) by Ed (X). Also generate the associated plot (after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Lazy” category). Is the relationship in the expected direction?

Crosstab: StereoBlks / Ed

(All respondents white)

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Lazy

So-so

Work-Hard

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Lazy” row = ) Prediction:

Finding:

Chapter 3: Culture

93

(c) Re-Open the Stereo2006 file; under Commands, do an Info to confirm that you are working with the original data file (2,382 cases).

(d) Then use the Modify command to Omit “Whites” from the variable Race2. Do an Info to confirm that this modification was successful—you should now have only 366 cases (just the black respondents). Here, once again, give CHIP some extra time to do the modification; your screen may freeze for a short while.

(e) Then Crosstab StereoWhts (Y) by Ed (X). Also generate the associated plot (after doing the CrossTab, select the Line Chart option under the Table command and highlight the “Lazy” category). Is the relationship in the expected direction?

Crosstab: StereoWhts / Ed

(All respondents black)

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Lazy

So-so

Work-Hard

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Lazy” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

94

Answer/Discussion

3. Although social science research supports the general proposition that interracial contact reduces preju- dice, it also suggests that those with privilege—in the U.S. this would generally speaking be white Americans—are more likely to view interracial contact positively and thus more likely to experience a reduction in prejudice and stereotyping.3 Test out both the general and specific expectations with your Stereo2006 date set by doing the following:

(a) Use the Modify command to Omit “Blacks” from the variable Race2; under Commands, do an Info to confirm that this modification was successful—you should now have only 2,016 cases (just the white respondents). Here, as always when using this command, give CHIP some extra time to do the modification; your screen may freeze for a brief time.

(b) Then Crosstab StereoBlks (Y) by OtherRace (X). Is the relationship in the expected direction?

3 See, for example, Linda R. Tropp, “Perceived Discrimination and Interracial Contact: Predicting Interracial Close- ness among Black and White Americans,” Social Psychology Quarterly 70:1 (March 2007), pp. 70–81.

Chapter 3: Culture

95

Crosstab: StereoBlks / OtherRace

(All respondents white)

(Table: Percent Down)

No

Yes

Total

Lazy

So-so

Work-Hard

100% 100% N=

(Percentage difference between Yes and No for the “Lazy” row = ) Prediction:

Finding:

(c) Re-Open the Stereo2006 file; under Command, do an Info to confirm that you are working with the original data file (2,382 cases).

(d) Then use the Modify command to Omit “Whites” from the variable Race2. Do an Info to confirm that this modification was successful—you should now have only 366 cases (just the black respondents). Here, once again, give CHIP some extra time to do the modification; your screen may freeze for a short while.

(e) Then Crosstab StereoWhts (Y) by OtherRace (X). Is the relationship in the expected direction? (Recall that you have a different expectation for black respondents.)

Doing Sociology with Student CHIP

96

Crosstab: StereoWhts / OtherRace

(All respondents black)

(Table: Percent Down)

No

Yes

Total

Lazy

So-so

Work-Hard

100% 100% N=

(Percentage difference between Yes and No for the “Lazy” row = ) Prediction:

Finding:

Answer/Discussion

Chapter 3: Culture

97

4. Generally speaking, age is a powerful determinant of how an individual views the world.

(a) How do you think Age will relate to the tendency of a white person to hold unfavorable stereotypes of the work ethic of blacks? More specifically, do you think that older whites will be more, or less, likely to hold unfavorable stereotypes of blacks? Why do you think this is so?

(b) Check out your reasoning in (a) by crosstabulating StereoBlks (Y) by Age (X). Be sure to eliminate blacks from the data file for this exercise by re-Opening Stereo2006 and using the Modify command to Omit the “Black” category of Race2 (just as you did in 2a and 3a above). Also examine the relationship graphically: after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Lazy” category. Is the relationship in the expected direction?

Crosstab: StereoBlks / Age

(All respondents white)

(Table: Percent Down)

18–39

40–64

65+

Total

Lazy

So-so

Work-Hard

100% 100% 100% N=

(Percentage difference between 65+ and 18–39 for the “Lazy” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

98

Answer/Discussion

Chapter 3: Culture

99

Exploratory

I. Using the CHIP Stereo2006 data file, state a hypothesis relating an X and a Y variable that have not al- ready been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy-

pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 4×4 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

100

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic-

tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Chapter 3: Culture

101

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z—using the

appended table shells. (Note: you may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

103

Chapter 4. Society

The Occupational Structure of Post-Industrial Society

All societies have some elements in common, as well as many differences. For example, all societies develop institutions to solve common problems of survival: how to produce and distribute goods and services (solved via the economy), how to control sexual relations and procreation (the family), how to fix certain rules that everyone must follow and how to regulate human interactions (the polity or government), how to train individuals to fulfill work and other important social roles (education), and how to answer the metaphysical questions that beset the human mind (religion). In his classic 1970s study The Coming of Post-Industrial Socie- ty,1 Daniel Bell describes how technological change has been altering these core institutions in the indus- trialized democracies of the United States, western Europe, and Japan. One key alteration is the ascendancy of theoretic knowledge in all fields, and the consequent importance of education and professional/technical occupations.

The following computer exercises allow you to explore this alteration in U.S. society. What Bell described in Post-Industrial Society—the growth of professional and technical employment, and the decline of the blue- collar workforce—has continued unabated into the new millennium. For example, while about one in four workers in 1970 was an executive, professional, or technician, more than one in three had such an occupa- tion in 2009 (this broad category includes financial managers, accountants, lawyers, architects, engineers, computer programmers, scientists, teachers, doctors, and dental hygienists). In 1970, a little more than a third of all workers held “blue-collar” jobs (e.g., factory work, truck driving, and plumbing), but this figure declined to less than a fourth of the workforce by 2009. For the first time in U.S. history, more people are now performing executive, professional, or technical jobs than making or transporting goods.

7. Changes in the Occupational Structure of the United States (Post-Industrial Society)

File: Occ1974_2006 (Occupational structure. Source: 1974–1976, 1984–1986, 1994–1996, & 2004–2006 GSS)

Info: EraRace2SexAgeDadOccEdOccFamInc2 (3) (2) (2) (3) (5) (4) (5) (3)

Era 74–76, 84–86, 94–96, 04–06

Race2 White, Black

Sex Male, Female

Age 18–39, 40–64, 65+

DadOcc (Respondent’s father’s occupation) for the 1994–1996 and 2004–2006 GSSs, occupational codes were recoded as follows: Farm (473–499), BC (Blue Collar; 503–889), Serv-Cler (Service–Clerical; 303–469), Tech-Sales (Technical–Sales; 203–285), Profession (Professional and Managerial; 3–199); for the 1970s and 1980s GSSs, occupational codes were recoded: Farm (801–846), BC (Blue Collar; 401–796), Serv-Cler (Service–Clerical; 301–396, 901–986), Tech-Sales (Technical–Sales; 80–85, 150– 174, 260–296), Profession (Professional and Managerial; 1–76; 86–145; 175–246) 2

1 The Coming of Post-Industrial Society: A Venture in Social Forecasting (New York: Basic Books, 1973).

2 The 1988 General Social Survey included both the new (Occ80) and old (Occ) census codings for occupation. Using the codings for DadOcc and Occ in Occ1974_2006, the rank-order correlation coefficient between Occ80 and Occ in the 1988 GSS is .91; thus, there is some measurement error when trying to compare occupational categories between the new and old census codings, but not much.

