An “effective altruism” approach to the election

September 28, 2020 by Sam Wang

September 29: Modified to take into account change in Florida ratings.

If you’ve been giving to races we’ve identified as high voter power, you may have noticed that once in a while, we make small changes. The reason is that our best estimate of return on investment responds to events or additional analysis. We will continue to follow this process until November 3. If you have more to give, I recommend checking once a week.

Today we introduce a new criterion, how well races are already funded, information that our friends at Data2ThePeople have been investigating.

The voter power calculations in the sidebars of this website give estimates of the ability of a single vote (or small number of votes) to shift the probability of the outcome: for President, for individual Senate seats, and for control of state legislatures. In the case of state legislatures, power is measured by ability to change the redistricting of congressional seats – Redistricting Moneyball). These measurements are exactly what you need if you are registering or turning out voters. They are also an approximation of where campaign donations will be most effective. This has driven the choices in the WinRed and ActBlue links.

For purposes of donations, this can be taken a step further by considering one additional real world factor: how well a candidate is fundraising. This is especially important in state legislative races since at that level, an entire campaign budget can be in the tens of thousands of dollars. In key Kansas legislative races flagged by Data2ThePeople, September fundraising ranged from $4,570 to $20,113. Running in other states is more expensive. Still, raising as little as $10,000 is liable to get you a really nice thank-you note.

For example, close North Carolina state House races have raised $1.5 million more for Democrats than for Republicans. So Carolina Democrats are being removed, and Republicans are staying. Kansas Senate candidate Barbara Bollier (D) has raised $5 million more than her opponent Roger Marshall (R). The impact of additional money is likely to be marginal. So we are removing her (and leaving Marshall on).

Conversely, Alaska Democratic/Independent Senate challenger Al Gross lags incumbent senator Dan Sullivan in fundraising. Donations there are especially powerful because Alaska has so few voters.

Finally, Florida state legislative fundraising is quite weak, despite the overall money-saturated campaign at the Presidential level. The most recent CNalytics ratings show a shift in races that take the chamber further out of the 20-80% range (nominal probability of bipartisanship 13%). It is time to de-emphasize Florida.

The resulting optimized choices as of today are:

  • ActBlue federal: senate candidates Bullock (Montana), Gross (Alaska), and Harrison (South Carolina).
  • ActBlue state legislative (Redistricting Moneyball): selected geographic hotspots in Kansas and Texas.
  • WinRed: Senate candidates in Kansas, Montana, and South Carolina, and state legislative races in Minnesota and Connecticut.
  • Finally, if you find our analysis useful, consider giving to the Princeton Gerrymandering Project.

    14 Comments

    538_Refugee says:

    Couldn’t sleep so got up to do some reading. Following the side link on Lindsey Graham was interesting. One pollster that isn’t ranked and flagged as ‘partisan’ has a 3rd party candidate drawing enough support from Graham to potentially lose him the seat. The mainstream pollsters seem to think this is a non-factor but there may be enough Republicans fed up with Trump and Graham’s U-turn that they might vote 3rd party on this race instead.

    Sam Wang says:

    Third-party support generally fades in the finish. Also, Graham is about to play a prominent role in filling a SCOTUS seat. That should bring them home.

    James says:

    Sam, I’m wondering if the historical data are available for which candidates or groups were being recommended and when, why those candidates were recommended and possibly how much was raised through the sidebar link.
    Maybe you are already planning something like this after the votes are counted, but I’d like to do a “how’d we do” analysis. I want to include the “why” like whether the goal was to flip a chamber or prevent a super majority etc, since the desired outcome could still come about even if a particular recommended candidate lost.

    Sam Wang says:

    I think this is a great idea. In previous years it’s just been Senate races, not much grist for such an analysis. But Redistricting Moneyball provides a much richer landscape for assessing what went well, or not.

    Greg Hullender says:

    I’m worried about the House having to decide the election, in which case all that matters is having a majority of the representatives in a majority of the states. What I’d like to see would be a list of representatives to support to maximize the likelihood of at least reaching a tie in the House.

    Sam Wang says:

    Montana and Alaska
    https://projects.fivethirtyeight.com/polls/house/alaska/
    https://projects.fivethirtyeight.com/polls/house/montana/

    Froggy says:

    Certainly the ones that Sam mentioned, but let me add a few others.
    Michigan is currently 7D, 6R, and 1Amash. MI-03 (Amash’s seat) and MI-06 look to be the best possibilities for a Dem pickup that will give them the majority of the state’s delegation:
    https://projects.fivethirtyeight.com/polls/house/michigan/3/
    https://projects.fivethirtyeight.com/polls/house/michigan/6/
    Pennsylvania is currently split 7-7, and the best D pickup opportunity is PA-10: https://projects.fivethirtyeight.com/polls/house/pennsylvania/10/

    Sam Wang says:

    Froggy makes good choices here, especially Pennsylvania. However, I think the Michigan districts might be gerrymandered enough to make it tough for a Democrat to pick up in the end.

    Abe Fisher says:

    I’m a little concerned that your new methodology introduces a couple of problems into the system. Primarily, a candidate might be raising money effectively exactly because they are in a key spot and their opponent is raising money effectively (i.e. they’re in a financial arms race). Concluding that because they’ve been raising lots of money maybe we should devote resources elsewhere isn’t actually a slam dunk.

    538_Refugee says:

    It is (mostly) one person’s best educated guess influenced by informed opinion and applied to an imperfect system that is a moving target. Democrats are now focused on winning state delegations in the off chance the house is pressed into deciding the presidential election.
    And we are assuming the money will be spent effectively? In some smaller races name recognition is still a deciding factor.

    Sam Wang says:

    For federal races you can explore OpenSecrets for yourself.
    In South Carolina, Harrison (D) has raised $28.6 M while Graham (R-inc) has raised $29.9 M.
    In Alaska, Gross (I) has raised $5.2 M while Sullivan (R-inc) has raised $7.9 M.
    In Kansas, Bollier (D) has raised $7.8 M while Marshall (R) has raised $2.7 M.
    From these numbers I have formed the view that Bollier’s limiting factor is not money. And in fact, I agree with you that the other two contests have an arms-race quality to them.

    Leading Edge Boomer says:

    I am really tempted to donate substantially to close Senate races in states where advertising is not overly expensive. But I REALLY don’t want to be in ActBlue’s database because I will be hounded by them for the rest of my life.
    Going to a candidate’s website to donate almost always leads to ActBlue; they seem to have a near monopoly on donating to Democratic candidates. I am not federally required to divulge my phone or email address, but ActBlue requires them.
    How to donate without ActBlue?

    Laff Tea says:

    Thanks for your very useful analysis.
    I tried to give a donation through your actblue link. Kansans for a Democratic House was declined. I tried to search for this org in Google, but couldn’t find it. what gives?

    Karl Chwe says:

    The ActBlue page https://secure.actblue.com/donate/pec2020federal
    still talks about Kansas and the Kentucky nonprofit as donees, even though it appears no money goes to them. Perhaps they could update it?

    Leave a Reply

    Your email address will not be published. Required fields are marked *