By Ryan P. Burge, Eastern Illinois University
One of the things that I have to impress upon my graduate students is there aren’t that many “laws” in social science. Unlike the physical sciences where the behavior of cells can be predicted with great precision, we really don’t have any findings that are always true in every circumstance.
However, there are some that come pretty close. One of those is the gender gap. It’s a well studied phenomenon in political science that concludes that women are more likely to identify and vote for Democrats than their male counterparts. There are a lot of possible reasons for this. It may be that women are more likely to side with the Democrats because of their less restrictive views on abortion services. Or it could be that women are more likely to find themselves in need of government welfare programs because of being single mothers or taking on the caretaker role in families. Another possibility is that women tend to seek out diplomatic or peaceful solutions to problems, when men may be more likely to resort to violence. This literature is deep and broad and well worth reading to capture the different perspectives on why this gap persists.
Nevertheless, we know that women are less likely to support Republicans, ceteris paribus. For instance, in the 2016 presidential election, Hillary Clinton did 13 percentage points better among women than men – an all time high. But are there places where the gap narrows? For instance, the data clearly indicates that women are more likely to attend church than men and we know that, by and large, religiosity in America is becoming more and more linked to Republican politics. Can this nudge highly religious women closer to the Republican side of the partisan spectrum? Does religion narrow or eliminate the gender gap?
Let’s begin with a broad look. I separated the Cooperative Congressional Election Study into 16 religious groups and then calculated the percentage of women and the percentage of men who identified as Democrats for every year available dating back to 2006. The results of that analysis are visualized below.
In short, the gender gap is clear and persistent for almost every religious group. Obviously there are major differences in the y-intercepts for some of the traditions, however. Note that the share of white evangelicals who identify as Democrats is well below 20%, while for Atheists it’s at least 50%. Yet, in some of the largest religious traditions in the United States (evangelicals, mainline Protestants, and black Protestants) there is an unmistakable gap between the two lines, with women consistently more likely to align with the Democrats. And this gap is remarkably consistent across the last decade – never does it seem to narrow in any systematic way. What is also striking is that the same gap also emerges among the religiously unaffiliated – women are typically ~10% more likely to identify with the Democrats. It does look like there has been a slight upturn in Democratic portion for atheist and agnostic women in the 2018 sample, however, that is only slightly evident among the male samples.
This deserved some more rigorous investigation. To do that I specified a simple logit regression model with identifying as a Democrat as the dependent variable using sixteen different religious groups as my subsamples.. The independent variable was being a female respondent and I also included controls for the age of the respondent, as well as their highest level of education, their household income, and their frequency of church attendance. The graph below is a coefficient plot of just one of those variables – gender – with the point representing the coefficient estimate and the lines representing the confidence intervals for each estimate for each religious group.
The interpretation here is very straightforward — every coefficient estimate is to the right of zero and is statistically significant. In simple terms it means that a female is more likely to identify as a Democrat than a male, holding constant my four control variables. That means a white evangelical woman is more likely to identify as a Democrat than a male white evangelical. The same is true for Mormons and Buddhists and atheists. The gender gap is basically universal across American religion.
However, there are some traditions where the gap is largerthan others. For instance, black Protestant women are much more likely to be a Democrat than a male Black Protestants. The religious group that seems to be the least affected by gender is followers of Islam. Another broad trend is that traditions that only included people of color (non-white evangelicals/non-white Catholics) evince a stronger gender gap than traditions that included only white respondents.
Because the previous model lumped together people who never attend church services with people who attend multiple times a week, I wanted to determine the effect of increased attendance on the gender gap. To do that I specified an interaction model of church attendance x gender x religious tradition. I also used the previously employed control variables: age, income, and education. The results of those interactions are visualized below. A downward sloping line indicates that higher levels of church attendance leads to lower levels of Democratic affiliation. The distance between the lines is the gender gap.
Speaking broadly, we see a lot of downward sloping lines. That indicates that increased religiosity leads to decreases in Democratic affiliation. However, what we are really looking for here is whether the gaps between the two lines change as one moves from lower to higher levels of church attendance. What we find is a pretty mixed bag.
For instance, it does look like the gender gap begins to narrow for both white evangelicals and white Catholics. Among those who attend multiple times a week the gap between men and women narrows to just about 5%. In other cases, however, this narrowing does not occur. For instance, the gender gap among mainline Protestants is basically the same at all levels of church attendance. Non-white traditions also see the same pattern – a persistent gender gap, with women being about 10% more likely to identify as Democrats at all levels of church attendance. Even among the religiously unaffiliated, the gender gap persists as people attend church more frequently.
There’s one more angle that I wanted to take on this – is the gender gap different for older Americans than younger ones? I did the same basic regression, except I substituted age for the x-axis with some of the same controls. The results are not necessarily what I expected.
For instance, for most white Christian traditions it seems that the gender gap is smaller among the youngest cohort (18 to 35 year olds), then it begins to widen as we look at older Americans. Note that a female white evangelical who is of retirement age is actually slightly more likely to be a Democrat than her younger evangelical counterpart. It may be that young white female evangelicals are merely adopting the political stances of their parents and they find their own political voice as they move further into adulthood. The pattern is reversed for white evangelical men – they become less likely to identify with the Democrats. That also holds for Mormons and white Catholics, as well. On the other hand, the religiously unaffiliated show a different pattern: a persistent gender gap at all age levels. That is also apparent among Jews, Muslims and Buddhists.
There are several takeaways for me. First, the gender gap is unbelievably persistent. In the case of the largest religious traditions in the U.S. (Protestants, Catholics, religiously unaffiliated) it has deviated very little in the last decade. Women, on average, are about 10% more likely to identify as Democrats than men. And this holds up in every regression model I tested. There is not a single religious tradition in which we can say that women are not more likely to identify as Democrats than men after controlling for education, income, age, and church attendance. There are times when the gap narrows, however. This is most notable among highly religious white evangelicals and Catholics. However, it most other cases, attendance doesn’t narrow the gap in a meaningful way. Finally, age seems to widen the gap for most white Christians. That’s because women are more likely to identify as Democrats among the oldest age group, but also because older men drift away.
In a recent report, the Center for American Women and Politics found that ten million more women voted in the 2016 election than men. If Democrats want to continue to make gains, they would do well to continue to target female constituents, even those who are affiliated with traditionally conservative religious traditions.
Ryan P. Burge teaches at Eastern Illinois University in Charleston, Illinois. He can be contacted via Twitter or his personal website. The syntax for this post can be found here.