Doing Sociology with Student CHIP

104

Ed (Years of schooling) <12yrs, 12yrs, 13–15yrs, 16+yrs

Occ (Respondent’s occupation) for the 1994–1996 and 2004–2006 GSSs, occupational codes were recoded as follows: Farm (473–499), BC (Blue Collar; 503–889), Serv-Cler (Service–Clerical 303– 469), Tech-Sales (Technical–Sales; 203–285), Profession (Professional and Managerial; 3–199); for the 1970s and 1980s GSSs, occupational codes were recoded: Farm (801–846), BC (Blue Collar; 401–796), Serv-Cler (Service–Clerical; 301–396, 901–986), Tech-Sales (Technical–Sales; 80–85, 150– 174, 260–296), Profession (Professional and Managerial; 1–76; 86–145; 175–246)

FamInc2 (Self-reported assessment of family income: “Compared with American families in general, would you say your income is far below average, below average, average, above average, far above average.” Recoded as:) Below, Average, Above

1. According to Daniel Bell, what would we expect to find if we crosstabulated Occ (Y) by Era (X)? More

specifically, what should we be find happening as time goes by to the proportion of:

(a) Professional and Managerial jobs?

(b) Blue-Collar jobs?

(c) Crosstab Occ (Y) by Era (X). Also examine the (a) and (b) relationships graphically: after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Profes- sion” category; then repeat this procedure for the “BC” category. Are the relationships in the expected directions? In short, were both of our expectations realized?

Chapter 4: Society

105

Original Table

Crosstab: Occ / Era

(Table: Percent Down)

74–76

84–86

94–96

04–06

Total

Profession

Tech-Sales

Serv-Cler

BC

Farm

100% 100% 100% 100% N=

(Percentage difference between 04–06 and 74–76 for the “Profession” row = )

(Percentage difference between 04–06 and 74–76 for the “BC” row = ) Answer/Discussion

Doing Sociology with Student CHIP

106

2. Bell makes no distinctions between the sexes—that is, he expects the occupational structure to change for both men and women. Test this expectation by crosstabulating and Occ (Y) and Era (X), then control for Sex. Was this expectation confirmed? Focus your answer on the Profession and BC categories of Occ.

Partial Table #1

Crosstab: Occ / Era

Control: Sex (Male)

(Table: Percent Down)

74–76

84–86

94–96

04–06

Total

Profession

Tech-Sales

Serv-Cler

BC

Farm

100% 100% 100% 100% N=

(Percentage difference between 04–06 and 74–76 for the “Profession” row = )

(Percentage difference between 04–06 and 74–76 for the “BC” row = )

Partial Table #2

Crosstab: Occ / Era

Control: Sex (Female)

(Table: Percent Down)

74–76

84–86

94–96

04–06

Total

Profession

Tech-Sales

Serv-Cler

BC

Farm

100% 100% 100% 100% N=

(Percentage difference between 04–06 and 74–76 for the “Profession” row = )

(Percentage difference between 04–06 and 74–76 for the “BC” row = )

Chapter 4: Society

107

Answer/Discussion 3. Has the job prestige gap between men and women narrowed over time? More specifically, are high

prestige jobs in U.S. society becoming increasingly filled by women? If so, why do you think this has happened? Begin your answers by comparing the “Profession” row in Partial Table 1 versus Partial Table 2 above (hint: compare men and women in 1974–1976, then do the same comparison for 2004–2006; what happened?).

4. Bell makes no distinctions among the races—that is, he expects the occupational structure to change for

all of them. Test this expectation by crosstabulating Occ (Y) and Era (X), then control for Race. Was this expectation confirmed? Focus your answer on the Profession, Tech-Sales, and BC categories of Occ.

Partial Table #1

Crosstab: Occ / Era

Control: Race (White)

(Table: Percent Down)

74–76

84–86

94–96

04–06

Total

Profession

Tech-Sales

Serv-Cler

BC

Farm

100% 100% 100% 100% N=

(Percentage difference between 04–06 and 74–76 for the “Profession” row = )

(Percentage difference between 04–06 and 74–76 for the “BC” row = )

Doing Sociology with Student CHIP

108

Partial Table #2

Crosstab: Occ / Era

Control: Race (Black)

(Table: Percent Down)

74–76

84–86

94–96

04–06

Total

Profession

Tech-Sales

Serv-Cler

BC

Farm

100% 100% 100% 100% N=

(Percentage difference between 04–06 and 74–76 for the “Profession” row = )

(Percentage difference between 04–06 and 74–76 for the “BC” row = )

Answer/Discussion

Chapter 4: Society

119

Exploratory

I. Using one of the CHIP data files for this chapter (Occ1974_2006 or World2006), state a hypothesis relating an X and a Y variable that have not already been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy-

pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 5×5 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

120

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic-

tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Chapter 4: Society

121

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z. (Note: you

may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

123

of a teacher).

Chapter 5. Socialization

9. Social Class and Parental Values

Sociologists emphasize social interaction processes as essential to socialization and the development of self- conception and personality. Melvin Kohn has established that social class influences the value orientations parents instill in their children.1 More specifically, he has shown how—independent of religion and national background—middle-class parents are more likely to raise children to be self-directed, whereas working-class parents are more likely to raise children to conform to external authority. Why does this particular rela-tionship between social class and values exist? The short answer is that economic success for working-class parents depends upon their following their bosses’ orders, while success for middle-class individuals more often depends upon their creativity in problem-solving (e.g., factory-floor work is usually much more de-limited than, say, the diagnostic work of a physician or the lesson-planning

The following computer exercises allow you to explore the impact of social class on value orientation. File: Values2006 (Social class differences in value socialization. Source: 2004 & 2006 GSS)

Info: ImmigrantHispanicRaceSexEdOccObeyThink (2) (2) (3) (2) (4) (5) (2) (2)

Immigrant No, Yes

Hispanic No, Yes

Race White, Black, Other

Sex Male, Female

Ed (Years of schooling) <12yrs, 12yrs, 13–15yrs, 16+yrs

Occ Farm (GSS codes 473–499), BC (Blue Collar; 503–889), Serv-Cler (Service–Clerical; 303–469), Tech-Sales (Technical–Sales; 203–285), Profession (Professional and Managerial; 3–199)

Obey (“If you had to choose, which thing on this list would you pick as the most important for a child to learn to prepare him or her for life?: a. to obey; b. to be well-liked or popular; c. to think for himself or herself; d. to work hard; e. to help others when they need help.”) Other (respondent chose b–e as first choice), Obey (respondent chose “a. to obey” as first choice)

Think (“If you had to choose, which thing on this list would you pick as the most important for a child to learn to prepare him or her for life?: a. to obey; b. to be well-liked or popular; c. to think for himself or herself; d. to work hard; e. to help others when they need help.”) Other (respondent chose a, b, d, or e as first choice), Think (respondent chose “c. to think for himself or herself” as first choice

1Class and Conformity: A Study in Values (Homewood, IL: Dorsey Press, 1969); key portions of which are reprinted in Part 5 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010).

