Executive Summary

Donald Trump has presented immigration and international trade as the causes of the diminished economic prospects of white working-class voters, the core of the former president’s political base. This narrative suggests that voters’ attitudes about immigration and trade reflect their actual lived experiences with immigration and trade—in other words, Trump supporters voted the way they did because they experienced direct negative impacts from both immigration and U.S. trade in a global economy. The power of Trump’s antitrade and anti-immigrant political and policy message is that—as some journalists assumed—voters supported him because he promised to reduce trade and immigration to benefit those who felt vulnerable to their effects. If this were true, it would logically follow that his support among economically self-interested voters would be greatest in places with higher levels of immigration and international trade.

However, our research found that to the contrary, Trump voters in 2016 and 2020 tended to live in places with relatively low levels of immigration and international trade. Anti-immigrant and antitrade attitudes more consistently and strongly explained voting for Trump in 2016 and 2020 than actual levels of immigration and trade where voters lived. The research uncovered a phenomenon the author calls the “Trump Paradox,”¹ in which support for Trump was highest in counties with the lowest exposure to immigration and trade, but which had the strongest anti-immigrant and antitrade attitudes.

This issue brief further examines the Trump Paradox by race, ethnicity, and gender, in 2016 and 2020.

The main findings are:

1. The Trump Paradox among white voters strengthened between 2016 and 2020.
2. White males, not females, were the main driver of the Trump Paradox in 2020.
3. Latino voters did not show a strong Trump Paradox in 2020.
4. New Latino female voters —those who voted in 2020 but did not vote in 2016 — were more likely to vote for Trump in 2020 than existing ones, and they had lower educational attainment on average.

First, we found that non-Hispanic white voters were the basis of Trump and Republican support, after controlling for immigration and trade attitudes as well as demographic, partisan, and economic characteristics.² Research shows that the Trump Paradox was fundamentally a white phenomenon, with white support for Trump highest in counties with the lowest levels of immigration and international trade, but with the strongest anti-immigrant and antitrade attitudes.

Further, we found support that from 2016 to 2020, the political power of Trump’s narrative appeared to unravel as his support declined in those counties most exposed to immigration and international trade. To further understand this evolution in the Trump Paradox, we conducted a gendered analysis—for white men and white women—that uncovered significant variations in attitudes on immigration and trade and Trump-voting behavior. The analysis found that the Trump Paradox increasingly revealed itself as a white male Trump Paradox while the significant differences of white women attitudes and voting played a growing role in the unraveling of the political power of the Trump Paradox.

We then performed a similar analysis for Hispanic voters, which found that in 2016, Hispanic voters from regions with more imports from China and Mexico were less likely to vote for Trump. However, the effect was not as significant in 2020.

Finally, this issue brief further examines Hispanic voters and other voter groups to better understand how changes in the composition of the U.S. electorate impacted the election dynamics from 2016 to 2020. This section includes a Trump Paradox analysis for Hispanic voters in 2016 and 2020 and statistics on certain aspects of various groups by gender, race, and ethnicity. Interestingly, we found that new Hispanic female voters (who did not vote in 2016 but voted in 2020) exhibited a much higher propensity to vote for Trump than existing Hispanic female voters (who voted in both 2016 and 2020), a phenomenon that was not seen in other gender, racial, or ethnic groups. Of all Hispanic female Trump voters in 2020, about 25% were new voters, also higher than other selected groups.

Introduction

Donald Trump presented immigration and international trade as the cause of the diminished economic prospects of white working-class voters, the core of his political base. Our past research in 2016—the first that compared actual immigration and trade exposure with attitudes and Trump voting—demonstrated that white voting for Trump was unrelated to immigration levels and, paradoxically, was strongest in counties with low levels of international trade. Instead of actual immigration and trade levels, anti-immigrant and antitrade attitudes more consistently and strongly explained voting for Trump in 2016. We termed this discrepancy the “Trump Paradox.”

