US congressional candidates received over $1.7B in donations for the 2018 midterm elections. If you break that down by party, Democrats outraised Republicans - the tally was about $1B to $700M. Incumbents alone raised a whopping $800M. And this doesn’t even touch elections for the Senate, where a single candidate - Beto O’Rourke - took in over $80M. Any way you cut it, that is a lot of money.
Given those massive numbers, it is no surprise that there have been frequent, widespread concerns about the influence of of money in politics. It seems quite reasonable to assume that dollars have an outsized effect on our elections. And yet, as we are about to explore, that effect is surprisingly difficult to prove. To put it simply: Did all of that money make a difference at all?
Admittedly, the answer feels clear: Of course money matters. If not, why would candidates be so focused on fundraising around-the-clock? Campaigns require resources and resources mean dollars: Without fundraising, a campaign would fall flat. Field organizers and digital marketing teams are not free.
For such an intuitive answer, however, the numbers are surprisingly complex.
Do candidates that raise more, win more?
At first glance, the obvious answer looks to be obviously true. The below plot represents the 2018 congressional elections and compares how well candidates fundraised with how well they eventually performed. Specifically, each dot is a candidate, and the two axes reflect the following metrics:
Vote Share: The percentage of the vote received by that candidate in their district’s congressional election (excluding minor or third-party candidates).
Fundraising Share: The percentage of fundraising dollars received by that candidate in their district’s congressional election (excluding minor or third-party candidates). For example, if Candidate A received $40 in donations and Candidate B received $60 in donations, then Candidate A’s fundraising share would be 40% and Candidate B’s fundraising share would be 60%.
Ostensibly, if fundraising helps candidates win, then candidates who raised more money than their opponents should also wind up with more votes. Take a look for yourself to see if that holds true:
Note: Chart excludes uncontested races (defined as races without a Republican or Democrat candidate); A small number of candidates excluded due to lack of clean FEC data
This chart is entirely unsurprising. As expected, there is a neat diagonal line between candidates in the bottom left (raised less money, got less votes) and candidates in the top right (raised more money, got more votes), which aligns precisely with our intuitions.
Look more closely at the edges, however, and you might notice something a bit odd. There are a number of candidates who appear to have received practically no donations whatsoever (see the flurry of dots down by the “0%” mark at the bottom of the graph). Tom Hanson is an example of one of those dots - Tom was the Republican candidate in my hometown district (IL-5), and he seems to have gotten $0 in donations entirely. While it feels unlikely that Tom Hanson actually received no money whatsoever (odds are this is a quirk in how the FEC collects or records data), if any candidate were to lie on that bottom row, it makes sense that it would be him. The Illinois 5th district is in the heart of Chicago and is heavily, heavily blue. To put that another way: It’s hard to imagine many donations would be coming in Tom’s direction.
Why bother thinking about these heavily partisan districts? Mainly because they call into question the direction of causality of the relationship between raising more money and receiving more votes.
When we noticed that neat diagonal line in the chart, it was easy to assume that it was because a candidate raised more money that the candidate performed better in the election. However, what if partisanship is playing the primary role? In other words, is it really true that raising less money caused Tom Hanson to lose? Or was it running in a heavily blue district that caused Tom to both raise less money and lose in a landslide?
What happens when we take partisanship out of the equation?
In order to get a more accurate view of the true relationship between fundraising and electoral performance, we need to control for the partisanship of each district. While there is no magic number that provides a perfect measure, we can estimate partisanship by using 2016 presidential results.
In other words, let’s call a district “Democratic” for our purposes if Hillary won that district in 2016 - the more Hillary won by, the more “Democratic” we estimate the district to be.
As you would expect, there is a strong relationship between 2016 presidential results in a given district and the results of the 2018 midterms:
Note: Chart excludes districts with uncontested congressional races (defined as races without a Republican or Democrat candidate)
While there are certainly districts where Hillary lost in 2016 and Democrats won two years later (as well as vice versa), those battles happened on the margins. In the grand scheme of things, blue districts stayed blue, red districts stayed red, and swing districts had close elections in both years. In essence, district partisanship drove the bulk of the results.