Doing Sociology with Student CHIP

124

Basic

1. According to Melvin Kohn, what would we expect to find if we crosstabulated Obey (Y) by Occ (X)? Do the crosstabulation, as well as the associated plot (after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Obey” category). Is this expectation realized?

Crosstab: Obey / Occ

(Table: Percent Down)

Farm

BC

Serv-Cler

Tech-Sales

Profession

Total

Obey

Other

100% 100% 100% 100% 100% N=

(Percentage difference between Profession and BC for the “Obey” row = ) Prediction:

Finding:

Chapter 5: Socialization

125

Answer/Discussion 2. Repeat #(1), but let X=Ed.

Crosstab: Obey / Ed

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Obey

Other

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Obey” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

126

Answer/Discussion 3. Repeat #(1), but let Y=Think

Crosstab: Think / Occ

(Table: Percent Down)

Farm

BC

Serv-Cler

Tech-Sales

Profession

Total

Think

Other

100% 100% 100% 100% 100% N=

(Percentage difference between Profession and BC for the “Think” row = ) Prediction:

Finding:

Chapter 5: Socialization

127

Answer/Discussion 4. Repeat #(2), but let Y=Think.

Crosstab: Think / Ed

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Think

Other

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Think” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

128

Answer/Discussion 5. Does the overall pattern of your findings in #(1)—#(4) confirm Kohn’s research? Discuss.

Chapter 5: Socialization

129

6. Assume, for the moment, that ethnicity (Hispanic) is significantly associated with the Values2006 dependent variable Obey. Do you think those with a Hispanic background would be more—or less— likely to emphasize “obeying” as a value to be instilled into a child? Why?

7. Test your expectation in #(6) by crosstabbing Obey (Y) by Hispanic (X). Was your expectation con- firmed? If not, why do you think it was not?

Crosstab: Obey / Hispanic

(Table: Percent Down)

No

Yes

Total

Obey

Other

100% 100% N=

(Percentage difference between Yes and No on the “Obey” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

130

Answer/Discussion

Chapter 5: Socialization

131

Advanced

8. Perhaps the relationship between Hispanic and Obey can be explained by the intervening variable of occupation; more specifically, perhaps Hispanics are more likely to be blue-collar workers, who in turn are more likely to give “to obey” as their first choice on the GSS Obey item. Sketch the model that would show that the relationship between Hispanic and Obey can be accounted for by Occ.

9. If the model you have sketched in #(8) is valid, what should happen to the Obey/Hispanic relationship

after we control for Occ? (Hint: you may want to go back to the chapter A Primer on Elementary Data Analysis and review the section entitled “The Art of Reading Partial Tables”)

10. Test your expectation in #(9) by doing the following:

(a) Use Modify to Omit the Farm, Serv-Cler, and Tech-Sales categories of Occ; Verify that you have executed your Modify command correctly; do an Info; you should now see only 949 observations; also note that Occ now has only two categories.

(b) Crosstabulate Obey (Y) by Hispanic (X), then control for Occ (Z). Was your expectation in #(9) confirmed?

Original Table

Crosstab: Obey / Hispanic

(Table: Percent Down)

No

Yes

Total

Obey

Other

100% 100% N=

(Percentage difference between Yes and No on the “Obey” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

132

Partial Table #1

Control: Occ (BC)

Crosstab: Obey / Hispanic

(Table: Percent Down)

No

Yes

Total

Obey

Other

100% 100% N=

(Percentage difference between Yes and No on the “Obey” row = )

Partial Table #2

Control: Occ (Profession)

Crosstab: Obey / Hispanic

(Table: Percent Down)

No

Yes

Total

Obey

Other

100% 100% N=

(Percentage difference between Yes and No on the “Obey” row = ) Answer/Discussion

Chapter 5: Socialization

133

Exploratory

I. Using the CHIP data file for this chapter (Values2006), state a hypothesis relating an X and a Y variable that have not already been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy-

pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 5×5 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

134

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic-

tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Chapter 5: Socialization

135

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z. (Note: you

may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

137

Chapter 6. Groups

10. Physical Health, the Quality of Primary-Group Ties, and Social Class

Essential to understanding the sociological perspective is realizing that different group memberships can generate different perceptions, behaviors, and states of health. Sidney Cobb has demonstrated that individ- uals with close primary-group ties survive illness and personal crisis better than those without such ties.1

The following computer exercises allow you to explore the impact of social ties on physical well-being. You can compare the self-reported healthiness of individuals with varying amounts of social support. You can also explore the influences of social class and age on healthiness and use these variables as controls to test for the possibility that the social supporthealthiness relationship is spurious. File: Health95 (Quality of social ties and health/happiness; source: 1991–1994 GSS)

Info: RaceSexAgeMarried EdFamInc3SatFFHealthHappy (2) (2) (3) (4) (4) (3) (3) (3) (3)

(Note: The 1998 and 2000 General Social Surveys did not include items on SatFF. Thus, here we use the Health95 CHIP data file that appeared in the third edition of this workbook.)

Race White, Black

Sex Male, Female

Age 18–39, 40–64, 65+

Married Never Married, Div/Sep, Widowed, Married

Ed (Years of schooling) <12yrs, 12yrs, 13–15yrs, 16+yrs

FamInc3 (Family Income) <$20K, $20K–$50K, $50K+

SatFF (Satisfaction with family life and friends. Sum of A and B, where: A= “How much satisfaction do you get from your family life”: 6=very great, 5=great, 4=quite a lot, 3=fair, 2=some, 1=little, 0=none; B= “How much satisfaction do you get from your friendships”: 6=very great, 5=great, 4=quite a lot, 3=fair, 2=some, 1=little, 0=none) So–So (0–9), Good (10–11), Excellent (12)

Health (“Would you say your own health, in general, is . . .”) Poor/Fair, Good, Excellent

Happy (“Taken all together, how would you say things are these days; would you say that you are very happy, pretty happy, not too happy?”) Not Too Happy, Pretty Happy, Very Happy

Basic

1. Sidney Cobb argues that “social support”—being involved in a network of communication and mutual obligation, as well as being cared for, loved, esteemed, and valued—diminishes the negative impacts of life’s stresses and transitions. All things equal, it would seem that individuals involved in satisfying relationships with their family and friends should be, on average, healthier. Test this hypothesis by crosstabulating Health (Y) by SatFF (X); also do the associated plot (after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Excellent” category). For this question and those that follow, focus your analysis on the “Excellent” category of Health. Was the hypothesis confirmed?

1 “Social Support as a Moderator of Life Stress,” Psychosomatic Medicine 38:5 (Sept.–Oct. 1976), pp. 300–314; reprinted in Part 6 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010).