We also find support that from 2016 to 2020, Trump’s narrative unraveled as his support declined in those counties most exposed to immigration and international trade. Although Trump elaborated a white nationalist narrative based on antiimmigrant and antitrade politics that many of his supporters accepted as truth, it was shown that virtually no aspects of Trump’s simple narrative on trade and immigration have significant factual basis. Our analysis suggests that Trump rallied white voters to support him by stoking xenophobic and racist sentiments about immigration and trade, rather than the presence or threat of actual immigration and trade driving turnout for Trump.

Recent research found that from 2016 to 2020, Trump gained support among Hispanic voters, especially among Hispanic women, Hispanics without a college degree, and new Hispanic voters. Pew Research Center found that Trump gained about 10 percentage points among Hispanic voters from 2016 to 2020.³ While Joe Biden still won the majority of Hispanic voters in 2020, Trump’s support surged among Hispanic voters lacking a college degree.⁴ The gap between Biden’s advantage over Trump was notably smaller for those without a college education (55% to 41%) than for those with a college education (69% to 30%). The progressive research firm Equis also found a gender element at play: Support for Trump grew dramatically from 2016 to 2020 among Hispanic women.⁵

The significance of the new Hispanic voters cannot be overlooked. A report by the Democratic data firm Catalist showed the number of new Hispanic voters increased by 31% from 2016 to 2020, making them the fastest-growing voter group,⁶ and almost 25% of Hispanic voters voted for the first time in 2020, compared to 14% of the electorate overall.⁷ Equis data suggest the swing toward Trump was the result of a combination of the persuasion of existing voters who were previously left-leaning and the mobilization of new and infrequent voters.⁸ Carlos Odio, co-founder and senior vice president at Equis Labs, and his colleagues found that Hispanics supporting Trump “grew on the margins thanks to a combination of defections and new voters—with likely a greater number of the latter.”⁹

In this issue brief, we expand upon the 2016 research to include a gendered, racial, and ethnic analysis of the Trump Paradox and analyze data for the 2020 election.

Methodology¹⁰

We updated the Trump Paradox analysis using data for 2020, which we then compared with the analysis for 2016. In this analysis, we look at how exposure to actual immigration and international trade (after controlling for other individual-level factors) affected voters’ behavior—specifically, whether they voted for Trump in the 2016 and 2020 Presidential Elections.

Actual county-level data on shares of Hispanic immigrants come from the U.S. Census Bureau’s American Community Survey (ACS). County-level data on imports from Mexico and China are from the U.S. Census Bureau’s County Business Patterns. Individual-level data are from the Cooperative Election Study (CES, formerly the Cooperative Congressional Election Study, CCES), which is a nationally representative sample of more than 60,000 respondents, who were asked about their attitudes on immigration and trade. The surveys also include information on partisanship, education, gender, age, income, and employment status, which we consider important demographic, political, and economic controls. For education and age, we use college educated (bachelor’s degree or higher) and older than 65 years old, as commonly operationalized in studies of voter behaviors.¹¹ Family income, status as temporarily laid off or unemployed, and employment in the manufacturing sector represent socioeconomic conditions.

For individual attitudes toward immigration and trade, we used items from CES. For immigration, the items were agreement or disagreement on whether the U.S. government should deport undocumented immigrants (2016) or whether it should build the border wall (2020); for trade, the items were whether voters supported the Trans-Pacific Partnership (TPP) in both 2016 and 2020. In 2016, the specific question asked about attitude on immigration was whether agree “illegal aliens should be identified and deported,” but the same question was not asked in the 2020 survey. Instead, we used an alternative measure from 2020 CES, which was whether the respondent agreed that “the U.S. should increase spending on border security by $25 billion, including build a wall between the U.S. and Mexico.” Although the immigration questions were quite distinct, the trade questions were similar but slightly different. The 2016 item queried simply whether respondents were for or against the TPP, while the 2020 question asked whether the United States should withdraw from the TPP. To combine the county and individual effects, we linked the individuals to the counties in which they resided.

Then we added gender, race, and ethnicity to our analysis to better understand which changes in the composition of voters contributed to Trump’s gain. Summary statistics are also based on data from CES.

