Attitudes about racial and ethnic groups helped fuel Donald Trump’s apparent victory in the Republican presidential primary. Trump performed best among Republicans who held unfavorable views of African Americans, Muslims, immigrants, and minority groups in general.  Perceptions that whites are treated unfairly in the United States, and that the country’s growing diversity is a bad thing, were also significantly associated with support for Trump in the primaries.

But whether attitudes about race and ethnicity will matter in November’s general election is a different question. Voting in presidential general elections are consistently shaped by factors, most notably party identification and the incumbent president’s job performance, that could potentially overwhelm other considerations.

We can gain some early insights into this question, though, by comparing the impact of racial attitudes on the choice between Trump and Hillary Clinton to their impact on the choice between Clinton and other Republicans.

This type of comparison has been useful in previous elections. For example, in the spring of 2008, polling showed that racial resentment, anti-black stereotypes, attitudes about Muslims, ethnocentrism, opposition to interracial dating, and even living in areas with high rates of racist Google searches, were all strongly associated with the choice between Barack Obama and John McCain, but much less strongly associated to the choice between Hillary Clinton and McCain. This suggested that Obama activated people’s underlying racial attitudes.

If Trump’s candidacy has the same impact in 2016 then we should also see a stronger impact of racial attitudes in a Clinton-Trump match-up than if Clinton were facing Republicans like Ted Cruz and Marco Rubio.  That’s exactly what the graphs below show:

These graphs are based on the March 2016 wave of RAND Corp.’s Presidential Election Panel Survey (PEPS). The left-hand graph shows that the relationship between attitudes about blacks and support for Clinton was stronger when she was matched up against Trump than it was when she was matched up against Ted Cruz and Marco Rubio.  Against Clinton, Trump performed considerably better than Rubio and Cruz did among whites who rated African-Americans most unfavorably and slightly worse among whites who rate blacks favorably.

The right-hand graph shows that the same finding emerges even after accounting for party identification, ideological identification, and other demographic characteristics.

The second set of graphs below shows the same thing, but using two different measures of racial attitudes: racially resentful beliefs that race-based inequality is due to cultural deficiencies in African American communities and ethnocentric preferences for whites over such minority groups as blacks, Hispanics, Asians and Muslims.

Both racial resentment and ethnocentrism had noticeably stronger negative effects on support for Clinton against Trump than they did in trial heats with Cruz and Rubio.

In short, it appears that racial and ethnocentric attitudes will matter more in November than they would have if Donald Trump was not the GOP nominee.

Interestingly, however, ideology will likely matter less. Where people placed themselves on the liberal-conservative spectrum was more weakly associated with voter preferences in a Clinton vs. Trump match-up than it was in a match-up between Clinton and Cruz or Rubio. That supports the finding that Trump performed best among ideologically inconsistent Republicans.

Of course, these patterns could change over the next five months. But early results suggest that Trump’s general election candidacy alters the ingredients of some Americans’ vote decisions, making the 2016 election more about race and ethnicity and less about ideology than it would have been with Trump as the nominee.

Michael Tesler is assistant professor of political science at UC Irvine and author of “Post-Racial or Most-Racial? Race and Politics in the Obama Era.