Post by eric on Aug 8, 2018 21:11:34 GMT
Incumbent U.S. politicians enjoy enormous rates of re-election, routinely in the 90%s. I wondered, which direction does the causation go? By analogy, were the 2018 Warriors good because they won the Larry O'Brien Trophy in 2017, or did they win in 2017 because they were good in the first place and the two years were (largely) continuous? I think we can agree it's mostly the latter, but it's plausible that having that championship cachet is also helpful in many ways, from free agent recruitment to locker room management to financial incentives like ticket sales and endorsements.
So I looked at the Congressional Representatives who were not in office in 2008 and were in 2010, then saw how they performed in 2012 so long as they were in the same district and neither margin of victory exceeded the Cook Partisan Voting Index by more than 50 points. The first criterion I think passes muster on its face, but for the second it turns out that many Congressional elections are literally or effectively unopposed, and I don't think those are relevant to our investigation, especially if only one of the two was unopposed.
This left us with 85 elections. (As an aside the incumbents won 75 of them, good for 88%.) Here's the graph:
What this is telling us is that even if a Representative doesn't outperform their PVI at all in their maiden election, on average they gain about 5 points in their second. We also see that even massive outperformance tends to come back towards even, although it's still a stronger position than no outperformance at all.
If we restrict the sample such that neither margin exceeds PVI by 30, we get a lower R^2 (less correlation) but a similar function of .27 * margin + 5.6.
It's hard to say what specific phenomenon this is measuring. My guess is that it's as simple as track record - most Representatives will be slightly below average or better, and people prefer even the low level devil they know over a potential catastrophe.
Also worth considering is that some of these new Representatives defeated the previous incumbent (as opposed to winning an open seat), which would mean we are double counting the incumbency advantage. As we've seen though incumbents are really hard to beat and even if half of the 85 came to Congress over an incumbent it would only drop the flat incumbent advantage to about 4 points.
So I looked at the Congressional Representatives who were not in office in 2008 and were in 2010, then saw how they performed in 2012 so long as they were in the same district and neither margin of victory exceeded the Cook Partisan Voting Index by more than 50 points. The first criterion I think passes muster on its face, but for the second it turns out that many Congressional elections are literally or effectively unopposed, and I don't think those are relevant to our investigation, especially if only one of the two was unopposed.
This left us with 85 elections. (As an aside the incumbents won 75 of them, good for 88%.) Here's the graph:
What this is telling us is that even if a Representative doesn't outperform their PVI at all in their maiden election, on average they gain about 5 points in their second. We also see that even massive outperformance tends to come back towards even, although it's still a stronger position than no outperformance at all.
If we restrict the sample such that neither margin exceeds PVI by 30, we get a lower R^2 (less correlation) but a similar function of .27 * margin + 5.6.
It's hard to say what specific phenomenon this is measuring. My guess is that it's as simple as track record - most Representatives will be slightly below average or better, and people prefer even the low level devil they know over a potential catastrophe.
Also worth considering is that some of these new Representatives defeated the previous incumbent (as opposed to winning an open seat), which would mean we are double counting the incumbency advantage. As we've seen though incumbents are really hard to beat and even if half of the 85 came to Congress over an incumbent it would only drop the flat incumbent advantage to about 4 points.