Findings

Our findings are as follows:

1. The Trump Paradox among white voters strengthened between 2016 and 2020.
2. White males, not females, were the main driver of the Trump Paradox in 2020.
3. Latino voters did not show a strong Trump Paradox in 2020.
4. New Latino female voters —those who voted in 2020 but did not vote in 2016 — were more likely to vote for Trump in 2020 than existing ones, and they had lower educational attainment on average.

1. The Trump Paradox among White Voters Strengthened Between 2016 and 2020

Among white voters, the Trump Paradox was stronger in 2020 than in 2016. After controlling for other factors, we found that whites who resided in counties with higher concentrations of Hispanic immigrants were less likely to vote for Trump in 2020, but this effect was not present in 2016. We also found negative and statistically significant correlations between Trump voting in both 2016 and 2020 and levels of Mexican and Chinese imports. In other words, after controlling for other factors, whites who resided in counties with more imports from China and Mexico were less likely to vote for Trump in 2016 and 2020.

At the same time, in both 2016 and 2020, we found that anti-immigrant and antitrade individual attitudes were significantly correlated with higher likelihood to vote for Trump.

This finding provides strong evidence that despite Trump’s message about Hispanic immigrants taking jobs and bringing crime, white voting for Trump had little to do with their actual presence. Table A1 shows the correlations between white votes for Trump and actual levels of Hispanic immigration and international trade as well as voter attitudes toward trade and immigration. Positive numbers represent a positive correlation, and vice versa. The number of asterisks represents how significant the correlation is; the more asterisks, the stronger the correlation.

Another way to understand the strengthening of the Trump Paradox is by looking at the characteristics of the counties that supported Trump. In 2020, white voting for Trump became even more paradoxical as his support became more concentrated in counties with lower levels of immigration and trade. Figure A1 confirms the negative link between Trump voting and actual immigration and trade. Counties that remained Democratic from 2016 to 2020 had the highest levels of immigration and trade, while counties that remained Republican from 2016 to 2020 had the lowest levels of immigration and trade.

2. White Males were the Main Driver of the Trump Paradox in 2020

If we look more closely into the nuances of white voting behavior, we see that gender plays an important role (see Table A2). For 2016 and 2020, a smaller share of white women voted for Trump than white men, and White women overall demonstrated weaker anti-immigrant and antitrade attitudes.

The Trump Paradox manifested itself in different ways for white men and white women. In the analysis, white men were the main Trump Paradox driver, meaning their Trump voting was more paradoxical than that of white women. After controlling for other individual-level factors, we found that white men were more likely to vote for Trump in counties with lower levels of immigration and trade (and vice versa). At the same time, white women’s Trump voting was less correlated with county-level immigration and trade (see Table A3). Instead, their voting was more strongly correlated to other factors such as attitudes on immigration and trade, partisanship, and educational attainment. And for both white men and white women, anti-immigrant and antitrade attitudes consistently and strongly explain voting for Trump in 2016 and 2020.

3. Hispanic Voters did not Show Strong Trump Paradox

Although the Trump Paradox was not strong among Hispanic voters, it was stronger in 2016 than in 2020 (see Table A6). In 2020, Hispanic voters’ choice to vote for Trump did not show clear correlation with actual county-level immigration and trade, but was again significantly related to their anti-immigrant and antitrade attitudes.

Further analysis shows interesting differences in voting behavior by gender as well as race and ethnicity. All racial and ethnic groups we analyzed showed important gender differences in Trump support, with men supporting Trump more than women across all groups. We also found important gender differences on attitudes concerning immigration and trade (see Table A4).

While the total number of Hispanics voting for Trump grew from 2016 to 2020, we observed a gender difference in this change: Hispanic women who voted for Trump grew more than Hispanic men. Interestingly, for existing voters (who voted in both 2016 and 2020), the largest increase in voting for Trump occurred among Hispanic women (see Table A5). About 18% of the Hispanic women who voted for a non-Trump presidential candidate in 2016 voted for Trump in 2020.

