Legislators are far less likely to skip party-line votes
Recently, I wrote that Utah’s legislators skip a LOT of votes. Today, let’s ask why. There are two ways to come at this. First, we can ask which votes get skipped the most. Second, we can ask which legislators skip the most votes. Today, I’ll address the first question.
I collected data on every vote held in the Utah House or Senate during the 2007 through 2011 general sessions. I ignore votes held in committee, focusing only on floor votes. I also ignore votes on motions, including motions to amend. I’m looking only at votes on complete bills. In total, I have data on 7,473 roll call votes that were held in the Utah legislature during these years. Of those, 4,317 were held in the Senate, while 3,156 were held in the House.1
For each vote, I calculated the “absentee rate,” which is the percent of legislators who missed the vote. I use percents to make this comparable across chambers. If 8 of 75 Representatives (10.7%) miss a vote in the House, that’s comparable to 3 of 29 Senators (10.3%) missing a vote in the Senate.
I analyzed the data using a technique that allows me to isolate the effect of various conditions. Here’s what I found.
The biggest movers and shakers
The biggest influence on absenteeism appears to be controversy. The legislature passes many bills by overwhelming majorities, with both Republicans and Democrats voting together. However, 9.0% of votes are party-line votes, meaning that most Republicans vote one way and most Democrats vote the other. Legislators are far less likely to skip party-line votes; other things being equal, the absentee rate is 6.4 percentage points lower. This effect is especially strong in the Senate, with the absentee rate falling by 15.1 percentage points in party-line votes.
As another manifestation of controversy, the vote margin influences absenteeism. Legislators show up for close votes, but they skip lopsided votes. If the winning side has only 1 vote more than the losing side–a margin of 1 vote–expect absenteeism to be 0.13 percentage points lower than if the margin were 2 votes. Now, 0.13 may sound small, but it can add up. If the vote were decided by a 1 vote margin instead of a 11 vote margin, for example, then expect absenteeism to fall by around 1.3 percentage points.
It turns out that a bill’s chamber of origin also matters. Every bill must pass through both chambers to become a law, but the bill can start its journey in either the House or the Senate. If it starts in the House, it will have higher absenteeism in the Senate than in the House–and vice versa. Absenteeism is 2.4 percentage points lower when voting on bills that originated in the chamber holding the vote (e.g. when the Senate votes on a Senate bill). This effect is especially strong in the Senate. Senators are far more likely to skip votes on House bills (absenteeism rises by 2.8 percentage points) than Representatives are to skip votes on Senate bills (absenteeism rises by 0.6 percentage points). I’d be curious to hear what members of the House have to say about this finding.
Leadership bills also have less absenteeism. If the bill being voted on is sponsored (or floor sponsored) by a member of the chamber’s majority leadership, absenteeism falls by around 0.5 percentage points. This is especially true in the House, where absenteeism falls by around 0.9 percentage points on leadership bills. In the Senate, the effect appears to be too close to zero to be meaningful; apparently Senators are more willing to offend their leaders than Representatives are.
A hugely important caveat
These are probabilistic results, not deterministic results. It appears to be true, on average, that absenteeism falls by roughly 6.4 percentage points on party line votes. That’s a probabilistic interpretation. It is not true, though, that absenteeism always falls by 6.4 percentage points on party line votes. That would be a deterministic interpretation.
The statistical technique I am using to produce these results returns probabilistic estimates that tell you roughly what to expect on average. There will, of course, be many anomalous cases. Just because the average white American woman is 5’5″ tall doesn’t mean that there aren’t white American women who are shorter than 4′ or taller than 7′. It happens.
I included many variables in my analysis. This table lists most of them.
|If this happens…||The predicted absentee rate changes by this much (other things being equal)…|
|Chamber of origin. The bill originated in the chamber holding the vote.||Rises by 2.4 percentage points|
|Party line vote. A majority of Republicans vote against a majority of Democrats.||Drops by 6.4 percentage points|
|Leadership bill. The bill is sponsored2 by a member of the chamber’s majority leadership||Drops by 0.5 percentage points (no effect if sponsored by minority leadership)|
|Repeated votes. The bill has already been voted on once before (in either chamber)||Drops by 1.4 percentage points for each previous vote. If a bill is on its third vote, absenteeism drops by 4.3 percentage points.|
|Vote margin. Margin is the number of “ayes” minus the number of “nays.”||Drops by 0.13 percentage points with each increase in margin.|
|Failed votes. There are more “nay” votes than “yea” votes.||Drops by 2.8 percentage points|
|Day of vote. The vote is held later in the session.||Rises by 0.019 percentage points each day. After 45 days, that’s a 0.8 percentage point rise.|
|Vote held in House, not in Senate.||Rises by 1.3 percentage points|
|Concurrence vote. One chamber is voting whether to “concur” with changes made to a bill by the other chamber.
||Rises by 2.7 percentage points.|
I reported yesterday that absenteeism is more frequent in the Senate than in the House, so you may be wondering why the table above shows the opposite. The statistical technique estimates the effect of each variable when all the other ones are accounted for. Apparently, the higher absenteeism in the Senate can be explained by other variables listed here, such as vote margins, party line votes, and so on. Once those things are taken into account, it appears that absenteeism is actually more common in the House.
What this really implies is that it’s easier to predict absenteeism in the Senate using the variables listed here, while it’s harder to predict absenteeism in the House. (Warning: Statistical mumbo jumbo ahead.) When I run the model using only Senate votes, the model explains a respectable 29% of the variance (that’s R-squared); when I run it using only House votes, the model explains only 9% of the variance. (Okay, no more statistical mumbo jumbo.)
Okay. We see that we can predict that absenteeism will be higher on certain types of votes. But there are two questions we want to ask. First, we want to ask which votes get skipped the most. We just did that. But second, we want to ask which legislators skip the most votes. I’ll address that next time. Stay tuned.
Update: See “Which legislators miss the most votes?” for a list of names.