The Political Power Of The Pew

A new study shows how churchgoing affects voting preferences

By Robert J. Barro

Church attendance is a powerful predictor of voting behavior in U.S. Presidential elections. Exit polls from the 2004 election show that, among voters for the two main candidates, 37% of those with zero church attendance voted for George W. Bush, vs. 61% for those who attended at least weekly. Income also has explanatory power, but not as much. So it is important to consider religion when analyzing politics.

A forthcoming article in the Quarterly Journal of Economics -- "Strategic Extremism: Why Republicans and Democrats Divide on Religious Values," by Edward L. Glaeser, Giacomo A.M. Ponzetto, and Jesse M. Shapiro -- develops a model to explain why religion and politics are so intertwined. In the model, politicians sometimes cater to extreme positions, such as the ardently pro-life views of the Religious Right or the ardently pro-choice views of the secular left. A successful appeal yields a large response by the targeted group in voter turnout or campaign contributions. This part is straightforward. The new idea is that a successful appeal has to be somewhat private. Otherwise, catering to an extreme -- say, pro-life -- has the downside of encouraging too much voter turnout and campaign contributions from the opposite pole -- pro-choice.

THE GLAESER ET AL. STUDY analyzes which groups end up with sizable political influence. The membership cannot be too small because then any perceived catering to the group loses too many votes from the bulk of the population relative to the small number gained. But the membership cannot be too large, because then targeted messages are impossible. The research shows that the most effective groups comprise a little less than half the population. The membership also has to be cohesive enough to facilitate private communication. U.S. churches fit with both characteristics. U.S. labor unions fit once upon a time, as well, but have since become too small.

The study applies the theory internationally by examining how monthly attendance at formal religious services predicts self-described right-wing orientation. The data show that more religious people are more likely to be right-wing. However, the link between religiousness and political outlook is weak when countries have very low or very high religious participation. For instance, whether in Scandinavia and Russia, where few people attend church, or highly religious nations like the Philippines and Bangladesh, an individual's attendance predicts little about political orientation. Instead, religiousness predicts the most about politics in countries where roughly half the population attends formal religious services at least monthly -- places such as the U.S., Turkey, India, and Argentina.

The study also looks at voting patterns in U.S. Presidential elections from 1972 to 2000. For any given gender, age, race, and income level, a person who attends church at least monthly is 10 percentage points more likely to vote Republican. But this relationship has changed over time. Because Democratic candidate Jimmy Carter in 1976 and 1980 was perceived as strongly religious, churchgoing had only a small impact on Republican support. The religiousness differential in favor of the GOP peaked in 1992 and 1996 at 17 and 14 percentage points, when Democratic candidate Bill Clinton appeared to be highly secular (but still won). In 2000, when the evangelical George W. Bush beat Al Gore, the effect was still a strong 12 percentage points. Full data for 2004 are not yet available, but the religion effect was likely larger than the one in 2000.

The impact of religiousness on U.S. voting patterns varies by state. In the most religious, such as Mississippi and the Carolinas, where over 60% of the populace attends church monthly, individual church attendance predicts little. As in highly religious countries, it isn't feasible to send private messages to these large constituencies. The strongest link between churchgoing and voting behavior is in the least religious states, such as Oregon and Alaska, where 35% to 40% of the population attends church at least monthly. In these places, churchgoers are numerous enough to be worth attention but not too plentiful to preclude targeted messages.

One might have thought that the Internet's free flow of information would crimp the ability to keep messages private. The reality is that evangelicals do not want to read Web sites aimed at atheists, and vice versa. So, the political cleavages based on religious differences are likely to be a permanent feature of American and international politics.

Robert J. Barro is a professor of economics at Harvard University and a senior fellow of the Hoover Institution (

    Before it's here, it's on the Bloomberg Terminal.