You are using an outdated browser.
Please upgrade your browser
and improve your visit to our site.
Skip Navigation

Are Democrats Too Centrist?

Political scientist Alan Abramowitz has done a fascinating pair of studies of candidate ideology and Senate races. The first produces a somewhat intuitive finding: all things being equal, a more conservative Republican, based on voting record, is more likely to lose a Senate race than a more moderate win. In other words, moving further away from the center carries a political cost. (Obviously, the more conservative the state, the more conservative the voting record a Republican can safely hold.)

So far this makes perfect sense. But in a second study, Abramowitz finds the opposite dynamic for Democrats does not hold: a more liberal voting record does not reduce a Democrat's chance of winning a Senate election:

The findings in Table 1 once again show that among Republican incumbents, conservatism had a substantial and statistically significant negative influence on vote share after controlling for the strength of the Republican Party in each incumbent's state and the national political climate at the time of each election. In contrast, among Democratic incumbents, liberalism had only a small and statistically insignificant influence on vote share.

It's a pretty strange and fascinating finding. My best guess to explain it would be that, since 2000 (when the data set begins), Republicans have been much more aggressive about forcing their Senators to toe the line on unpopular positions than Democrats have. Democrats have less party coherence and have been far more reluctant to take risks to advance the party agenda than have Republicans. In other words, Republican "conservatism" has meant embracing some really unpopular ideas -- like opposing the minimum wage, health care for children or endorsing tax cuts for the rich -- while the Democrats have steered clear of unpopular liberal proposals.

But it's just a hypothesis. I'd like to hear more about why this could be happening.