Some (Rough) Thoughts on the Value of Campaign Contributions

I have a number of non-EA friends who contribute to political campaigns, and have seen a few EAs online wonder (sort of in passing) whether campaign contributions are worthwhile. I started to wonder about this as well. However, I did a bit of digging around, and my initial conclusions are that your default attitude (as an EA-minded person) should probably be to not donate to political campaigns unless it’s an exceptionally good opportunity. The vast majority of the time, the marginal contribution of an individual donor is negligible.

A few caveats:

The discussion below pertains exclusively to federal elections in the US (elections for the House, the Senate, and the presidency). However, I expect many of the conclusions will have relevance for local and state elections, and maybe even for elections in other countries.
There are many ways you can influence politics with money. I’m not going to argue money plays no role in influencing politics, or that all possible ways of influencing politics with money are quite as low impact as campaign donations (for the explicit purpose of increasing the probability your candidate gets into office).
I figure many EAs already accepted some version of the conclusion I reach below. I don’t think very many EA donations, if any at all, are going to political campaigns. Still, there is something valuable in providing EA-style analyses of plausible interventions, if only to rule them out.

With all of that said, a basic argument in defense of campaign contributions might go like this:

US federal elections are dominated by the biggest spenders. Candidates spend in aggregate $4-6 billion dollars each election cycle, and the candidate who spends more almost always wins. Elected officials devote a substantial portion (between 17-18%) of their time to fundraising. It would seem, from all of the money being gathered and thrown around in federal elections, that the amount of money a candidate spends, and therefore the amount of money a candidate raises, is a significant factor in how successful their campaign will be.

Moreover, the outcomes of (some of) these elections are (probably) enormously consequential. It’s hard to precisely determine just how consequential, but I think it’s safe to say the following: (1) decisions made by elected officials in the US will have a large impact on current and future generations (2) in some cases, one candidate will have a much greater positive impact as an elected official than the other (in the counterfactual sense that the world in which one candidate is elected will be meaningfully better than the world in which the other candidate is elected).

This amounts to rough EV reasoning: (1) money can increase the chance candidate X is elected and (2) candidate X’s being elected will produce a lot of good. However, I think both of these premises are a little misleading. The first is easier to debunk than the second, so I’ll start there.

Does Money Actually Influence Election Outcomes?

The FEC has fairly strict regulations on reporting campaign contributions, so we fortunately have a lot of data on how much money candidates spend and how they spend it. Political science research in the past few decades has used this data to study the relationship between spending and electoral success. Yasmin Dawood summarizes the state of the literature in this comprehensive literature review from 2015:

Both incumbents and challengers put a great deal of effort into fundraising, which suggests that they think that raising money is required for winning elections. The academic literature, however, has not been able to conclusively establish a causal connection between incumbent spending and electoral success (Milyo 1999, Stratmann 2005). Incumbent reelection rates are at sky-high levels, and because incumbents outspend challengers by approximately a three-to-one ratio, one might think that campaign spending provides an electoral advantage. The puzzle, though, is that incumbent spending does not appear to be effective (Jacobson 1990, Abramowitz 1991, Milyo 1999, Jacobson 2003). Campaign spending by US House incumbents does not affect vote shares, although challenger spending does increase vote shares (Levitt 1994).

There are a few important bits to draw out from this quote. The first is that the literature seems to agree that it’s much more complicated than “spend X more dollars for Y more votes.” The second is that, historically, challenger spending has been viewed as much more effective than incumbent spending. And while this may be true, it seems unlikely to me that the difference is very significant.

In fact, the study cited above (Levitt 1994) actually came to the conclusion that challenger spending was less effective than previous studies indicated, and the reason is quite informative. Referencing work done by James Snyder, he writes: “Failure to control for candidate quality will lead to an upward bias in the estimation of the impact of challenger spending because high quality challengers will have a greater likelihood of winning and therefore will be able to raise a greater volume of campaign contributions.” In other words, higher quality challengers (i.e. challengers who are more charismatic, or propose better policies, or mesh more easily with the electorate, or whatever) tend to attract more donations. Explaining the correlation between spending and electoral success with the claim that spending more causes electoral success seems to be an almost textbook example of the correlation/causation fallacy.

This fallacy explains, in large part, why the academic literature has failed to find a causal relationship between spending and electoral success in general. Gary Jacobson writes:

Questions about how campaign spending matters are really questions about how campaigns matter, for although the point is often left implicit in the literature, the amount of money spent is just a handy surrogate for what is really expected to influence voters, the total campaign effect, its quality as well as quantity. To be sure, spending measures campaigning with considerable error, but there is no evidence that the error is systematic, and as long as we do not pretend to estimate effects with greater precision than warranted (“spend another seventy-five thousand dollars and you’ll get 1,572 more votes”), we are not likely to be led too far astray.

So while spending probably has some impact on election outcomes in most cases, it has much less of an impact than one might at first suspect. Spending is probably still a good measure of the size and popularity of a political campaign, so how much money a campaign spends generally gives you a good estimate of how likely that campaign is to be successful. Spending, however, is only one small component of how political campaigns are successful, along with candidate quality, the staff’s quality, the campaigning strategy, and many other factors. And, of course, if a campaign has no more money, that’s the end of the road for that campaign. But my guess is that the lack of funding is usually a symptom, rather than the cause, of the real problems with that campaign.

The most significant impact spending has is when a campaign is just starting out (especially for challengers, who need to get their name out there), but after a certain amount of money, additional funding starts to have diminishing returns.

