Does Money Matter in House Elections?

Does Money Matter in House Elections?

This House midterm cycle has been notorious for its high price tag. The special election in Georgia’s sixth congressional district, which took place more than a year before election day 2018, has already broken campaign finance records — and the cycle isn’t even officially underway. But do donors who would have three dollar signs on Yelp actually hold sway over who gets a seat in Congress? Do challengers who out-raise their incumbent opponents actually perform better than challengers who don’t?

In this post, we will analyze campaign finance data and election returns from the 1970-2014 House elections to (1) compare recent fundraising benchmarks to past hauls and (2) determine whether money does have a significant effect on congressional races.

So let’s start with the obvious. How do Democrats compare to past cycles with the fundraising numbers they’re posting this time around?


Democrats are Beating Fundraising Expectations for the 2018 Midterms


First and foremost, Democratic challengers are outspending the Republican incumbents in their districts. According to data compiled by political science professor Dr. Gary Jacobson and DailyKos (and supplemented by official returns from the Federal Elections Commission), Democratic challengers are spending about 40% of the total campaign finance dollars in Congressional districts with Republican incumbents. This represents a notable departure from past years when Democrats were spending an average of just 24% of campaign cash. This trend is depicted below:

Generally, this should be encouraging to Democrats who are trying to unseat Republicans. Especially those in competitive districts where they have, if not a good shot, at least a better shot of pulling off an unseat than they do elsewhere. According to Roll Call, 30% of Democratic House challengers in competitive districts raised more money than their incumbent Republican opponents (note that we’re using money raised, not spent, here). This includes districts currently represented by some vulnerable members of Congress such as John Culberson (R, TX-05) and Mike Coffman (R, CO-06), but it also includes many seats rated Likely Republican that are much less likely to place a Democrat in office. Other districts where Democrats need the cash are noticably dry.

Nate Cohn of the New York Times Upshot noted this discrepancy a few weeks back. Cohn writes “Over all, there are 11 districts (out of the 50 districts that ought to be most competitive, by our estimates) where the Democrats don’t have a candidate who raised $100,000.” That is a bad sign for Democrats, Cohn concludes, because Democrats will need strong fundraisers to overcome the advantages their Republican opponents have in most of these districts.

It also may be helpful for us to look at the difference in spending between Democratic challengers and Republican challengers over the years. This allows us to evaluate the viability of non-incumbent candidates in places where their success would deliver an additional seat for their party. By this metric, too, Democrats are in good shape. In fact, they are in the best shape they have ever been in, currently outspending Republican challengers by more than 4-to-1. (Of course, these numbers and proportions will change before election day 2018 — which is 366 days from now!)

This being said, numbers are just numbers. What we really want to know is exactly how many challengers are spending more money than the incumbents in their districts. This year there are roughly 40 Democratic or Republican candidates outspending their competition. That’s a huge departure from recent history. Since 2000 the average number of incumbents outspent by candidates running to dethrone them was a measly 15. In recent years it has been even less. That makes this cycle’s performance by congressional hopefuls an extremely unlikely event. Their competitiveness is a more than 1-in-5,000 event — or, it will be, if the same spending ratios hold true through November of 2018.

With these numbers in mind, we also see the beginnings of a pattern. By the looks of the previous two graphics, the most notable “wave” elections occurred when one party’s challengers largely outspent the other party’s. That’s what happened in 2006 when Democrats nearly tripled Republican spending and in 2010 when Republicans spent $150 million more than candidates on the left.

But there are some notable exceptions to this theory; in 2014, for example, Republicans picked up thirteen seats in the House, despite spending just $9M more than Democrats. And in 2016, Republicans spent only a quarter of the amount that Democrats spent but lost only six seats. We do see suggestions of a pattern here, but the relationship is rather noisy and doesn’t hold if we exclude 2010 and 2006 from our analysis. We may think that Democratic spending for the 2018 midterm hints at a blue wave, but this is tricky to find in the data. If this were a game of Clue we’d still need to know where Colonel Mustard hid the candlestick — though other data (like presidential approval, the generic congressional ballot, and special election results) definitely suggest that he did it in the conservatory.