Doing Sociology with Student CHIP

138

Crosstab: Health / SatFF

(Table: Percent Down)

So-So

Good

Excellent

Total

Excellent

Good

Poor/Fair

100% 100% 100% N=

(Percentage difference between Excellent and So-So for the “Excellent” row = ) Prediction:

Finding:

Answer/Discussion

Chapter 6: Groups

139

2. Social class—as indicated by education, income, and occupational prestige—is a predictor of infant birth weight. Low birth weights are correlated with a number of ill-health effects.

(a) In general, why would social class affect the health of individuals?

(b) Test the hypothesis that social class and healthiness are positively related by crosstabulating Health (Y) by FamInc3 (X). Also, examine the relationship graphically—after doing the Cross- tab, select the Line Chart option under the Table command and highlight the “Excellent” cate- gory. What did you find?

Crosstab: Health / FamInc3

(Table: Percent Down)

<$20K

$20K–$50K

$50K+

Total

Excellent

Good

Poor/Fair

100% 100% 100% N=

(Percentage difference between $50K+ and <$20K for the “Excellent” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

140

Answer/Discussion 3. As people grow older, their health problems mount. However, Cobb argues that older individuals with

social support are “protected from the consequences of the stress of growing old and infirm.” Test this hypothesis by:

(a) Crosstabbing Health (Y) and Age (X). Also examine the relationship graphically—after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Excel- lent” category.

Chapter 6: Groups

141

Crosstab: Health / Age

(Table: Percent Down)

18–39

40–64

65+

Total

Excellent

Good

Poor/Fair

100% 100% 100% N=

(Percentage difference between 65+ and 18–39 for the “Excellent” row = ) Prediction:

Finding:

(b) Then control for SatFF.

Doing Sociology with Student CHIP

142

Crosstab: Health / Age

Control: SatFF (So-So)

(Table: Percent Down)

18–39

40–64

65+

Total

Excellent

Good

Poor/Fair

100% 100% 100% N=

Crosstab: Health / Age

Control: SatFF (Good)

(Table: Percent Down)

18–39

40–64

65+

Total

Excellent

Good

Poor/Fair

100% 100% 100% N=

Crosstab: Health / Age

Control: SatFF (Excellent)

(Table: Percent Down)

18–39

40–64

65+

Total

Excellent

Good

Poor/Fair

100% 100% 100% N=

Chapter 6: Groups

143

(c) Does the crosstabulation in (a) confirm the notion that as people age their health problems increase?

(d) However, do the partial tables in (b) confirm Cobb’s contention that those older individuals with high social support (SatFF=“Excellent”) tend to be healthier? Defend your answer by referring to the correct percentages that can support it. (Hint: You need to compare three per- centages to answer this question—one from each partial table.)

Doing Sociology with Student CHIP

144

Advanced

4. Sketch the model that would show the relationship between SatFF and Health to be spurious, using FamInc3 as your antecedent variable. Defend this alternative model.

(a) Model sketch:

(b) Interpretation of the FamInc3–SatFF relationship:

(c) Interpretation of the FamInc3–Health relationship:

Chapter 6: Groups

145

5. Before testing the alternative model proposed in (4), do the following modifications to your variables (using your Modify command):

(a) Combine the two lowest FamInc3 categories and label them “<$50K”;

(b) Combine the two lowest categories of SatFF (So-So & Good) and label the new value “Not Excellent”;

(c) Combine the two lowest categories of Health (Poor/Fair and Good) and call this new cate- gory “Less-Than Excellent”; then verify that you have reduced the number of categories for all three variables by doing an Info under Command—you should see just two values now for FamInc3, SatFF, and Health.

(d) Now, crosstab Health (Y) by SatFF (X), then control for FamInc3 (Z). Does the relationship between SatFF and Health maintain itself when controlling for Income?

Original Table

Crosstab: Health / SatFF

(Table: Percent Down)

Not Excellent

Excellent

Total

Excellent

Less-Than Excellent

100% 100% N=

(Percentage difference between “Excellent” SatFF and “Not Excellent” SatFF for the “Excellent” row of Health = )

Prediction:

Finding:

Doing Sociology with Student CHIP

146

Partial Table #1

Crosstab: Health / SatFF

Control: FamInc3 (<$50K)

(Table: Percent Down)

Not Excellent

Excellent

Total

Excellent

Less Than Excellent

100% 100% N=

(Percentage difference between “Excellent” SatFF and “Not Excellent” SatFF for the “Excellent” row of Health = )

Partial Table #2

Crosstab: Health / SatFF

Control: FamInc3 ($50K+)

(Table: Percent Down)

Not Excellent

Excellent

Total

Excellent

Less Than Excellent

100% 100% N=

(Percentage difference between “Excellent” SatFF and “Not Excellent” SatFF for the “Excellent” row of Health = )

Answer/Discussion

6. Is there a multivariable model evident when using SatFF and FamInc3 to predict Health? That is, do both X and Z (SatFF and FamInc3) have independent effects? Refer to the correct percentage-differences to support your answer.

Chapter 6: Groups

147

11. Psychological Health, the Quality of Primary-Group Ties, and Social Class

The point previously made bears repeating: Essential to understanding the sociological perspective is to realize that different group memberships can generate different perceptions, behaviors, and states of health. Similar to Sidney Cobb’s demonstrating that individuals with close primary-group ties survive illness better than those without such ties, John Mirowsky and Catherine E. Ross show how psychological distress de- pends upon the groups to which individuals belong. For example, they demonstrate that married individ- uals are less likely to suffer distress compared to their unmarried counterparts (the divorced, widowed, or those never married) because the married are more likely to feel loved, cared for, esteemed, and valued—all of which offer benefit in handling the ups and downs of everyday living.1

The following computer exercises allow you to explore the impact of social ties on psychological well-being. You can compare the self-reported happiness of individuals with varying amounts of social support. You can also explore the influences of social class, marital status, and other social factors on happiness.

Basic

1. Mirowsky and Ross argue the “social support” that comes with social ties diminishes the negative im- pacts of life’s stresses. All things equal, it would seem that individuals involved in satisfying relation- ships with their family and friends should be, on average, happier. Test this hypothesis by crosstabulat- ing Happy (Y) by SatFF (X), as found in your Health95 CHIP data file (you will need to re-Open it if you have just finished the exercises in the preceding section). Also examine the relationship graphically— after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Very Happy” category. For this question and those that follow, focus your analysis on the “Very Happy” category of Happy. Was the hypothesis confirmed?

Crosstab: Happy / SatFF

(Table: Percent Down)

So-So

Good

Excellent

Total

Very Happy

Pretty Happy

Not Too

100% 100% 100% N=

(Percentage difference between Excellent and So-So for the “Very Happy” row = ) Prediction:

Finding:

1 Social Causes of Psychological Distress (New York: Aldine De Gruyter, 2003), pp. 75–89; key portions of which are reprinted in Part 6 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology, 5th edition (Boston: Allyn & Bacon, 2010).

Doing Sociology with Student CHIP

148

Answer/Discussion 2. “Money can’t buy happiness” is an old saying. On the other hand, the Beatles offer the alternative hy-

pothesis that money does buy happiness in their song “Money” (they concede that money can’t buy everything, but then note that the things it can’t buy are things that they don’t use very much!).

