"Why Buy the Cow When You Can Get the Milk for Free?"

by Susan Cotts Watkins

On the first day of my class in Introduction to Sociology: The Family, I used to despair. The students and I are there for incompatible reasons. They are looking forward to the families that they want to establish, and looking backward to the families that they are leaving (usually, more than half the class are freshmen or sophomores). As if we are all in a cartoon, I fancy I can see their anxious questions in balloons over their heads: How will I find a mate? My parents divorced--what has this done to me? How can I avoid repeating my parents' mistakes, and protect my marriage (which I want to last until death do us part)? But I don't know how Lisa or Andrew will find a mate, or, if they do, whether their parents' history is their own destiny. I want them to take a sociological perspective and ask other questions: What are the patterns of family formation and dissolution? How do those in the contemporary U.S. differ from those in our past, or in Kenya, where I am currently doing research on fertility change? What are the theories that help us to recognize these patterns? How do we confront our theories with empirical evidence? How do we evaluate the quality of that evidence?

When I began teaching The Family at Penn in 1982, the gulf often seemed too great. After a few years I hit on an approach I think has been successful (an intuition supported by higher student evaluations). Now, on the first day of class I survey their attitudes to the family, taking some options from national surveys: People should only get married because they are deeply in love...Marriages are better when the husband works while the wife runs the home and cares for the children...When parents divorce, children develop permanent emotional problems. The data from the survey are entered into the computer (courtesy of Bob Douglas, director of social science computing) and by the third week of classes I can show the class where they as individuals fit into the aggregate distribution.

"Do you expect to cohabit?"

By the time I introduce the class data, we've read the chapters in the text on the family and gender, class, and race and ethnicity, and we've read about the crystallization of 19th century U.S. family ideals (separate spheres, where the man is the breadwinner and the woman in charge of the domestic sphere). How far have they come? I ask them. We turn to the class data, and year after year we find that the class has abandoned the notion of separate spheres (though not the expectation that women will do most of the housework and childcare). They still believe firmly in love as a basis for a marriage that will last forever--even if most of them, it turns out, believe that when a marriage is emotionally or sexually unsatisfying, di-vorce is a solution, even if there are children. We begin to formulate hy-potheses about patterns in the data. Do they expect that women in the class will have different attitudes toward the "Ozzie and Harriet" family? They usually say yes, and have sensible explanations (drawn more from their own intuitions than from the sociological theory of the text): the women in the class are getting an education and they expect them to want returns on this "investment in human capital." We then examine the class data, and usually find that their hypotheses are not supported: there are no significant differences by gender. I then gently lead them from aggregation at the level of the class to aggregation at the level of the nation. We talk about the representativeness of our class sample, and they don't find it hard to believe that Penn students are different from the national population: better educated, with parents likely to have higher incomes and levels of education than the national average.

It becomes easier and easier to move from their individual attitudes to the patterns in the class and to national patterns. What about cohabitation, I ask: do you expect to cohabit? The data show that most of them do. Why? They formulate hypotheses. Many of them believe that cohabitation is a prophylactic against divorce--as Michele phrases it, one can try living with a partner, with the option of abandoning ship before the vows are publicly pledged. We look at national data on reasons for cohabitation, and, lo and behold, Penn students aren't so different after all: on a national survey, getting to know a partner is the modal reason for cohabiting before marriage. Does cohabitation protect against divorce? We look at the national data comparing the duration of marriage for those who have cohabited before marriage and those who have not. Alas, cohabitation is not a help. Why not? If I'm lucky, a bright student will point out before I do that people who choose to cohabit may be different from those who choose not to--different in ways that may also affect the probabilities of divorce. They are now well into discovering one of the major differences between social sciences and natural sciences, that social sciences can't randomly assign people to categories (Lisa, you cohabit, Andrew, you don't). Selection biases become intuitively clear, and the possibility that cohabitation may yet be a prophylactic against divorce revives their spirits.

By the middle of the semester, they are ready to formulate a hypothesis on their own and test it with the class data for their paper. They are, I've discovered, reluctant to examine hypotheses about race, but hypotheses about the effect of parental divorce on children's attitudes are popular. Many believe firmly that those whose parents have divorced are likely to be more cautious in their attitudes towards marriage. They can test this with a cross-tabulation of Living together before marriage makes good sense by Are your parents presently married to one another? (with the option of going on to separated/divorced/remarried/never married/deceased/other). Also popular are hypotheses having to do with female labor force participation: Does having had a working mother make a student more likely to expect to work while her (yet unborn) children are aged 0-6?... Are members of sororities or fraternities more "promiscuous" (their words, not mine) than non-Greeks? They can cross-tab Greek membership by It's o.k. to have sex on the first date.

The paper is to be written in a standard format: the clear statement of a testable hypothesis and its theoretical justification; a discussion of the data and methods (simple cross-tabs, but some students move on to odds ratios and significance tests); results; discussion. They are to attach the computer printouts. (Before I hit on this assignment, I worried a lot about plagiarism. But the numbers change with each survey, and including the original printouts is a further protection). The TAs offer to read their hypotheses before they start testing them (invariably, some students want to test hypotheses that can't be tested with the available data, e.g. Women at Penn will put their career before their family). The TAs also offer extra office hours in the UDAL lab in McNeil to those who struggle with our handouts on how to use Stata, the statistical package.

"Ask a different question..."

When a cherished hypothesis is not supported, they are downcast, despite my repeated Popper-ian points about how science marches on by disconfirmation. Popper is o.k. in the abstract, it seems, but not in practice. They come to office hours (or, now, e-mail) worried that they don't have a paper, and rephrasing Popper doesn't cheer them up much. They start to do what scientists often do, to recognize the inadequacies of the available data or the sample. Perhaps students aren't willing to admit on a survey that they believe that sex on the first date is o.k.-- or perhaps some (e.g. men) are, but others (e.g. women) aren't? So they cross-tab the data by the gender of the respondent. Perhaps the age category 0-6 is too broad, and the survey should ask about expectations for working when the children are 0-2? Ask a different question next time, they suggest. Or they claim that Penn students are different, a better sample is required.

Some of them will go on to take more sociology courses, and a few of them become majors. Most of them disappear from my sight. No doubt they will forget many of the "facts," and perhaps they should: after all, the total fertility rate is likely to change, and one theme of the course is change in family patterns associated with changes in economic organization and in ideologies. But I'd like to think that they carry some of the sociological perspectives with them, and some of their criticisms about data. When they hear The American family is disintegrating, I hope that some, at least, will say How do you measure disintegration? What are your data? How did you test that hypothesis?

And in the meantime, some of the papers are terrific: one of my personal favorites is the one on cohabitation that provided my title.

Talk About Teaching is in its second year as a series sponsored by the Lindback Society and the College of Arts and Sciences.

Dr. Watkins is a professor of sociology and a member of the Population Studies Center


February 20, 1996
Volume 42 Number 21

Return to Almanac Home Page.

Return to index for this issue.