4. New Hispanic Female Voters were more Likely to Vote for Trump in 2020 than Existing Voters and they had Lower Educational Attainment on Average

We also looked at the new voters—those who did not vote in 2016 but turned out to vote in 2020,—and found that a larger share of Hispanic women that voted for Trump were new voters, as compared to other gender, racial and ethnic groups, as shown below:

In 2020:

• About 12% of all Trump voters were new voters who had not cast a vote in 2016.
• About 11% of all white female Trump voters were new voters.
• About 9% of all white male Trump voters were new voters.
• About 18% of all Hispanic male Trump voters were new voters.
• About 25% of all Hispanic female Trump voters were new voters.

Further analysis shows that compared to other groups, Hispanic women who were new voters in 2020 demonstrated much more support for Trump than existing voters who voted in both 2016 and 2020 (see Figure A2).

In terms of educational attainment, among Trump voters in 2020 who did not vote in 2016:

• About 82% of all Trump voters had less than a 4-year degree.
• About 81% of white female Trump voters had less than a 4-year degree.
• About 84% of white male Trump voters had less than a 4-year degree.
• About 84% of Hispanic male Trump voters had less than a 4-year degree.
• About 89% of Hispanic female Trump voters had less than a 4-year degree.

These numbers show that a larger share of Hispanic female new Trump voters have less than a 4-year degree, as compared to other gender, racial and ethnic groups.

Conclusion and Recommendations/Further Discussion

Our analysis has confirmed the existence of the Trump Paradox in general as well as many elements and nuances of the Paradox. As we proceeded from 2016 to 2020, we found a stronger paradoxical relationship between immigration and Trump voting for white voters: the less immigration in the county they reside, the more likely they vote for Trump. We also found negative and statistically significant correlations between Trump voting in both 2016 and 2020 and levels of Mexican and Chinese imports for white voters.

We also see some differences in gender. Women in general were not as supportive of Trump as men. But we also saw an unusual uptake in Trump’s support among Hispanic women in 2020, which can be attributed to new voters supporting Trump at a higher rate compared to existing voters. It is also important to point out that in 2020, a higher percentage of new Hispanic female Trump voters have a lower educational attainment compared to other gender, racial and ethnic groups.

End Notes

¹ Raul Hinojosa-Ojeda and Edward Telles, The Trump Paradox: Migration, Trade, and Racial Politics in U.S.-Mexico Integration (Oakland, CA: University of California Press, March 2021).

² In the following text, non-Hispanic white will be referred to as “white,” and the same rule applies for non-Hispanic Black (“Black”) and non-Hispanic Asian (Asian).

³ Ruth Igielnik, Scott Keeter, and Hannah Hartig, Behind Biden’s 2020 Victory: An Examination of the 2020 Electorate, Based on Validated Voters, Pew Research Center, June 30, 2021, available online.

⁴ Igielnik, Keeter, and Hartig, Behind Biden’s 2020 Victory.

⁵ Sabrina Rodriguez, “The GOP Is Gaining among Texas Hispanics. Women Are Leading the Charge,” Politico, February 7, 2022, available online.

⁶ Nicole Narea, “New Data Helps Explain Trump’s Gains among Latino Voters in 2020,” Vox, May 19, 2021, available online.

⁷ Yair Ghitza and Jonathan Robinson, “What Happened in 2020,” Catalist, accessed March 6, 2023, available online.

⁸ Eric Levitz, “Latinas Drove Trump’s Gains with Hispanic Voters in 2020,” New York Magazine, April 8, 2021, available online.

⁹ Equis Research, “2020 Post-Mortem (Part One): Portrait of a Persuadable Latino,” Medium, April 2, 2021, available online.

¹⁰ For details on data sources and methodology, see Raul Hinojosa Ojeda and Edward Telles, “Trump Paradox: How Immigration and Trade Affected white Voting and Attitudes,” Socius: Sociological Research for a Dynamic World, vol. 7, 2021, pp. 1–12, available online.

¹¹ John Altick, Lonna Rae Atkeson and Wendy Hansen. “Economic Voting and the 2016 Election,” Unpublished manuscript.