Is President Donald Trump likely to be reelected in November of 2020? In this post, we evaluate the president’s reelection prospects and the structure of his support among voters with an analysis of data from the November 2018 and October 2019 editions of the Grinnell College National Poll (GCNP).
Our findings indicate that the president is facing an uphill road to reelection. In both polls, the percent of likely voters who said they would definitely vote for the president was nearly 10 points less than the percent who said they would definitely vote for someone else. We were struck by the overall stability of the electorate on this question even though nearly a year passed between the two polls.
When it comes to opinion on whether Donald Trump should be reelected, most people say they have already made up their mind and only a small portion are persuadable.
The November 2018 and October 2019 editions of the GCNP both asked likely 2020 voters the following question: “If a vote for president of the United States were held today, would you definitely vote to reelect President Trump, consider someone else, or definitely vote to elect someone else, or would you not vote?”
Respondents to both polls gave similar answers. In November 2018, 35 percent of likely voters said they would definitely reelect the president, compared to 45 percent who said they would definitely vote for someone else. In October 2019, the divide was 38 percent to 47 percent. From a statistical standpoint, these responses are effectively the same.
To better understand the structure of the president’s support among our respondents, we first conducted an Ordinary Least Squares regression analysis in which we estimated the relationship between a set of standard demographic variables such as age, education, gender and party identification with support for the president’s reelection. These variables are routinely correlated with voting choices. See, for example, the extensive discussion of how variables like party and ideology are related to voting decisions in Presidential Elections: Strategies and Structures of American Politics (Polsby, et. al. 2012).
Figure 1 plots regression coefficients (dots) and 95 percent confidence intervals (lines) for key variables that might reasonably be associated with a president’s reelection chances. We use the color blue for the November 2018 poll and red for the October 2019 poll.
There are two main takeaways. First, the variables that have a statistically significant impact on opinion about the president’s reelection are party identification, ideology, education, age, a respondent’s race, and a respondent’s views on the state of their personal finances. Second, the findings between the two polls are statistically indistinguishable between the two time periods.
The variable for the president’s reelection is coded 0 for respondents who would definitely vote for someone else, 1 for respondents who would consider a candidate other than the president, and 2 for respondents who would definitely vote for President Trump. For each independent variable we analyze, a positive relationship means that support for the president’s reelection increases as the independent variable increases. A negative relationship means that support for the president’s reelection declines as the independent variable increases. In the chart above, independent variables can be considered positive and statistically significant if their 95 percent confidence interval is to the right of zero, and negative and statistically significant if their 95 percent confidence interval is to the left of zero.
Not surprisingly, party identification has a strong relationship with support for the president’s reelection. In our analysis, we treat Democrats and independents as dummy variables, using Republican as a base category, or point of comparison. Both Democrats and independents are less likely at statistically significant levels to say they will definitely vote for President Trump’s reelection. Similarly, moderates and liberals (also treated here as dummy variables, with conservatives as the base category), are less likely to support the president’s reelection than conservatives.
Both polls showed that age has a small positive effect on support for the President’s reelection, indicating that support is higher among older respondents. This finding was of about equal magnitude in both polls, but only statistically significant in November of 2018. Education also has a small but statistically significant negative effect in both polls, indicating that respondents with higher levels of education are more likely to oppose the president’s reelection than those with lower levels of education.
Respondents to the GCNP are asked if they identify with a particular racial or ethnic group. The resulting sample only contains large enough numbers of people who identify as white, black or Latinx for us to write about our findings in regard to race with statistical confidence. Our samples of other groups are too small for accurate analysis, and so we put them together into a group we label “Additional” for the purpose of this analysis.
As above, we treat racial identification as a dummy variable in which white respondents serve as the base category and black and Latinx respondents are in the model. Both polls show that respondents who identify as black are less likely at statistically significant levels than white respondents to support the president’s reelection.
Finally, we analyze whether or not respondents who believe that they are moving closer to their personal financial goals are more likely to support the president’s reelection. Both polls show a positive and statistically significant relationship between support for reelection and the belief that one’s personal finances are improving.
Next, we turn to a variable by variable analysis of support for the president’s reelection among likely voters using data from the November 2018 poll.
The first relationship we looked at is age. Figure 2 shows a weak positive correlation between age and support for the president’s reelection. Just 16 percent of respondents who are in the under-25 age group support Trump’s reelection compared to 41 percent among those 65 and older. Interestingly, the highest amount of support we measured comes from those who are aged 40-44 at 47 percent.
Figure 3 depicts the relationship between support for Trump’s reelection and party identification. This variable is scaled to assess the strength of a respondent’s affinity to their political party and is based on the Question 102 of the GCNP. Predictably, we find a strong relationship between party identification and support for the president’s reelection. The highest proportion of support comes from respondents who identify as a strong Republican, with 86 percent saying they will definitely vote to reelect the president. The president’s support is weaker among respondents who lean Republican at 52 percent. By comparison, just under 31 percent of independents say they will definitely vote to reelect the president and virtually no Democrats say they will do so.
The next significant variable we looked at is political ideology, with results shown in Figure 4. The classifications are conservative, moderate, and liberal. The greatest proportion of those who support Trump’s reelection are respondents who identify as conservative at 68 percent, followed by moderates at 18 percent, and liberals at 9 percent.
Previous research finds that presidents who govern over periods of economic growth are likely to be reelected (see, for example, the October 2016 special issue of PS: Political Science & Politics for a series of articles evaluating the role of the economy in presidential elections). While the November 2018 GCNP did not ask respondents about their attitudes about the economy overall, it did ask whether they believed they were growing closer or farther away to their goals for their personal finances. Figure 5 illustrates support for the president’s reelection by respondents’ evaluation of their personal finances. Over 38 percent of those who believe they are moving closer to their financial goals say they will definitely vote to reelect the president, compared to nearly 23 percent of those who say they are moving farther away.
Figure 6 illustrates the relationship between a respondent’s self-identified race and support for the president’s reelection. The proportion of respondents who are most likely to vote for Trump are those who identify as Latinx, at 44 percent, but the very wide confidence interval for this group gives us substantial uncertainty about this finding. Likely voters who identify as white report they will definitely vote for the president at a level of 40 percent. The smallest measurable level of support comes from those who identify as black at six percent.
Using data from the November 2018 and October 2019 editions of the Grinnell College National Poll, we analyzed support for President Trump’s reelection. Both polls showed that more likely voters say they will definitely vote for someone besides the president than for him. Additionally, the structure of his support among likely voters was highly stable across the two polls, with little movement in the groups that support and oppose the president.
What do these findings suggest about the 2020 presidential race? First, the approximately 10-point gap between those who support the president’s reelection versus oppose it suggests that the president has a difficult road ahead in the 2020 race. Previous research finds that presidents who govern over a strong economy should be well positioned to win reelection. Our results indicate that President Trump is not benefitting from the strong economy in the way that past presidents have. Despite the passage of a year between the polls, the overall stability in our findings suggests attitudes in the electorate toward the president have calcified. Very few voters remain who are open to persuasion. The task of the candidates over the next year will be primarily to mobilize their key supporters, and then to zero in on the small number of persuadable voters who may be the key in deciding a few closely contested states.
Polsby, Nelson, Aaron Wildavsky, Steven Schier, David Hopkins. 2012. Presidential Elections: Strategies and Structures of American Politics. Thirteenth Edition. Lanham, MD: Rowman & Littlefied.
Peter Hanson is associate professor of political science at Grinnell College and the director of the Grinnell College National Poll.
Max Hill ‘20 is a political science major at Grinnell College with a concentration in statistics.