However, there’s an unanswered question staring us right in the face. Through our analysis thus far we have been forced to assume that campaign spending actually affects election outcomes. But..

What if the money spent by incumbents and challengers doesn’t actually matter?


What’s the Value of Campaign Spending?


Data for this section comes from political scientists Dr. Gary Jacobson at the University of California, San Diego and Carlos Algara at the University of California, Davis. Special thanks to Carlos for providing some much-needed background on the research around campaign spending.

Could it be that House elections have such a strong relationship with other factors — like the partisanship of a district, the incumbent’s previous vote share there, the different qualities and characteristics of the candidates, etc. — that campaign spending simply doesn’t matter? That an extra $100,000 spent by the challenger won’t really buy them an extra percentage point of the vote? To answer simply: probably. There are a few ways we can go about testing this hypothesis.

We can do this with results from all of the House elections from 1970 to 2014 using multiple linear regression. This statistical tool will allow us to predict the incumbent’s vote share in each district using their previous share of the vote there, their party’s share of the presidential vote in the district, a variable indicating the quality of the challenger candidate (basically whether they are viable or not), and finally, the difference between the incumbent’s and challenger’s spending in the race.

Before we proceed it should be noted that, while this approach does allow us to control for some of the known factors of elections that could create the appearance an effect of spending, we can never be sure whether this model is capturing the importance of campaign finance or something we have yet to consider. Further, it should be noted that the spending itself isn’t persuading more voters to cast ballots for the incumbent candidate, but rather, more money means more campaign staff to knock on doors, buy ads to mobilize candidates that would be nonvoters, etc. However, the evidence linking campaign spending to these standard campaign practices is very weak. Moreover, there is even weaker (perhaps nonexistent) evidence linking these practices to actual shifts among the electorate. See a recent paper by political scientists David Broockman and Joshua Kalla for more on this.

At any rate these notes are just to say that if we do perceive an effect of campaign spending, it’s possible that we could be observing the effects of some other intervening factor (though it’s unlikely the effect would be entirely attributable to something else), and if the effect of spending is real it may be unlikely to have tangible electoral consequences in both the aggregate and in reality.

In the end, we stumble upon a muddy picture. According to this initial model there is evidence that campaign spending does not influence the share of the vote incumbents receive from their constituents.

Even more intriguing we see that spending did matter in the past. At the height of the effect of spending in 1974, our model estimates that every $10,000 more an incumbent spent than their challenger, they would gain roughly 0.75% of the vote. However, the literature on campaign spending points to large diminishing returns of incumbent spending. This is no surprise; you would expect that the first $10,000 margin spent by an incumbent would have a pretty big effect in terms of buying an advertising lead, hiring more staff, mobilizing more volunteers, etc. than the difference between a $390,000 lead and a $400,000 lead. Presumably, voters (at least, the attentive ones) would notice an incumbent suddenly running ahead of their challenger. They wouldn’t notice a blowout election turning slightly more into a blowout election — and if they did they likely would not care. We can run the model again while taking this into account.

Viewing campaign spending within this paradigm yields similar results; we observe a large effect of spending (now in terms that account for a loss of importance as spending increases) that decreases sharply in the 1970s and seemingly disappears as we move into the 21st century.

There is yet a final way to evaluate our hypothesis with the data we have, and that is not by using the difference in dollars spent but by the difference in the percentage of expenditures that go to the incumbent and the challenger. This allows us to control for the very real fact that campaign spending has ballooned over time, with some House races costing donors tens of millions of dollars on either side. This could make it fairly hard for the influence of campaign spending to show up in our models.

