Exercise 9: Candidate Character Traits and the Presidential Vote

Another reason for using three-variable tables in our data analysis is to examine the joint influence of two independent variables on a dependent variable. To illustrate this situation, we can consider the influence of how perceptions of candidate character traits influenced the presidential vote in 2016. Character traits were emphasized in the 2016 election campaign, partly because both Clinton and Trump were viewed unfavorably overall. Clinton was viewed as especially weak in integrity; Trump received low marks on experience, knowledge, and integrity, as well as seen by many as being too rash and impulsive. Since integrity was a particular problem for Clinton, one that many voters cited as a reason for not voting for her, it should be interesting to analyze how voters perceived the honesty of Clinton and Trump.

Perceptions of Honesty

This dataset contains several items of information about the perceptions of the character traits of the two presidential candidates, including questions that asked the respondents whether or not they thought that Clinton and Trump were honest (D04 and D10). These items asked respondents whether the phrase “he/she is honest” characterized each candidate extremely well, very well, moderately well, slightly well, or not well at all.

Step A. Create and interpret Tables 9A and 9B

To begin, you should create Frequency Tables 9A and 9B that show how voters perceived the honesty of the two candidates (D04 and D10). (Note: to create a frequency table for a single variable, enter the variable name in the box for the row variable and leave the column variable blank. Be sure to use the weighted data.)

These two tables show the frequency distribution of these two variables. Examine the tables and answer the following questions:

  1. What percentage of the respondents had a favorable view of Clinton’s integrity? What percentage had an unfavorable view? In answering this question, which categories of variable D04 would you consider to be a favorable assessment?
  2. What percentage of the respondents had a favorable view of Trump’s integrity? What percentage had an unfavorable view?
  3. How do the perceptions of Clinton’s honesty compare with perceptions of Trump’s honesty? Was one candidate seen as more honest than the other?

You should have concluded that while neither candidate received high marks for integrity, Trump had a clear advantage over Clinton on this character trait.

Step B. Create and interpret Tables 9C and 9D

The next question is whether these perceptions of honesty influenced the vote. You should create Table 9C that shows how perceptions of Clinton’s honesty (D04) were related to the vote. For the reasons suggested in exercise one, you should use the recoded version of A02 that you created for that exercise, so that you examine only the major-party vote (i.e., only the Clinton and Trump voters). After creating this table (Table 9C), create Table 9D, using the same question about Trump’s honesty (D10).

If you ran Tables 9C and 9D as suggested, you should have two tables, each with five columns and two rows. Perceptions of Clinton’s (or Trump’s) honesty (the independent variable) should be on the top of the table (the column variable), and the two-party presidential 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:

  1. How much difference in their presidential voting is there between those who thought that Clinton was honest and those who did not think so? Between which categories does the greatest difference in voting exist?
  2. How much difference in their presidential voting is there between those who thought that Trump was honest and those who did not think so? Between which categories does the greatest difference in voting exist?
  3. Is the relationship in Table 9C stronger, weaker, or about the same as the relationship in Table 9D?

Step C. Create Table 9E

Tables 9C and 9D each indicate that perceptions of both Clinton’s and Trump’s honesty were related to the vote. However, looking at each perception separately does not answer the most interesting question, which is what effect perceptions of differences between the candidates have on the vote. That is, some voters who thought that Clinton was not honest may also have felt the same about Trump, in which case they did not see a difference between the candidates on this attribute. Other voters who thought that Clinton was not honest may have felt that Trump was, and these voters clearly saw a difference between the candidates on this attribute. To examine the joint influence of D04 and D10 on the presidential vote, you should create Table 9E, a three-variable table. In this case, it does not matter whether D04 is treated as the independent variable and D10 as the control variable, or the reverse. In order to simplify the table, first recode D04 and D10 so that they have fewer categories.

If variables D01 and D07 are not recoded, using both variables together to predict the vote will produce a three-variable table containing five subtables, each with five columns, for a total of 25 columns. This would be a complex table that might be difficult to interpret, and some of the columns might have too few respondents in them, which would make the percentages unreliable estimates of the true figures. By recoding D04 and D10, you can produce a table that is simpler. To make your three-variable table as simple as possible, you could recode these two variables so that they each have just two categories: favorable and unfavorable.

In recoding variables D04 and D10, the responses “extremely well” and “very well” clearly could be considered to be favorable assessments of a candidate’s honesty. Similarly, “slightly well” and “not well at all” would seem to be unfavorable assessments. The response “moderately well” is somewhat ambiguous—it is clearly less positive than the first two responses, but more positive than the last two responses.

For some guidance on how this middle category is best recoded, we can look at the result of the previous tables that show how these variables are related to the presidential vote. Tables 9C and 9D show that voters who thought that the term honest characterized the candidate as moderately well were highly likely to vote for that candidate. In both tables, the difference in the vote between the “very well” column and the “moderately well” column is not that great. This suggests that we could recode the top three categories into one group, which is basically favorable in its assessment of the candidate.

An alternative would be to keep the “moderately well” category as a separate group, and we could do this, thereby recoding D04 and D10 into variables with three categories. However, to make this exercise as simple as possible, we recommend combining the top three categories together so that we have a simple dichotomous variable for the perception of each candidate’s honesty.

Step D. Interpret Table 9E

If you ran Table 9E as suggested, you should have a table that consists of two subtables. Each subtable should have two columns and two rows. Both subtables should have the perception of Clinton’s (or Trump’s) honesty (the independent variable) on the top of the table (the column variable), and the two-party presidential vote (the dependent variable) 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). There should be two subtables: one for each of the two categories of responses regarding Clinton’s (or Trump’s) honesty. It does not make any difference whether you treated D04 as the independent variable and D10 as the control variable or the reverse, since both are really independent variables in this analysis. In reading your subtables, 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:

  1. Among those who thought that Clinton was honest and Trump was not, what percentage of respondents voted for Clinton? Among those who thought that Trump was honest and Clinton was not, what percentage of respondents voted for Clinton? These two groups are the voters who saw big difference between the candidates on this attribute.
  2. Among those who thought that both Clinton and Trump were basically honest, what percentage of respondents voted for Clinton? How about among those who thought that both were not honest? These groups are the voters who did not see much difference between the candidates on this attribute.
  3. How many voters were in each of the four possible combinations of views of Clinton and Trump: (1) Clinton honest and Trump not; (2) Trump honest and Clinton not; (3) both not honest; (4) both honest?
  4. Is Table 9E superior to Tables 9C and 9D for examining how perceptions of honesty is related to the vote? What did you learn from Table 9E that you did not from Tables 9C and 9D?