The clearest conclusion we can draw from all of this is best provided by Jeffrey Milyo:

One important implication of recent empirical research on campaign finance is that the marginal value of a campaign contribution is quite small, both to the recipient and to the donor, because current law limits the size and source of contributions and because marginal campaign spending appears to have little effect on electoral outcomes.

How much of a positive impact could your politician have?

I think you can, in a sort of naïve and incomplete way, apply the STN framework to individual political campaigns. Although almost all of the time your contribution will make little-to-no difference, perhaps you could view campaigns as individual interventions, and assume they vary significantly in effectiveness, and figure that there are some campaigns (few-and-far-between) where your contribution may meaningfully increase the probability that a really good candidate is elected to office (and therefore be worthwhile from a marginal cost-effectiveness point-of-view):

(1) How knowledgeable and competent is the candidate and how consequential is the office they’re running for? What policies do they support?

(2) How effective are the campaigning strategies they are using (and how well is the campaign organized in general?)

(3) How early are you contributing in the electoral process and how much potential for funding does the candidate currently have?

(1) is scale/importance, (2) is tractability (how effectively will your contribution be used, how plausible is it that the candidate wins, etc.), and (3) is neglectedness/room-for-more-funding. The main difference between this framework and STN is that normally, an intervention only has to be unusually satisfactory along one axis of STN to qualify as “high impact”, whereas it seems like a campaign will need to perform well along all three of these axes to qualify as “high impact.” If the candidate does not seem especially competent, or the campaign is poorly organized, or the campaign is quite late in the election process and already has a lot of funding, making a donation would probably not be worth it.

So let’s accept that you’ve identified one of these rare campaigns, and (while we’ve seen earlier than contributions are not generally very effective) a contribution might be effective here. Why might you still be hesitant to contribute to the campaign?

Well, the first reason is that the candidate is just not particularly counterfactually effective. While I don’t have any data here to support this, I do have a strong feeling that many people vastly overrate the counterfactual impact of their favored candidate. My personal bias is that this has been happening a lot in the 2020 Democratic primary campaign. Someone will say something like: “if my candidate is elected, we’ll really start to see some profound structural/systemic changes!” They usually neglect to take the following sorts of questions seriously: what sort of structural/systemic impediments are there to your candidate implementing said policy? How confident are we that this candidate’s policy will have its desired effect? What sort of knock-on, second-order effects might the policy have? What sort of biases might be causing me to overrate how much better this candidate is from their competition? And so on.

As usual, we need to think about impact in a counterfactual way. And I think this is really hard to do. Some good considerations may be: How likely is a given candidate contribute to/alleviate the possibility of particular x-risks? Does this candidate have unusual focuses or abilities that will be unlikely to be matched by other candidates (such a candidate may support farm animal welfare or discuss his concern for the global (not merely national) poor)? My unfortunate hunch is that many politicians – unless they’re distinctly quirky/exceptional! – are somewhat interchangeable.

There’s been some EA reasoning in this domain. I found this article from 2016 where Rob Wiblin writes about the immense impact one president might have over another. I think a lot of his comments are spot on, but there are a few ways they don’t apply here:

Much of what he’s talking about is absolute impact, not counterfactual impact.
Most people would agree Trump was an anomalous candidate. The difference between Hilary and Trump is likely much greater than that between your average center-right Republican and center-left Democrat (and even less between two center-left Democrats, like many in the primary).
The impact of the US president is probably greater (by at least an order of magnitude) than a member of Congress.
Yes, if you were confident that one candidate had an X% lower chance of causing a war (and especially of causing an x-risk, as mentioned previously) that would be a good reason to assume they’d have a greater counterfactual impact.

You should also be concerned that, aside from counterfactual impact, you might just be wrong about absolute impact. Holden Karnofsky has also made some interesting and relevant comments about the risks of “adversarial philanthropy” in a blog post from 2013:

I’ve long been wary of giving opportunities that involve taking on other people as adversaries, and I think a lot of our audience shares my misgivings. One reason for this is that projects with active, intelligent opposition are likely to have more difficult—and unpredictable—paths to success. Another reason is the potential difficulty of being on the right side. When working on controversial issues, one can easily be blinded by personal biases and ideology into believing a particular change is more desirable than it is, with the result that even a “success” can end up doing more harm than good.

He goes on to say his fears have been lessened because he’s become more aware of the fact that there are “natural asymmetries of money and organization between different sides on a policy issue. The side favored by a consensus of informed humanitarians can be significantly (and importantly) under-resourced relative to the side with a structural advantage.” This is a good revision for many other kinds of philanthropic interventions in politics, but I think “natural asymmetries” apply less in the case of campaign contributions (often, as mentioned in the previous section, if a candidate has a real chance, they have enough funding). The “difficulty of being on the right side” is still very real.

Conclusions

If you were to assume money buys votes, and argue against contributions from a marginal cost-effectiveness point-of-view (spend $X to have a Y% greater chance of Z impact), I think you’d be successful. But it’s a tougher case to make than what I’m saying. Spending $X on most campaigns probably won’t budge that Y% at all. Even if that Y% budges, your counterfactual impact Z is likely to be quite low.

The one thing I didn’t add, because EAs have discussed it quite a bit, are the reputational advantages of staying apolitical. I haven’t thought too much about these, but they also seem to favor my position.

Finally, in this post, I’ve assumed that all your money is buying you is an increased chance your candidate is elected. One set of reasons to contribute might be if your money is buying you something else. For instance, maybe you’re in a position to benefit from signaling to the right tribes. Your donation may influence other donors to contribute, or, if you contribute enough, maybe you’ll get to meet the candidate and influence them in some way. If you’re in this unique position where your contribution is buying you these “extra bonuses”, then you should account for whether they’re worth the cost. For most of us, they don’t apply.