(a) Take the Beatles’ point of view: Why would social class (as measured by family income) af- fect the likelihood of being very happy?

Chapter 6: Groups

149

(b) Test the hypothesis that social class and happiness are positively related by crosstabulating Happy (Y) by FamInc3 (X). Also examine the relationship graphically—after doing the Cross- tab, select the Line Chart option under the Table command and highlight the “Very Happy” category. What did you find?

Crosstab: Happy / FamInc3

(Table: Percent Down)

<$20K

$20K–$50K

$50K+

Total

Very Happy

Pretty Happy

Not Too

100% 100% 100% N=

(Percentage difference between $50K+ and <$20K for the “Very Happy” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

150

Answer/Discussion

Chapter 6: Groups

151

3. As already noted, Mirowsky and Ross stress that married people are less likely to suffer distress com- pared to their unmarried counterparts (the divorced, widowed, or those never married). They argue that married individuals are more likely to feel loved, respected, and valued—all of which offer benefit in handling the stresses of everyday living. Test this hypothesis by crosstabulating Happy (Y) by Married (X). Also examine the relationship graphically—after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Very Happy” category. Was their hypothesis con- firmed?

Crosstab: Happy / Married

(Table: Percent Down)

Never Married

Div/Sep

Widowed

Married

Total

Very Happy

Pretty Happy

Not Too

100% 100% 100% 100% N=

(Percentage difference between Married and Div/Sep for the “Very Happy row” = ) Prediction:

Finding:

Doing Sociology with Student CHIP

152

Answer/Discussion

Chapter 6: Groups

171

Exploratory

I. Using one of the CHIP data files for this chapter (Health95, Seniors, or Divorce2006), state a hypothesis relating an X and a Y variable that have not already been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hy-

pothesis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 4×4 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

172

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is predic-

tive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Chapter 6: Groups

173

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z—using the

appended table shells. (Note: you may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

175

Chapter 7. Interaction 14. The Relationship of Internet Use to Social Background and Emotional Well- Being

One of the first systematic studies of the social and psychological effects of the Internet—conducted by researchers at Carnegie Mellon University in the late 1990s—showed the power of social context in determining the nature of interaction and its consequences. More specifically, the researchers found that the context of sitting in front of a computer screen instead of in the actual presence of someone (or communicating via the telephone) constrains the quality of the interaction, thereby increasing the odds of feeling lonely and depressed. For example, the researchers observed that “on-line friends are less likely than friends developed at school, work, church, or in the neighborhood to be available for help with tangible favors, such as offering small loans, rides, or babysitting. Because on-line friends are not embedded in the same day-to-day environment, they will be less like to understand the context for conversation, making discussion more difficult and rending support less applicable.”1 However, in a subsequent study, researchers at Michigan State University found only partial support for the Carnegie Mellon study: they did indeed find a causal link between Internet use and depression, but discovered that the relationship was mediated and modified by several considerations. For example, they found that Internet use can lead to decreased depression through the use of e-mail exchanges when the associates are known and are used to obtain social support.2 This finding is especially noteworthy in the context of the growing use of the Internet in the United States and elsewhere. The majority of U.S. adults use the Internet regularly—spending many hours a week online involved in: (1) the use of e-mail, (2) instant messaging, (3) Web browsing, (4) shopping, (5) finding out about entertainment, (6) reading news, and (7) participating in social networking sites such as Facebook and MySpace. Importantly, according to the latest research, they are most satisfied with their ability to communicate with friends, relatives, and others.3

As the Internet and other forms of communication technology—á la PDAs and related cell-phone tech- nologies—grow in popularity, their effects on social relationships and individual well-being will run deep, and it will become a central focus of sociological research for many years to come. Many of the prognostications have become increasingly hopeful. For example, sociologists Barry Wellman and Milena Gulia proffer that the Internet is beginning “. . . to integrate society and foster social trust. The architecture of the Internet facilitates weak and strong ties that cut across social milieus—be they interest groups, localities, organizations, or nations—so that the cyberlinks between people become social links between groups that otherwise would be socially and physically dispersed.”4 And, indeed, a recent major survey on Internet use by Jeffrey Boase—and his colleagues at the University of Toronto and the

1 Robert Kraut, et al., “Internet Paradox: A Social Technology That Reduces Social Involvement and Psychological Well- Being?” American Psychologist 53(1998):1017–1031. 2 Robert LaRose, et al., “Reformulating the Internet Paradox: Social-Cognitive Explanations of Internet Use and Depression.” Journal of Online Behavior (2001): http://www.behavior.net/JOB/v1n2/paradox.html. Accessed August 27, 2008. Reprinted in Part 7 of Gregg Lee Carter (ed.), Empirical Approaches to Sociology 4th edition (Boston: Allyn & Bacon, 2004). 3 See Jeffrey L. Cole, et al., The UCLA Internet Report 2001: Survey of the Digital Future, Year Two (Los Angeles: UCLA Center for Communication Policy, 2002): http://teachopolis.org/dlu/THE_UCLA_INTERNET_REPORT.pdf. The annual Survey of the Digital Future is now conducted at the University of Southern California; for recent survey findings, see http://www.digitalcenter.org/. Both of these URLs accessed August 27, 2008. 4 Barry Wellman and Milena Gulia, “Virtual Communities as Communities: Net Surfers Don’t Ride Alone.” Pp. 167– 194 in Marc A. Smith and Peter Kollock (eds.), Communities in Cyberspace (New York: Routledge, 1999).

Doing Sociology with Student CHIP

176

Pew Research Center in Washington, D.C.—confirms Wellman and Gulia’s observation. On balance, Boase et al. find that the Internet and related communication technologies are “good” for society— increasing social capital and contributing to the well-being and personal capital of the individuals having access to these technologies.1

Indeed, the most recent sociological research emphasizes the benefit of being on the Web. For example, MIT social scientist and director of the university’s laboratory on aging, Joseph Coughlin, observes that “the new future of old age is about staying in society, staying in the workplace and staying very con- nected . . . And technology is going to be a very big part of that, because the new reality is increasingly a virtual reality. It provides a way to make new connections, new friends and new senses of purpose.”2 Echoing this view is the conclusion of sociologist Antonina Bambina: She observes that on-line networks offer older people “a place where they do feel empowered, because they can make these connections and they can talk to people without having to ask a friend or a family member for one more thing.”3

The following CHIP exercises will allow you to investigate how Internet usage is related to emotional well-being, as indicated by feelings of trust and happiness, as well as to see how this usage is related to social background characteristics such as race, age, and social class. If Boase et al., Coughlin, and Bambina are right, then those left out of the Web will become increasingly marginalized in the general society—and bear the emotional and physical costs such marginalization entails.

