Learn: A Tutorial on Running and Interpreting Crosstabulations With a Third Control Variable

Your crosstabulations in the last tutorial revealed that city size is a potential "suspect." However, there are a lot of other potential "suspects" or explanatory factors for civic disengagement. These additional suspects can complicate social science investigation. Was civic disengagement due to something else? On the bottom of page 205-206, Putnam writes,

When Putnam refers to "holding constant" in the quote above, he is referring to the idea of scientific control. You are likely to be somewhat familiar with the idea of scientific control , which conjures up the image of a scientist working in a laboratory with a white lab coat.

Scientists in the laboratory are able to isolate potential explanatory factors. Sometimes social science investigators are able to isolate factors as well using the experimental research method, but much of social science research takes place outside the laboratory--in the messy world of society where crimes also take place. Therefore, social science investigators collect as many potential suspects as they can in order to "rule them out" as possible explanations for what they are investigating. This is what Putnam does in chapters 10-15 of his book, Bowling Alone.

Overall, in order to "rule out" potential suspects, social science investigators often use procedures of control statistically, rather than through experiments. One statistical procedure used to do this is controlling for a third variable in a crosstabulation.

A lot of the graphics that Putnam uses in the book illustrate controlling for third variables. For example, figure 49 on page 200 illustrates the relationship between work status and club meeting attendance, holding constant or controlling for whether one works out of necessity or for satisfaction. As Putnam writes, "full-time work significantly depresses club attendance, regardless of whether work is a choice or a necessity."

Let's extend the crosstabulation that you produced on city size and working on a community project by controlling for homeownership, one of the characteristics that Putnam holds constant on page 206.

You will get a lot of results, since the control procedure produces a separate crosstabulation for each code that you have for your control variable. In the case of homeownership, there are two codes (0 = no, 1 = yes). Therefore, your results are two tables--one table is only people who do not own a home and the other table is only homeowners. In this way, it is clear how you are able to hold constant or control for the additional independent variable of homeownership.

Let's interpret these results.

Return to the main page for Exercise 3 to practice producing and interpreting your own crosstabulation that controls for a third variable.