In July there was a flurry of news stories about Donald Trump’s “Catholic problem” that became evident with the release of a national survey from the Pew Research Center conducted in June. Thankfully, this organization is one of the few in this election cycle to still show an interest in how religious affiliation, or lack thereof, affects vote intentions.
In that June survey, Hillary Clinton led Trump 56% to 39% among self-identified Catholics. The media, commentators, and politicians picked up and ran with the result as such:
- “Trump is faring poorly among Roman Catholics.” National Review
- “Why are so many Catholics down on Donald Trump?” Huffington Post
- “Catholic voters, who have been key to picking the winning ticket in almost every modern election, reject Trump decisively.” Religion News Service
- “Experts on American Catholics say Democrats have an opportunity to attract religious Catholic voters in a way they have not for decades.” National Journal
Now with the conventions over and the campaigns headed toward Labor Day it appears Clinton may have also caught a “Catholic problem?” Is this stuff contagious? Did her support among Catholic registered voters drop 16 percentage points in two months?
The August poll includes the Green and Libertarian candidates. Obviously the Catholic shares are sub-samples with fewer respondents and thus larger margins of error than the overall poll results for both surveys. Some of the volatility may just be artifacts of these issues and in the end may not be very reflective of what we would have seen if the election were held in June or August.
Here is the reality… A majority of Americans see these candidates as unfavorable (Clinton, Trump). The 2016 election is not about voting for a candidate as much as it is voting against one. Turnout will be key. Of course not all registered voters are going to show up at the polls. Yet registered voters are the frame for many election polls at the moment. There are some polls looking at likely voters but the accuracy of these depends on the quality of the model. I don’t know why anyone would be confident at this point about predicting the likely turnout of voters given the candidacies of Trump and Bernie Sanders. There is a “new” segment of the electorate out there that hasn’t been active in the past and there are likely pieces of the old electorate that won’t bother showing up or voting for the top of the ticket.
One other note of caution. In the recent years there has been something quirky going on in election polling here and abroad (1, 2). Don’t be surprised if the election results look significantly different (beyond margin of error) than what is predicted in the pre-election polls or in how the exit polls turn out. Whether it is low response rates, poor sampling, or social desirability effects (respondents feeling embarrassed to state their vote intentions) there is a ghost in the election polling machine and it is likely to be visible again on Election Day here in the United States. Unfortunately, people tend to take polls too literally and this may only stir conspiracy theories of a “rigged” or “hacked” election.
What I can say is that the overall vote is likely to go as the vote of Catholics does. By no means is the “Catholic vote” a block but it is a historically definitive swing vote. While there are typically big differences between non-Hispanic white Catholic voters and Hispanic Catholic voters this matters most in states that are not competitive (e.g., California, Texas). The Trump campaign’s “Rust Belt” strategy is in states where Catholics are disproportionately non-Hispanic white and tend to vote Republican (doing so in 2012).
As a pollster I always want to trust an aggregation of polls over any single study. With only Pew taking religion seriously this election cycle we can’t aggregate Catholic results for a clearer portrait. With nearly all of the exit polls for the primaries excluding a religious affiliation question the data just aren’t out there. This is remarkable given that the “God Gap” is likely to be one of the decisive factors for Election 2016.
RealClear Politics allows one to view what is happening for the overall electorate by aggregating polls. What is evident is that Hillary Clinton’s lead over Trump in August varies, on average, from about 3.4 percentage points to 8.4 percentage points depending on whether one is looking at likely voters or registered voters and whether it is a two-candidate choice or a four-candidate choice. One assumes that the result most reflective of a potential outcome is with likely voters choosing among four candidates. But then again, how good is the likely voter model being used? The figures below show the trends for the overall electorate as aggregated on RealClear Politics:
How useful are these trends? Not much. Instead what really matters is the population-weighted popular votes of each state in terms of Electoral College outcomes. In key swing states, Clinton holds sizeable and consistent leads over Trump. Don’t make too much out of individual polls which show leads counter to other surveys in a state. These can happen by chance or by flaw. They often lead to a “shock” headline in the paper but amount to little on Election Day.
Realistically, Trump needs to win the states Romney did in 2012 and then add Ohio, Florida, Iowa, and either Pennsylvania or Virginia. Looking at recent polls, given margin of error and differences in poll structure, Clinton and Trump are tied in Iowa. Clinton appears to have an edge in Ohio (+5) and Florida (+6). Clinton has big, perhaps insurmountable, leads in Virginia (+13) and Pennsylvania (+10). Of course there are also some Romney states that Trump is at risk of losing as well. All of the election prediction models have Clinton at about 80% likely to win at this point given these advantages at the state level. In some regard this election is quite “small.” It’s about the voters in just a handful of states. Catholics will be part of that story. Perhaps in the exit polls we will get a clearer picture of just what role they played. For now be wary of claims either candidate has a “Catholic problem.” We have too little data and what we do have presents a mixed picture.
Update (8/26): Proving my point on the volatile wackiness of the polls this year, PRRI also conducted a survey in early August (Aug. 10-16) and found Clinton leading Trump 55% to 32% among Catholic registered voters.
Note: If you are a regular reader of this blog you already know that I (CARA researcher Mark Gray) am a political scientists and pollster who is profoundly apolitical. CARA is also an independent non-partisan research center. I am not registered to vote nor will I be. I am neither a Democrat nor a Republican. In political analysis and forecasting I always try to stick solely to the data.