File: NetUse2006 (Internet use, social background, and emotional well-being. Source: 2004 & 2006 GSS)

Info: RaceSexAgeSocClassNetUseTrustHappy (3) (2) (3) (4) (3) (2) (3)

Race White, Black, Other

Sex Male, Female

Age 18–39, 40–64, 65+

SocClass (Respondent’s socioeconomic status, recoded as:4) Lower (17.1–33.0), Lower-Middle (33.1–44.2), Upper-Middle (44.3–64.1), Upper (64.2–97.2)

NetUse (“Not counting e-mail, about how many hours per week do you use the Web? Include time you spend visiting regular web sites and time spent using interactive Internet services like chat rooms, Usernet groups, discussion forums, bulletin boards, and the like.”) 0 hrs, 1– 6hrs, 7+hrs

Trust (“Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?) Be Careful5, Can Trust

1 The Strength of Internet Ties (January 26, 2006), see, especially, pp. i–ix, 47–49 (Pew Internet & American Life Project: http://www.pewinternet.org/index.asp). Reprinted in Part 7 of Gregg Lee Carter (ed.), Empirical Approaches to Soci- ology 5th edition (Boston: Allyn & Bacon, 2010). 2 As quoted in Stephanie Clifford, “Online, ‘a Reason to Keep on Going’,” New York Times (June 1, 2009). 3 Quoted in Clifford, Ibid. Bambina is also the author of Online Social Support (Youngstown, NY: Cambria, 2007). 4 Socioeconomic index value based on respondent’s education, income, and occupational prestige; scores range from a minimum value of 17.1 to a maximum value of 97.2; see General Social Surveys, 1972–2006: Cumulative Codebook (Storrs, CT: Roper Center for Public Opinion Research, University of Connecticut, March 2007), p. 2235. 5 A small number of GSS respondents chose the response “Depends” for this question; in NetUse2006.chp “Depends” and “Be Careful” have been combined.

Chapter 7: Interaction

177

Happy (“Taken all together, how would you say things are these days; would you say that you are very happy, pretty happy, not too happy?”) Not Too Happy, Pretty Happy, Very Happy

Basic

1. (a) As observed above, high-volume Internet use might, theoretically speaking, constrain the quality of social interaction, thereby increasing the odds of feeling lonely and depressed. If this is true, then what should we expect to find if we crosstabulated Happy (Y) by NetUse (X)? In particular, would we expect that those who spend hours each day in chat rooms, discus- sion forums, social networking sites, and the like to have the smallest probability of being “very happy?”

(b) Do the crosstabulation, as well as the associated plot (after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Very Happy” category). Was our expectation realized? Focus your answer on the Very Happy row.

Crosstab: Happy / NetUse

(Table: Percent Down)

0 hrs

1–6hrs

7+hrs

Total

Very Happy

Pretty Happy

Not Too Happy

100% 100% 100% N=

(Percentage difference between 7+hrs and 0 hrs for the “Very Happy” row = ) Prediction: Finding:

Doing Sociology with Student CHIP

178

Answer/Discussion 2. (a) As noted above, sociologists Barry Wellman and Milena Gulia proffer that the Internet is

beginning to integrate society and foster social trust. The Internet facilitates social links that cut across milieus—like interest groups, localities, organizations, or nations—so that the cyberlinks between people become social links between groups that otherwise would be socially and physically spread apart. Moreover, Jeffrey Boase and his colleagues found that the Internet and related communication technologies are “good” for society—increasing social capital and contributing to the well-being of the individuals having access to these technologies. Recall from the “Introduction to Part 1” that social capital is built on shared values—which are, in turn, the prerequisite of all forms of group enterprise (from running a grocery store to lobbying Congress to raising children). Group enterprises are the foundation for communities, which both reflect and give orderliness to a society. Social capital is also built on trust. If Wellman, Gulia, Boase, et al. are on target, what should we expect to find if we crosstabulated Trust (Y) by NetUse (X)? In particular, would we expect that those who spend hours each day in chat rooms, discussion forums, social networking sites, and the like to have the highest probability of believing that we “can trust” most people?

Chapter 7: Interaction

179

(b) Do the crosstabulation of Trust (Y) by NetUse (X), as well as the associated plot (after doing the Crosstab, select the Line Chart option under the Table command and highlight the “Can Trust” category). Was our expectation realized? Focus your answer on the Can Trust row.

Crosstab: Trust / NetUse

(Table: Percent Down)

0 hrs

1–6hrs

7+hrs

Total

Can Trust

Be Careful

100% 100% 100% N=

(Percentage difference between 7+hrs and 0 hrs for the “Can Trust” row = ) Prediction: Finding:

Answer/Discussion

Doing Sociology with Student CHIP

180

3. As we have thus far discovered, Internet use does not appear to correlate with emotional well- being measured as happiness, but does seem to have a positive effect on social trust. If this relationship is valid, then those not on the Web—for whatever reason (from poverty to lack of desire to ignorance)—are missing out on an important resource, a resource that can build both personal and social capital. For each of the following social background variables, state a hypothesis on how you think it will be related to NetUse (Y):

(a) Race: (b) Age: (c) SocClass:

Chapter 7: Interaction

181

4. Test your predictions by crosstabulating NetUse by Race , Sex, Age, and SocClass. Were your expecta- tions realized? Be sure to restate your hypothesis, give your basic finding, and give an interpretation for each in your Answer/Discussion sections. Direct your attention especially to the 0 hrs row of each cross- tabulation (thus focusing on those individuals lacking the benefits of the Web in their lives).

(a) NetUse (Y) by Race (X):

Crosstab: NetUse / Race

(Table: Percent Down)

White

Black

Other

Total

7+ hrs

1–6 hrs

0 hrs

100% 100% 100% N=

(Percentage difference between Black and White for the “0 hrs” row = ) Prediction:

Finding:

Answer/Discussion

Doing Sociology with Student CHIP

182

(b) NetUse (Y) by Age (X):

Crosstab: NetUse / Age

(Table: Percent Down)

18–39

40–64

65+

Total

7+ hrs

1–6 hrs

0 hrs

100% 100% 100% N=

(Percentage difference between 65+ and 18–39 for the “0 hrs” row = ) Prediction:

Finding:

Answer/Discussion

Chapter 7: Interaction

183

(c) NetUse (Y) by SocClass (X):

Crosstab: NetUse / SocClass

((Table: Percent Down)

Lower

Lower-Middle

Upper-Middle

Upper

Total

7+ hrs

1–6 hrs

0 hrs

100% 100% 100% 100% N=

(Percentage difference between Upper and Lower for the “0 hrs” row = ) Prediction:

Finding:

Answer/Discussion

Doing Sociology with Student CHIP

184

Advanced

4. Sketch the model that would show the relationship between NetUse and Trust to be spurious, using SocClass as your antecedent variable. Defend this alternative model.

(a) Model sketch:

(b) Interpretation of the SocClass–NetUse relationship:

(c) Interpretation of the SocClass–Trust relationship: 2. Let’s test the model you have sketched and defended above. To simplify your analyses, do the follow-

ing modifications to your NetUse2006 data file: (1) Click on Modify, then Omit, and omit the Lower- Middle and Upper-Middle categories of the variable SocClass. Verify that you have done this modifica- tion correctly by clicking on Info; you should have “2” values for SocClass, and your number of cases (N=) should now equal 1,331 (note that you had 4 values for your original SocClass variable and 2,716 cases before you did this modification). Now, crosstab Trust (Y) by NetUse (X), controlling for SocClass (Z). Does the relationship between Internet use and trust maintain itself when controlling for social class? Is an interaction model evident? (Hint: you may want to go back to the chapter A Primer on Elementary Data Analysis and review the section entitled “The Art of Reading Partial Tables.”) Refer to the correct percentage-differences to support your answer.

