Exercise 2. Marital Status and the Congressional Vote
Step A. Create and interpret Table 2A
Voting behavior is related to several demographic or social characteristics. In this exercise, we examine the relationship of one such characteristic, marital status, to the congressional vote. To examine this relationship for 2016, you should create Table 2A that shows the relationship between the individual’s marital status (R04) and his or her vote in the U.S. House elections (A03).
Table 2A must be interpreted somewhat differently than Table 1A. In looking for a relationship between party identification and presidential vote in Table 1A, we looked for a pattern of decreased support for Clinton as we went from left to right in the table. However, in looking for a relationship between marital status and vote in Table 2A, we should simply look for differences between the columns, not necessarily a pattern of increasing support for one candidate as we go from one end of the table to the other. That is because party identification is an ordinal variable, whereas marital status is a nominal variable. You need to recognize whether you have ordinal or nominal variables in your analysis; the distinction is discussed in the section on data analysis.
To create a two-variable table, you must specify a row variable and a column variable. A common way of setting up a table is to put the independent variable on the top of the table (this is the column variable) and the dependent variable on the side of the table (this is the row variable). For a discussion of independent and dependent variables, see the section on principles of data analysis. When you are in the SDA crosstabulation program, enter your variables in the row and column dialog boxes.
If the table is set up in the above fashion, you normally would want percentages by columns. You should click this option under table options.
One could construct the table so that the independent variable is the row variable (on the side of the table) and the dependent variable is the column variable (on the top of the table). If the table is set up in this manner, then you normally would want percentages by rows. However, in all of the tables in this module, we have followed the convention of having the independent variable as the column variable and the dependent variable as the row variable.
You should be sure that you have the weight on and that you have selected the weighted Ns to appear in the table. If you do not have the weight on, then you will be analyzing the unweighted data, which will not be a representative sample of the American electorate. See the discussion of weighting the data in the section on survey research methods for an explanation of how and why the data are weighted.
Because the data are weighted, which means that individual respondents may count as more or less than one person (e.g., as .75 or 1.35 persons), the number of respondents in each cell (the Ns) probably will not be whole numbers. If you prefer to have Ns that are whole numbers, you can revise the output to do that by using the “revise the display” options that appears to the left of the table that you generated.
If statistics are desired, that option should be checked under table options. For a discussion of the statistics that are commonly used for contingency tables, see the section on data analysis. In these exercises, we have not asked you to generate statistics, but your instructor may suggest doing so.
The SDA crosstabulation program will produce both a table and a chart, but the chart is not necessary, as all of the information that you need will be contained in the table that you generate. You can revise the output to drop the chart if you like by using the “revise the display” option.
If you ran Table 2A as suggested, you should have a table with five columns and three rows. Marital status (the independent variable) should be on the top of the table (the column variable), and congressional vote (the dependent variable) should be on the side of the table (the row variable). Percentages should be calculated by column (i.e., they should sum to 100% for each column). In reading your table, take care to properly interpret the percentages, remembering that they are column percentages, not row percentages.
You should attempt to answer these questions to see if you are able to correctly read the table and interpret the data:
- Overall, what percentage of respondents in the table voted for a Republican candidate in the U.S. House elections? What percentage voted for a Democratic candidate?
- What percentage of married voters cast a ballot for a Democratic congressional candidates? How does this percentage compare to that for voters who never were married? Which group was most Democratic in its voting? Which was the most Republican?
- Overall, what differences exist in the voting behavior of people in different marital situations? How strong of a relationship is there between marital status and voting behavior? How would you briefly summarize this relationship?
Step B. Generate a simpler table
We might wish to create a simpler version of Table 2A by recoding marital status. General information about why it can be desirable to recode variables is in the section on principles of data analysis.
Create Table 2B, a new, simpler table by using a recoded version of marital status, as discussed above. Also, recode the congressional vote to eliminate minor party voters (just as you did in Exercise 1 with the presidential vote), which will simplify your dependent variable. Compare Table 2B to Table 2A.
Table 2A has five categories for marital status. We might want to simplify that table by recoding marital status so that it has fewer categories. This would be especially desirable if we feel that the differences between some categories are small. For example, if we feel that the difference in voting behavior between single individuals and divorced or separated individuals is quite small, and the difference between these two groups and those who are married is much larger, then we might want to recode the variable.
If we want to recode marital status so that it has just two categories, we probably would want to place single and divorced/separated individuals into one category and married individuals into a different category. Conceptually, single and divorced people are in a different family situation than married people are. Moreover, Table 2A shows us that married voters were much more likely to vote Republican than were the single or divorced voters, who were fairly similar in their voting.