Using the 2016 Presidential results as a guidepost, we can take a second pass at analyzing the relationship between fundraising and electoral performance. Specifically, we will “control” for the affect of partisanship by calculating the following metric:
- Adjusted Vote Share: For a Democrat, we can define Adjusted Vote Share to reflect the extent to which a candidate over- or under-performed Clinton in their given election. For a Republican, Adjusted Vote Share will reflect the extent to which a candidate over- or under-performed Trump.
Let’s continue to use the IL-5 as an example. In 2018, Mike Quigley (the Democrat opposite Tom Hanson) received 214K votes to Hanson’s 65K. Therefore, Mike Quigley had a “Vote Share” of ~77%. However, in 2016, Hillary received 75% of the vote compared to Trump’s 25% (excluding third-party). Therefore, we calculate Mike Quigley’s “Adjusted Vote Share” as 2%: 77% - 75%, the extent to which Quigley ‘outperformed’ Clinton. Similarly, Tom Hanson underperformed vs. Trump and wound up with a “Adjusted Vote Share” of -2%.
This is important because it allows us to set a fair baseline. A Democrat will clearly do well in the IL-5 election, whether or not they fundraise successfully. A more interesting question is whether raising more money will allow them to perform even better than expected (or, on the other hand, whether raising less money will lead them to underperform).
If raising more money really does cause a candidate to do better, then candidates who outraise their opponents should outperform expectations and have a higher “Adjusted Vote Share”. On the flip side, candidates who do not fundraise well should underperform expectations and have a lower “Adjusted Vote Share” . Let’s see whether that is the case:
Chart excludes uncontested races (defined as races without a Republican or Democrat candidate); A small number of candidates excluded due to lack of clean FEC data
After taking partisanship into account, that nice diagonal line disappears. There might be a slight tilt, but only if you squint - it is no longer nearly as clear that there is a genuine relationship between fundraising efficacy and campaign succes. In fact, there are many candidates who raised more money than their opponent (a “Fundraising Share” above 50%, in the top half of the graph) but who underperformed vs. partisan expectations (an “Adjusted Vote Share” below 0%, in the left side of the graph). Similarly, candidates frequently outperformed expectations despite raising fewer dollars.
What does all of this mean?
At the very least, we should admit the picture is not as clear as one might expect. It seems intensely odd that despite so much attention placed on campaign finance and - as a reminder - almost 2 billion dollars raised in the 2018 elections, we do not have surefire evidence that all this money actually makes a difference. I am far from the only person to have pointed this out - see, for example, this FiveThirtyEight article written in September 2018, which raises similar questions. When it comes to the impact of all that cash, the jury is still out.
Can we conclude with certainty that money does not matter? Of course not! There are all sorts of nuances worth considering. To list just a few of them:
The FiveThirtyEight article cites one Political Science paper that suggests fundraising matters more for incumbents than for challengers. Perhaps fundraising impacts different kinds of candidates differently.
One could imagine a “threshold” above which the additional dollar has negligible effect. Another title for this hypothesis could be “Money Matters - Up to a Point”. Perhaps the first $10M Beto raised was critical to his candidacy, but dollars $10-80M were effectively meaningless.
In the above exploration, we controlled for partisanship on the “votes” side but not the “dollars” side - maybe if a Democrat were to raise far more than the standard Democrat raises in their district, that would provide their candidacy a significant boost.
Perhaps it is not how much money raised that matters but the way in which those campaign dollars are spent. For example, suppose digital advertising is far more effective than TV ads (take this truly as an example - I have no idea whether this is true). Then, more money spent on digital should lead to more votes. However, more money spent on TV ads would not. If campaigns are universally spending their resources on ineffective strategies, that would mask the potential impact of dollars raised.
Each of the above may impact this analysis, and there are undoubtedly countless other possibilities. Personally, I still find it difficult to believe that money does not play a major role in elections. My intuitions there are strong. However, it also seems possible that, one way or another, a good chunk of that $1.7B spent on congressional elections had significantly less of an impact than we imagined.
P.S. This is a question I have been fascinated by since spending 6-months exploring electoral data and campaign finance data while working for a political start-up. If there are any other political data nerds interested in exploring the impact of political spending, let me know - I would love to hear your thoughts. Most of the data is public, if a bit messy and tough to wrangle.
P.P.S. Speaking of data, the data used in the above post comes from the MIT Election Data and Science Lab, the FEC, and DailyKos.