Chapter 7: Interaction

185

Crosstab: Trust / NetUse

(Table: Percent Down)

0 hrs

1–6hrs

7+hrs

Total

Can Trust

Be Careful

100% 100% 100% N=

(Percentage difference between 7+hrs and 0 hrs for the “Can Trust” row = ) Prediction:

Finding:

Partial Table #1

Crosstab: Trust / NetUse

Control: SocClass (Lower)

(Table: Percent Down)

0 hrs

1–6hrs

7+hrs

Total

Can Trust

Be Careful

100% 100% 100% N=

(Percentage difference between 7+hrs and 0 hrs for the “Can Trust” row = )

Partial Table #2

Crosstab: Trust / NetUse

Control: SocClass (Upper)

(Table: Percent Down)

0 hrs

1–6hrs

7+hrs

Total

Can Trust

Be Careful

100% 100% 100% N=

(Percentage difference between 7+hrs and 0 hrs for the “Can Trust” row = )

Doing Sociology with Student CHIP

186

Summary of findings for the partial tables:

Chapter 7: Interaction

187

Exploratory

I. Using the NetUse2006 CHIP data file for this chapter, state a hypothesis relating an X and a Y variable that have not already been analyzed together.

II. Sketch the bivariate model.

III. Give a brief interpretation of your hypothesis—that is, describe what is going on in the world such that

we would expect to find data patterned in the way in which you have predicted. IV. (a) Test your hypothesis with a Crosstab, putting your Y variable on the rows. Was your hypoth-

esis confirmed? (Note: you may need to delete one or two rows and/or one or two columns; the following 4×4 table shell is simply a starting point.)

Original Table

Crosstab: ____________(Y) /____________ (X)

(Table: Percent Down)

Total

100% 100% 100% 100% N=

(Percentage difference between the highest and lowest values of X on the highest value of Y = ) Prediction:

Finding:

Doing Sociology with Student CHIP

188

(b) Use one of the plots under the Table option to display the above relationship graphically. Feel free to be creative—trying out each of the plot types (line, bar, pie, stacked). Print out and attach the plot that you think best captures the relationship between your X and Y.

Do either parts V–VIII or parts IX–XII below. V. Perhaps the relationship you uncovered in #IV is spurious; that is, perhaps a third variable is

predictive of both X and Y; if this is so, then the relationship between X and Y would exist not because X is causing Y, but simply because of their covariation with this third variable. If this third variable is held constant, then the relationship between X and Y will weaken greatly or disappear. Choose a third variable that might possibly be generating a spurious relationship between X and Y. Sketch the model showing the relationship between this third variable and X, and between this third variable and Y, as well as the lack of causal relationship between X and Y. Hint: refer back to the discussion on page 19 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

VI. A good social scientist does not choose just any variable to test for spuriosity. Just as you were able to

defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between Z and X, then between Z and Y.

(a) Interpretation of the Z–X relationship:

(b) Interpretation of the Z–Y relationship: VII. Test the alternative model sketched in #V by crosstabbing Y by X and controlling for Z—using the

appended table shells.

Chapter 7: Interaction

189

VIII. What are your conclusions? For example, is the original X–Y relationship spurious? Is it nonspurious (i.e., causal)? Is a multivariable model evident?

IX. Examining all the variables in your data set, which one do you think might be serving as a causal

mechanism connecting your X with your Y? In other words, which variable would you choose as “Z” in the following sketch: XZY? Hint: refer back to the discussion on page 21 in the introductory chapter entitled “A Primer on Elementary Data Analysis.”

X. A good social scientist does not choose just any variable to test as a causal mechanism (intervening

variable). Just as you were able to defend the hypothesized relationship between X and Y in #III, develop a brief interpretation to defend the hypothesized relationship between X and Z, then between Z and Y.

(a) Interpretation of the X–Z relationship:

(b) Interpretation of the Z–Y relationship: XI. Test the alternative model sketched in #IX by crosstabbing X and Y and controlling for Z—using the

appended table shells. (Note: you may need more than the four partial tables provided; of course, you will need one partial table for each value of Z.)

XII. What are your conclusions? Most importantly, do your findings support the notion that your Z is

acting as an intervening variable (causal mechanism) connecting your X and your Y?

207

Chapter 9. Inequality

17. An Examination of the “Status Attainment” Model and the Predictors of Individual Economic Success

The following computer exercises address the issue of who is most likely to succeed economically in U.S. society. The CHIP file Status2006 contains many of the key variables associated with status attainment research begun by Blau and Duncan1 and continued by Jencks2 and others at Harvard University and by the Duncans and Featherman3 and their colleagues at the University of Michigan and the University of Wisconsin. The models for status attainment developed by these investigators are among the most rigorous- ly tested in all of sociology; you can test the most significant aspects of them with the GSS data in Status2006. File: Status2006 (Status attainment model. Source: 2004 & 2006 GSS; Note: full-time workers only)

Info: RegionRaceHispanicSexAgeDadEdSibsEdIncome (4) (3) (2) (2) (3) (4) (3) (4) (3)

Region Northeast (CT, ME, MA, NH, NJ, NY, PA, RI, and VT) Midwest (IL, IN, IA, KS, MI, MN, MO, NE, ND, OH, SD, and WI) South (AL, AR, DE, FL, GA, KY, LS, MD, MS, NC, OK, SC, TN, TX, VA, and WV) West (AK, AZ, CA, CO, HI, ID, MT, NV, NM, OR, UT, WA, and WY)

Race White, Black, Other

Hispanic No, Yes

Sex Male, Female

Age 18–39, 40–64, 65+

DadEd (Years of schooling of respondent’s father) <12yrs, 12yrs, 13–15yrs, 16+yrs

Sibs (“How many brothers and sisters did you have? Please count those born alive, but no longer liv- ing, as well as those alive now. Also include step-brothers and step-sisters, and children adopted by your parents.”) 0–1, 2–3, 4+

Ed (Years of schooling): <12yrs, 12yrs, 13–15yrs, 16+yrs

Income (Respondent’s individual income): Bottom-3rd, Middle-3rd, Top-3rd

Basic

1. Who is most likely to succeed economically in American society? Develop a profile of this individual by crosstabulating Income (Y) by each of the following, focusing your attention on the Top-3rd Income row.

1 Peter M. Blau and Otis Dudley Duncan, The American Occupational Structure (New York: Free Press), 1967. 2 Christopher Jencks, et al., Inequality: A Reassessment of the Effect of Family and Schooling in America (New York: Basic Books), 1972. 3 Otis Dudley Duncan, David L. Featherman, and Beverly Duncan, Socioeconomic Background and Achievement (New York: Seminar Press), 1972. For recent a review of the status attainment literature and a strong confirmation of the basic model, see Samuel Bowles, Herbert Gintis, and Melissa Osborne Groves (eds.), Unequal Chances: Family Background and Economic Success (Princeton, NJ: Princeton University Press, 2005), as well as Ingrid Schoon, “A Trans- generational Model of Status Attainment: the Potential Mediating Role of School Motivation and Education,” Nation- al Institute Economic Review 205:1 (2008), pp, 72–82.

