Friday, October 29, 2010

Today On "Not My Job"

Come November 3rd, there's going to be a lot of declarations about what expected Republican gains in the House and Senate "mean." What is this election "about?"--it's a pretty stupid question, but not one that I am above engaging in. Of course elections are about many things to many people, and within the pundit class, asking it is simply a way of defining prestige. It is essentially a game of who will be the first to voice what becomes the consensus position.

On a more serious note, the election will certainly have concrete consequences for the American political agenda. But what matters most? Control of the House or the number of Republican seats picked up in the Senate? Does the party of the speaker or the number of seats that the majority controls matter more? To start answering these questions, I did the simplest possible test- is the number of major legislative accomplishments correlated with the President's party controlling more seats in the House and Senate. If we believe that "elections matter," we would expect to see a fairly strong correlation. Note: I am not a political scientist, hence the title of this post, but I do own a copy of Microsoft Excel and have statistical training.

The data represented in these graphs is the number of "Major Legislative Accomplishments" listed on Wikipedia for each of the 86th-111th Congresses, i.e. the ones in which we have 100 senators. Over the course of this period, the size of the House has held constant at 435 members with floor votes, with the exception of the 86th Congress, which included an extra two voting members to accomodate Hawaii and Alaska joining the Union. Wikipedia's number of legislative accomplishments is a useful tool here, as it suggests agreement among an informed public about what constitutes a major legislative moment. A more ideal model might involve asking a number of political scientists to quantify exactly how successful each Congress has been, but I can't think of a good reason that those numbers would substantially differ from the raw count of Wikipedia-listed accomplishments.

The number seats belonging to the President's party is directly correlated with legislative accomplishments in both chambers. This matches our intuition that elections have consequences even without reference to the fulfillment of liberal or conservative policy goals. In essence, if a voter wants more things to get accomplished, the rational thing to do is to vote for members the President's party.

The little "r =" numbers tell us how well correlated the numbers of legislative accomplishments is to the number of seats in each house. The Pearson's r for the Senate (.181) tells us that there may be a weak correlation, but the truth is that the relationship isn't statistically significant. Partisan possession of the Senate just doesn't have that big of an effect on how much Congress gets done. The better predictor is partisan make-up of the House. Pearson's r is .505 for the House, which is quite strong in social science terms. The probability of seeing this relationship purely by chance is about 1-in-120, so we can be pretty sure that the relationship is real.

The next question that presents itself is "is the relationship meaningful?" House partisan make-up may not determine the number of legislative accomplishments; they might be results of the same phenomenon, such as Presidential popularity or the economic situation. I would simply be delighted if someone would point me to (or perform) some research on the Presidential popularity possibility. However, the actual make-up of the Congress probably has better a priori explanatory power for what Congress accomplishes than an external consideration like Presidential popularity. As to an economic explanation, we expect high anti-incumbency voting during economic downturns, but the political need for Congressional action increases during these periods. The logic cuts the other way.

My regression model tells us that if Democrats lose about 53 seats in the House (which Nate Silver predicts is the median election night outcome), we could expect a still-above median 14.6 major legislative accomplishments in the 112th Congress. The median number of major legislative accomplishments for a Congress since 1958 is 12.

Is that a meaningful prediction? Probably not, and it certainly doesn't tell us anything about Congressional popularity, President Obama's popularity, the state of the economy, or any number of important questions for the 2012 elections. But it will tell us whether Congress is under or over performing its expected workload in the next two years--relative to a pretty crude expectation. Telling you how crude this projection is? That's somebody else's job.

Update: With only 10 districts not called on the NYTimes electoral map, we can make a final prediction: assuming 243 Republicans in the House (i.e. 192 Democrats) and a 53-47 Senate in the Dems' favor, expect about 14 major pieces of legislation in the next two years.

No comments:

Post a Comment