Incumbency and GCB overperformance

[DataGuy]

I was reading a thread recently where someone remarked that the party in power is usually underestimated in the generic ballot. I looked into whether this was true and I have posted the data below from 2002-2016.

I found that if "party in power" is defined as the presidential party, the presidential party was underestimated on average by 0.8% over the eight congressional elections in this time frame. But in three of those eight elections, the presidential party was actually overestimated, so I would be hesitant to call it a rule.

The correlation is slightly stronger if we count only the four midterm elections. In that case, the presidential party was underestimated on average by 1.5%. The one exception was 2014, when the Democrats actually did 3.3% worse than RCP's average.

Interestingly, the strongest pattern occurred when "party in power" is defined not as the presidential party, but rather as the party that controls the House. Between 2002-2016, the GCB underestimated the party that controlled the House at the time of the election by an average of 2.1%. The only exception was 2012, meaning that this "rule" has held true in 7 of the 8 last congressional elections, or 88% of the time. Furthermore, it held true in all four midterm elections during that time frame.

Overall, the trend seems to be that polls typically exaggerate the size of "wave elections." Right now, Republicans control the White House. And perhaps more importantly for congressional elections, they control the House of Representatives, which historically makes them a clear favorite to overperform the polls.

Democrats lead the generic ballot by 7.6% right now. So if we apply the historical average of the incumbent House party overperforming by 2.1%, the Democrats would win the NPV by only 5.5%. Coincidentally, that is the exact point at which 538 would consider the House a 50-50 tossup.

The past doesn't always predict the future, but I think this is something to think about.


2002: Republicans control WH & House

RCP average: R +1.7
Result: R +4.8

Outcome v. polls: R +3.1


2004: Republicans control WH & House

RCP average: R +0.0
Result: R +2.6

Outcome v. polls: R +2.6


2006: Republicans control WH & House

RCP average: D +11.5
Result: D +8.0

Outcome v. polls: R +3.5


2008: Republicans control WH, Democrats control House

RCP average: D +9.0
Result: D +10.6

Outcome v. polls: D +1.6


2010: Democrats control WH & House

RCP average: R +9.4
Result: R +6.8

Outcome v. polls: D +2.6


2012: Democrats control WH, Republicans control House

RCP average: R +0.2
Result: D +1.2

Outcome v. polls: D +1.4


2014: Democrats control WH, Republicans control House

RCP average: R +2.4
Result: R +5.7

Outcome v. polls: R +3.3


2016: Democrats control WH, Republicans control House

RCP average: D +0.6
Result: R +1.1

Outcome v. polls: R +1.7

[DataGuy]

Good analysis! I tend to agree that the theory (hope on the Dem side) that undecideds will break towards Dems (on the grounds that Republicans are the incumbents) is probably exaggerated. That may happen, but it may also not happen. As you outline well, there are examples in recent history of both of those things happening.

Relative to the presidential party, the polls have broken in different directions in different years. So I really don't see too clear of a correlation between a party having the presidency and overperforming the GCB. On the other hand, I do see a pattern of the majority House party generally doing better in congressional elections than predicted by the polls, regardless of who's president.

Like I mentioned, that pattern has held true in 7 of 8 congressional elections since 2002 (including presidential years), and in 4 of the 4 midterms since that time.

[DataGuy]

Good analysis! I tend to agree that the theory (hope on the Dem side) that undecideds will break towards Dems (on the grounds that Republicans are the incumbents) is probably exaggerated. That may happen, but it may also not happen. As you outline well, there are examples in recent history of both of those things happening.

Relative to the presidential party, the polls have broken in different directions in different years. So I really don't see too clear of a correlation between a party having the presidency and overperforming the GCB. On the other hand, I do see a pattern of the majority House party generally doing better in congressional elections than predicted by the polls, regardless of who's president.

Like I mentioned, that pattern has held true in 7 of 8 congressional elections since 2002 (including presidential years), and in 4 of the 4 midterms since that time.

Yes, this tends to happen because of fundraising. Incumbents in the house tend to have more cash on hand and will generally over perform expectations. However, if you look at the fundraising this cycle for house candidates that incumbent advantage has disappeared.he

It's true that Democrats have an unprecedented fundraising advantage this year, but I would be careful about extrapolating that to actual votes. Money just doesn't have as clear a correlation with victory as it used to. For some anecdotal examples, Trump was vastly outspent by both his GOP primary opponents and Clinton yet still won. This year, Beto O'Rourke has been a fundraising machine but is still trailing.

More generally, Democrats have raised 65% of total donations this year in competitive House districts. That's definitely good. In 2006, they took the House when that metric was at 51%. But on the flip side, Democrats were soundly defeated in 2010 and 2014 despite raising 54% and 56%, respectively, of donations in competitive districts. In other words, they lost even though they had better fundraising than 2006 in the districts that mattered most.

Of course, this year's numbers do point to a marked improvement for Democrats, and based on other data points they are favored (although not 100% guaranteed) to take the House anyway. But since fundraising specifically has had a poor and often contradictory track record in predicting recent election cycles, I don't put too much stock in it compared to other more reliable metrics.