Doing Sociology with Student CHIP

208

(a) Region (b) Race (c) Hispanic (d) Sex (e) Age (f) Sibs (g) DadEd (h) Ed

Crosstab: Income / Region

(Table: Percent Down)

Northeast

Midwest

South

West

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between West and South for the “Top-3rd” row = (Percentage difference between Midwest and South for the “Top-3rd” row = (Percentage difference between Northeast and South for the “Top-3rd” row =

) ) )

Prediction:

Findings:

Chapter 9: Inequality

209

Crosstab: Income / Race

(Table: Percent Down)

White

Black

Other

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% N=

(Percentage difference between Black and White for the “Top-3rd” row = ) Prediction: Finding:

Crosstab: Income / Hispanic

(Table: Percent Down)

No

Yes

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% N=

(Percentage difference between Yes and No for the “Top-3rd” row = ) Prediction:

Finding:

Doing Sociology with Student CHIP

210

Crosstab: Income / Sex

(Table: Percent Down)

Male

Female

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% N=

(Percentage difference between Female and Male for the “Top-3rd” row = ) Prediction:

Finding:

Crosstab: Income / Age

(Table: Percent Down)

18–39

40–64

65+

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% N=

(Percentage difference between 40–64 and 65+ for the “Top-3rd” row = ) (Percentage difference between 40–64 and 18–39 for the “Top-3rd” row = )

Prediction:

Findings:

Chapter 9: Inequality

211

Crosstab: Income / Sibs

(Table: Percent Down)

0–1

2–3

4+

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% N=

(Percentage difference between 4+ and 0–1 for the “Top-3rd” row = ) Prediction:

Finding:

Crosstab: Income / DadEd

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Top-3rd” row =

)

Prediction:

Finding:

Doing Sociology with Student CHIP

212

Crosstab: Income / Ed

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Top-3rd” row = ) Prediction:

Finding:

Who is most likely to succeed? (e.g., a western, white, female . . . 8 characteristics must be noted): 2. Using “percentage-difference” as your criterion,

(a) which independent variable above best predicts Income? That is, given a change in the value of X, in which crosstab did you observe the greatest percentage change in the Top-3rd row?

(b) which independent variable is the second best predictor?

Chapter 9: Inequality

213

3. Based on your reading, class discussion, and/or your own intuition:

(a) why do you think Region is related to Income in the way in which it is?

(b) why do you think Race is related to Income in the way in which it is?

(c) why do you think Hispanic is related to Income in the way in which it is?

Doing Sociology with Student CHIP

214

(d) why do you think Sex is related to Income in the way in which it is?

(e) why do you think Age is related to Income in the way in which it is?

(f) why do you think Sibs is related to Income in the way in which it is?

Chapter 9: Inequality

215

(g) why do you think DadEd is related to Income in the way in which it is?

(h) why do you think Ed is related to Income in the way in which it is?

(i) Does the Ed  Income relationship make you glad that you’re in college? 

Doing Sociology with Student CHIP

216

Advanced

4. Perhaps the relationship between education and status attainment uncovered above is spurious. Sketch the model that would show this relationship as spurious when controlling for the antecedent variable Race. Defend this alternative model (for this exercise we will use only the White and Black categories for the Race variable).

(a) Model sketch:

(b) Interpretation of the Race–Ed relationship:

(c) Interpretation of the Race–Income relationship:

Chapter 9: Inequality

217

5. Check out your alternative model in #(4). First click on Modify to Omit the “Other” category of Race. Here, as always when using this command, give CHIP some extra time to do the modification; your screen may freeze for a brief while. Verify that you have done the modification to your data file correctly by clicking on Info: you should now have just 2 values for the variable Race (instead of your original 3 values), and the number of cases should be reduced to 1,910:

(a) Crosstabulate Income (Y) by Ed (X), then control for Race (Z).

Original Table

Crosstab: Income / Ed

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Top-3rd” row = ) Prediction:

Finding:

Partial Table #1

Crosstab: Income / Ed

Control: Race (White)

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Top-3rd” row = )

Doing Sociology with Student CHIP

218

Partial Table #2

Crosstab: Income / Ed

Control: Race (Black)

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Top-3rd” row = )

(b) Was the original relationship between education and income spurious?

(c) Who is most likely to be in the Top-3rd category? Does this surprise you? Why or why not?

(d) Who is most likely to be in the Bottom-3rd category? Does this surprise you? Why or why not?

Chapter 9: Inequality

219

6. In research on status attainment (that is, on who becomes most prosperous and who gets the best jobs), the relationship between the education of a respondent’s father and the respondent’s income—as un- covered in #(1g) above—is accounted for, in part, by the relationship between father’s education and respondent’s education. That is, fathers with high education tend to raise children who get a lot of schooling; in turn, those with a lot of schooling tend to make higher incomes. This causal pathway des- cribes the “indirect” effect of a father’s education on a respondent’s income attainment. The argument here is that children from higher social classes tend to end up making more money not because of who they are or whom they know, but because of the education that their parents are able to afford for them. Thus, the upper-middle-class child who does not take advantage of the educational opportunities afforded him or her—say by dropping out of school—will likely wind up in a low prestige, low-paying job. On the other hand, the child from humble roots who manages to get an education is not hindered by his or her humble beginnings (e.g., the medical school doesn’t ask who the parents of an applicant are—rather, the applicant is asked to provide his or her GPA and standardized test scores).

In addition to the indirect effects of father’s education on respondent’s income, there is also a direct effect. That is, beyond educational opportunities (or lack of opportunities for poorer children), parents also can bequeath their children income directly—through, for example, the inheritance of income- producing wealth (e.g., stocks, bonds, trust funds) or actual jobs (e.g., taking over a family business or being able to enter a field through the intervention of a parent). Sketch the model that would represent both the direct and indirect effects of DadEd on Income.

7. If the model sketched above is true, then what will happen to the relationship between DadEd and In-

come when Ed is held constant? Why?

Doing Sociology with Student CHIP

220

8. Test your prediction in #(7) by crosstabulating Income (Y) by DadEd (X), then control for Ed (Z). What did you find?

Crosstab: Income / DadEd

(Table: Percent Down)

<12yrs

12yrs

13–15yrs

16+yrs

Total

Top-3rd

Middle-3rd

Bottom-3rd

100% 100% 100% 100% N=

(Percentage difference between 16+yrs and <12yrs for the “Top-3rd” row =

Our website has a team of professional writers who can help you write any of your homework. They will write your papers from scratch. We also have a team of editors just to make sure all papers are of HIGH QUALITY & PLAGIARISM FREE. To make an Order you only need to click Ask A Question and we will direct you to our Order Page at WriteDemy. Then fill Our Order Form with all your assignment instructions. Select your deadline and pay for your paper. You will get it few hours before your set deadline.