It is less clear where we should place those who are widowed. Should widowed individuals be placed with the married individuals or with the single or divorced/separated individuals? The widowed individuals are conceptually between these two groups in terms of their marital situation. They are not currently married, making them similar to the separated or divorced individuals in that sense. On the other hand, most of them probably were married for a long time and became widowed only in old age, and they presumably did not want their marriage to end, giving them more in common with those who are now married. In terms of their likelihood of voting Republican (or Democratic), they are somewhere between these two groups—more Republican than single people, but less Republican than married people. One might suggest that if widowed individuals are not clearly in either category, they should be left as a separate category. However, there are only about 100 widowed voters in the survey, so this might result in small Ns in our tables if we try to further explore this relationship by introducing another variable (e.g., gender or income) into the analysis.
Those who are not married but who have a domestic partner present a similar problem for recoding. Having a domestic partner would in some ways make them close to those who are married, as both groups are people who presumably are in a committed, monogamous, romantic relationship, so it would seem that they should be recoded into the married group, not into the single or divorced group. On the other hand, these people have chosen not to get married, so they are in a different legal status than those who are married. Also, this group is very strongly Democratic in its voting, making them quite different from the married individuals in their political behavior, even if they are similar in some ways in their marital situation.
While a case can be made for either decision regarding how widowed individuals or those with domestic partners should be categorized, we think that it makes more sense to put both groups with the married individuals, since they seem more similar in their marital or domestic situations or experiences, so that is how we suggest that you recode marital status. Thus, your recoded marital status variable should have two categories:
- individuals who are married, widowed, or living with a domestic partner
- single and divorced or separated individuals
Although one could argue that both the widowed and living with a domestic partner groups should each be retained as separate categories, rather than recoded into the married group, doing so would leave us with four groups after recoding, which would not be much different from the five groups in the original variable. If our goal is to simplify the original table, reducing marital status from five categories to four does not achieve very much.
Another possibility would be to combine widowed individuals with those who have domestic partners into one separate category, thus yielding three categories:
- married
- widowed and domestic partners
- single, divorced, and separated individuals.
However, this method combines two quite different groups, widowed individuals and those with a domestic partner, into one category even though they do not seem to have much in common. Each group seems to have more in common with married people than they do with each other.
We emphasize that the method of recoding marital status that we suggest is not the only logical way of doing so. Someone else might think that it makes more sense to place those who are living with a partner into the same category as those who are single or divorced. Others might think that widowed individuals belong in that category as well. Still others might opt for having three categories. Recoding often involves making decisions on theoretical grounds. One must think about the underlying concept that the variable is attempting to measure and recode in a manner that best captures that concept. Sometimes there is not one clearly best way to recode a variable.
SDA contains detailed instructions, including examples, on how to use the program to recode a variable.
To recode a variable, enter the recoding instructions in the recoding dialog box that appears when you click on the recode link next to the dialog box where you enter the variable name in the crosstabs builder.
The basic syntax for recoding is simple. Begin with an “r:”, then specifiy which new values (those on your new, recoded variable) should be equal to which old values (those on the original variable). For example, to recode party identification (A07) in the manner suggested above, the recode syntax should look like this: r: 1=1-3; 2=4; 3=5-7. This will create a new value of “1” that will be composed of old values 1, 2, and 3 (strong, weak, and independent Democrats); a new value of 2 that will be equal to old value 4 (independents); and a new value of 3 that will be equal to old values 5, 6, and 7 (independent, weak, and strong Republicans).
If you want to drop some categories in the original variable, just do not include them in the recode statement. For example, to recode presidential vote (A02) to exclude minor party voters, specify the following recode: r: 1=1; 2=2. This drops the value of “3” on the original variable, leaving only the Clinton and Trump voters.
When you recode a variable, it usually is helpful to attach labels to the new values. To do this, simply add the new label in quotation marks after the statement that indicates what new value equals what old values. For example, the following recode specification will add labels to the recoded version of party identification described above: r: 1=1-3 “Democrat”; 2=4 “Independent”; 3=5-7 “Republican”.
You may intend to generate a number of tables with the same variable. To avoid having to enter the recode specification each time that you want to run a table with the variable, you can copy the recode statement to some location (e.g., a Word document) and paste it to the recode syntax window for subsequent table constructions that use the variable.
Step C. Interpret Table 2B
If you ran Table 2B as suggested, you should have a table with two columns and two rows. Recoded marital status (the independent variable) should be on the top of the table (the column variable), and congressional vote (the dependent variable) should be on the side of the table (the row variable). Percentages should be calculated by column (i.e., they should sum to 100 percent for each column). In reading your table, take care to properly interpret the percentages, remembering that they are column percentages, not row percentages.
You should be able to answer these questions to see if you are able to understand the data and the tables:
- How does Table 2B compare to Table 2A? Do you find this table easier to read and interpret?
- Do you come to the same basic conclusion about how marital status is related to voting behavior regardless of whether you examine Table 2A or Table 2B?
- How strong is the relationship between marital status and voting behavior in Table 2B? Would you describe it as fairly strong or fairly weak?
- Why do you think that the differences in Table 2B exist? What might explain why married individuals were more likely to vote Republican than were those who were single or divorced/separated? What additional variable might be introduced into the analysis to test your explanation?