Politicians are often portrayed as craven opportunists, flip-floppers who bend to public opinion without any real core beliefs to call their own. The idea there is understandable. For people to get elected, they have to win over the median voter, the one smack dab in the middle of the district's political spectrum. The best way to win that voter over, presumably, is to share their policy preferences. So switching your own preferences to be closer to theirs' — or as close as possible without alienating the median primary voter — makes sense.
But the tricky part is that voters' beliefs themselves are constantly changing — often in response to what politicians believe. Political scientists have known for a long time that political elites have a powerful effect on public opinion, but experimental findings on the topic have been hard to come by. So UC - Berkeley's David Broockman and Washington University in St. Louis' Daniel Butler set out to change that by enlisting state legislators in an experiment. They had the legislators randomly alert their constituents of opinions that the constituents likely didn't share, and then watched to see how the constituents' views changed.
How the study worked
Broockman and Butler are both veterans at running this kind of study; they previously ran an experiment together showing that white state legislators of either party are less likely to respond to correspondence from constituents with black-sounding names, and minority legislators are more likely to do so. But their latest work is trickier, in that it actually required the cooperation of state legislators to work. Moreover, it required them to not only agree to alert constituents of points of disagreement — a potentially dangerous move for people who need to be reelected — but to have the content of those alerts varied at random, by the researchers.
"I don't think it's that surprising" that they cooperated, Broockman says. "I think most people who decide to enter public service do it because they really care about the issues. This was an opportunity for people who passionately believe in causes like raising the minimum wage or making lives better for undocumented immigrants to understand if and how they can build public support for those causes."
In total, they got eight Democratic state legislators from "a Midwestern state" on board. First, the legislators were asked to name five to ten policy initiatives they were supporting and working on passing. Then, Broockman and Butler polled their constituents to find the four initiatives that were least popular for each legislator's district. They identified constituents who disagreed on at least one of the four; only 5 percent of respondents agreed with their representatives on all four, so this was the vast majority of the constituents surveyed.
Broockman and Butler conducted the study in two parts. First, they had only one legislator send a letter to a random group of constituents stating his views on one or two issues where he and the constituent in question disagreed; a control group of his constituents didn't get a letter at all. The four issues in question here were (a) should mining be regulated at a state or local level (b) should the government fund school vouchers (c) should the state income tax be cut and (d) should school districts be allowed to raise property taxes.
Then they expanded the study to the other seven legislators, with some alterations. This time, instead of not getting a letter, the control group got a letter from their legislator that didn't mention issue positions at all. Additionally, the content of the policy letters was varied, with some constituents getting brief ones merely stating their legislator's position, and other getting a long letter arguing for the position in detail. The issues here were more varied, including a minimum wage increase, vouchers, the introduction of a state-sponsored pension plan, and a plan for non-partisan redistricting.
For both waves of the study, the constituents were surveyed again after receiving (or not) their letters.
How voter opinions changed
For the first study, with one legislator, Broockman and Butler find that the letters "significantly moved his constituents' opinions to be more in line with his policy positions." Voters who disagreed with the legislator but got a letter were 6.5 percent more likely to agree with him in the follow-up survey. For the second, bigger study, voters getting a letter laying out their legislator's disagreements were about 5 percent more likely to agree in the follow-up than people who got a letter without issue positions stated.
There's reason to believe that most respondents read the letters they got. In the first study, over 50 percent of respondents getting a letter remembered getting it in a survey performed after the follow-up, compared to just 20 percent of those not getting the letter who erroneously reported receiving one. in the second study, over 60 percent of respondents said they remembered receiving a letter, and voters who got policy letters were likelier to correctly identify their legislators' positions.
They further found that opinion change was no more likely when an extensive argument was included in the letter; the legislators aren't persuading people with reason and evidence, but with the bare fact that they're the ones holding the positions in question. And legislators didn't suffer a loss in support from constituents they didn't convince: "citizens who received letters from their legislators taking positions they had disagreed with previously evaluated their legislators no less favorably." And while responses to the letters varied for different issue areas, they didn't differ so much that the results were "driven by a small set of atypical issues."
Broockman and Butler point out that the effect they found should, if anything, be stronger in real-life situations. Few people know who their state legislator is, or have much of an idea of what their opinions are on anything, and the study only provided a single letter, whereas for higher offices voters get "dozens of reminders about where well-known and potentially well-liked political leaders stand on issues." For example, the study would predict that Republican Senator Rob Portman's conversion to supporting same-sex marriage increased support for marriage equality among Ohioans, but the fact that Portman's conversion was huge news that many voters likely heard multiple times likely magnified the effect's size.
There's a lot more to explore here. The study can't tell us how much bigger the effect is when, say, the legislator and constituent are members of the same party, and we don't know whether the phenomenon revealed here overwhelms the opposite effect, of voter preferences on politicians' stances on issues. But the very fact that politicians have a powerful effect on voter opinions is troubling for many conceptions of democracy. The paper ends by quoting the Oxford political theorist David Miller defining democracy as "the aggregation of independently-formed preferences." "Our studies," Broockman and Butler conclude, "provide a rare window into democracy functioning in precisely the opposite manner: distributing issue positions taken by politicians to citizens."