As a reminder, all we’re doing here is running a linear model where the incumbent’s percent of the two-party vote in a district is equal to (1) their previous vote share, (2) their party’s share of the presidential vote in the district (3) a variable indicating the quality of the challenger candidate (basically whether they are viable or not), and finally, (4) the difference between the incumbent’s and challenger’s percent of spending in the contest.

Using this approach we do observe small effects of campaign spending; the investigation reveals a roughly 0.08% increase in the incumbent’s vote share for every 1 percentage point in the margins of spending between them and the challenger. In other words, if a challenger spends 40% of the money in a district and the incumbent spends 60%, the incumbent is estimated to get 1.6 points more of the vote than if spending was even. However, there is a wide confidence interval on the estimate, and the effect could be as low as 0.04% or as high as 0.12%.

Even still, the question remains of whether including campaign finance data in our predictive models will help gauge the performance of incumbents. Specifically, we want to know if the accuracy of our House elections predictions increases if we include that percentage point difference in spending in our prediction models (one akin to the models being run in my 2018 House midterms election forecast). There is bad news here for those who want to see that campaign finance data can enhance our predictions.

This comparison comes almost all the way to confirm our original hypothesis that campaign finance data doesn’t influence House elections. In fact, the reality of this final manipulation of spending data — an analysis of percentage point, not real, margins of expenditure — is that it may, in fact, decrease prediction accuracy.

Thus, not only does the difference between what an incumbent spends on a race and what their challenger spends (1) have little effect on the race (at most) but (2) attempting to account for that effect actually hurts analyses of contest dynamics.

There are some traditional explanations as to why spending may not matter. It could be simply because spending has run out of control, decreasing the effective power of the extra dollar. It could be that incumbents spend far more than challengers (think 85-90% of candidate dollars in 2016), all but preventing challengers from even raising enough money to show up in our data. It could be that the traditional explanations hold; that spending is simply a proxy for some other campaign tools (like voter mobilization or volunteer coordination) that hardly ever show up as mattering.

Finally — and this is my favorite hypothesis — it could be that money simply doesn’t shift the mostly sticky psychological tools that voters use in making decisions about voting. Research posits that no reasonable amount of spending would change the partisan, ideological, or emotional distribution of an electorate; e.g. that even if one candidate could spend enough money on an ad that makes some opposing partisans more likely to vote for them, the same ad would either turn some of their supporters away or be canceled out by the spending of the opposition campaign.

Though this investigation cuts slightly against the grain of the conventional wisdom — I find that spending does have at least some effect, even if essentially minute — the explanations espoused above and given by classic political science literature seem compatible with our findings. The same reasons why spending doesn’t matter can explain why spending would matter just a small amount.

And yet, this conventional wisdom may not even be on most politicos’ radars. Given the enormous fuss over recent finance reports (with headlines such as “Democrats’ early money haul stuns GOP” and “The early numbers don’t look good for California Republicans”) we may all be thinking that campaign finance does have a huge impact the outcome of House races. Well, now we know that that is certainly a stretch of the truth.

Of course, no one can argue sensibly that campaign spending doesn’t matter at all. Can you imagine a candidate winning an election who spent literally $0 dollars on their campaign? Indeed, spending does buy candidates something in their bids for office, but what exactly it buys them may be of limited importance to their chances of winning a seat in the U.S. House of Representatives.



Recent headlines about the 7-and-8-digit fundraising numbers posted by Democratic candidates prompted enthusiasm about their ability to flip important House races in 2018. However, these headlines have ignored past research into the amounts and effects of spending in House campaigns. Even more important, this research is either outdated or incomplete and the holes we have filled in paint an even different picture with the data.

Campaign spending likely matters little — if at all — and this fact should be enough to focus much more attention on the partisan dynamics of a contest, choices made by candidates and campaigners, and voices of voters than the dollar signs attached to numbers in campaign emails. Painting a target on the latter could do more than distract from the other important dynamics of a contest, as we’ve seen, it could even shift our estimates of candidate performance in the wrong direction.