Do constituents influence the work of legislators?

 Image source: NYT

Image source: NYT

By Philip D. Waggoner

Decades of normative and empirical research on American representation has taken up the question of whether the represented actually influence the representatives, and thus whether the representatives are listening to the represented. Some suggest legislators operate as “delegates,” directly reflecting the desires of their constituents, while others expect legislators to act as “trustees,” aiming for more opaque alignment with constituents’ preferences. Either way, there is an expectation of constituents’ influence to be seen in their representatives’ work, as representatives are elected to be the voice of their constituents in a crowded, competitive government. But while constituents technically retain this power to influence, do they actually impact the work of their representatives? It turns out, the answer is not really.

I systematically take up this question in my recently published paper in American Politics Research, and add a few updates to this longstanding query. First, regarding the behavior of legislators, I look to a relatively underappreciated form of legislative behavior in the empirical study of Congress: bill sponsorship. This is a valuable place to begin from the elite side of the equation, given American legislators’ mostly uninhibited ability  to sponsor as many bills on any issues they wish. It follows that their policy priorities should be visible to some extent in their bill sponsorship behavior. Studying this form of behavior also has the added value of being mostly uninfluenced by party, in contrast to roll call voting or votes in committee, for example. And second, I offer a new way of measuring and mapping the policy preferences of constituents at the granular district level. Pairing these “calls” and “responses” across numerous issues, I am able to gain new leverage on the question of constituent representation. 

In measuring  constituents’ preferences, I take a slightly different approach and focus on specific issues. Other approaches tend to make inferences about constituents’ “ideologies” based on their responses to position-specific questions, such as “Do you favor or oppose legalization of marijuana?” Ideological preferences of constituents are then determined by aggregating  a battery of responses to similar questions. While valuable, there are numerous assumptions built in to such approaches, including whether or not ideology as a concept even exists (much less whether or not constituents possess it), and whether responses to inherently non-ideological questions (e.g., “favor or oppose”) actually reveal ideological information. With such approaches, the result is the researcher placing constituents in ideological space.

My approach  avoids these assumptions by allowing constituents to place themselves in policy space, based on aggregation and stratification of their responses to the “most important problem” question, which has been asked of numerous nationally representative samples over many decades. The product is a measure of constituents’ policy preferences on specific issues in relation to many other issues, resulting in a map of stated issue preferences, and assumes nothing of ideology.

Consider Figure 1, which includes maps of the distribution of average constituent preferences by state based on my measurement strategy across four issues considered as the “most important problem” facing the country in 2008: immigration (upper left), environment (upper right), gay marriage (lower left), and the economy (lower right).

Figure 1: Constituent Preferences based on Average Multilevel Regression with Poststratification (MRP) Estimates by State

Phil Waggoner - Figure 1.jpg

Note the variance in responses across geographic pockets in line with what we know about these regions. For example, regarding the economy (lower right map), 2008 is when the Great Recession began, and was felt most by the Midwest and Northeast (darker shades of gray). Or consider immigration (upper left map), which shows highest average rates of selecting immigration as “most important” along the Mexican border states.

Returning to the  question of constituent representation, my approach was twofold. First, I used rates of employment in every congressional district in specific industries as proxies for preferences of constituents. The idea here is that employment should reflect a prioritization of that same issue (e.g., farmers should prioritize agriculture over most other issues). In this first stage, I find mostly strong results, suggesting legislators are aware of the employment patterns in their districts, and they too make the same assumption of employment in a specific industry reflecting the preferences of their constituents.

The problem with this approach, though, is that employment patterns are mere proxies for preferences. Indeed, the assumption of employment reflecting preferences could be wrong. In the second stage of the analysis, then, I used the individual direct issue measures to see whether fluctuations in these measures correlate with fluctuations in legislators’ sponsorship on the same issues. In other words, does an increase in constituents citing the environment as the “most important problem” influence legislators to focus more of their bill sponsorship efforts on the environment?

Surprisingly, I find that the effects from the proxy tests in stage one disappear, suggesting legislators are not looking to the specific policy problems constituents highlight, at least insofar as their bill sponsorship fails to reflect these preferences. Rather, legislators appear to be most influenced by their committee assignments and employment patterns in their districts.

Stepping back, these results suggest  that legislators are mostly unconcerned with listening to the issue-specific preferences of their constituents. However, though based on the assumption that employment reflects preferences, legislators are at least attempting to reflect the interests of their districts with their sponsorship decisions. While not altogether ignoring their districts with their work, they are certainly not responding directly to the specific policy preferences of their constituencies either.

In sum, my study calls into question the expectations of the delegate model of representation, and finds that legislators act more as trustees. They are given power through election to office by constituents, but voting may be the extent of constituents’ direct influence.

Philip D. Waggoner is a Ph.D. candidate in the Department of Political Science at the University of Houston.