Princeton Election Consortium

A first draft of electoral history. Since 2004

House Control Meta-Margin: R+0.5 Also Check out the new Beta!

Optimal Donations 2018: Senate and Governor’s races

August 26th, 2018, 4:40pm by Sam Wang


In 2016, I promised to get away from giving you probabilities. Here I show that under current conditions, you can optimize your donations without such calculations.

As has been the case in past elections, I’d like to point out key races where supporters of either Democrats or Republicans will get the maximum impact for their donation. (I’m assuming that control of either chamber of Congress has comparable value – see reader James McDonald’s point.)

This year, I don’t recommend giving to House races for two reasons.

  • There are just too many of them. Over 80 districts are less Republican-leaning than Ohio’s 12th district, whose margin was razor-thin in the recent special election. Compared with the six key Senate races I’ve listed at left, your donations would be diluted by more than a factor of 13. This is the big reason.
  • Under current conditions, we are closer to threshold for switching Senate control than for switching House control. Therefore Senate control is closer to the knife’s edge.

Taken together, these two reasons dilute the impact of your donation by a factor of 30-100*. This can change, depending on how the overall odds of House or Senate control change over the next two months.

Instead of giving you probabilities, which leads to possible complacency, I will simply ask: how far are Democrats/Republicans from threshold for House or Senate control?

In regard to effective giving, start with individual races. Once the margin between two candidates gets beyond about 5 percentage points, the probabilities get saturated. Money and activism don’t really move the needle. Instead, it is more strategic to look for races that are closer to threshold. Giving to, say, Dianne Feinstein’s re-elect campaign would be a waste of money. The same would be true for giving money to Senator Roger Wicker – again, no point. But the Nevada Senate race between Jacky Rosen (D) and incumbent Dean Heller (R) – now there’s a close race.

The same logic applies to House and Senate control. But now we have to both identify close individual races, and also ask where the chamber stands as a whole. Here are some examples of my quantitative analysis from 2014 and 2012.

With all that in mind, here is some information about 2018.

I estimate that for Democrats to have an even-odds chance to take control of the House, they would have to win the national popular vote by about 6 percentage points. I like to look at two indicators that reflect real voter behavior at a national level: special elections (i.e. real voting) and the generic Congressional ballot question (i.e. opinion polls).

Midterm opinion polling is hard to do accurately, and was five points off in 2014. Also, it can change over time. So let me put that aside until a later post.

In special elections since 2016, Democrats have outperformed Hillary Clinton’s performance against Donald Trump by an average of 11 percentage points. She won the popular vote by 2 percentage points, so if this pattern were applied to the whole nation, it would translate to a 13-point popular margin. That would be the biggest margin for either party since the post-Watergate year of 1976, when Democrats won the national popular vote by 13.6 points (and before that, 17 points in the Watergate year of 1974). Historically, special election performance has been within about 6 points of the mark, so we could expect to see a margin of 7 to 19 points this fall.

Thirteen percentage points is a lot. Even a margin half as large would still be enough to switch several dozen seats from Republican to Democratic, and make Democratic control likelier than not.  Because 13 points is 7 points above the magic threshold, I’d give Democrats a handicap (to use golf terminology) of 7 percentage points favoring them.

The Senate is considerably closer to the knife’s edge.Senate control is said to be a difficult challenge for Democrats. However, the eventual seat margin will be close, and the number of critical races is small. If we look at current polling margins, a swing of 3 points would be enough to put Democrats on the brink of having 51 seats. So in the Senate, Republicans have a handicap of 3 percentage points favoring them.

I should throw in here that close Senate races tend to break mostly in the same direction on Election Day. Which way they’ll break isn’t known; one way gets Democrats to 51-52 seats, and the other way gets them to 45-46 seats. It appears that Senate control could go either way.

Bottom line, Senate control is on a knife’s edge more than House control. I could convert that to  probabilities, but my feeling is that does not have a clear benefit. (Ballpark, it’s 0.9 for Democrats in the House, 0.7 for Republicans in the Senate.) For now, I will go out on a giant limb and say that the basic reason for paying more attention to the Senate is

7 > 3.

By the way, longtime readers of PEC will recognize these quantities as being conceptually related to the meta-margin, which I have calculated for Presidential races. To give you an idea of uncertainties in this quantity in tight situations, the meta-margin was around Clinton +1.3% in 2016 and ended up as Trump +0.7%, a 2-point error that was highly consequential.

>>>

In addition, I have added governor’s races. I focus on one race: Wisconsin, with Tony Evers (D) challenging the incumbent, Scott Walker (R). This is a long-term investment. Governors elected this year will oversee redistricting in 2021. Of the Egregious Eight gerrymandered states, Wisconsin is the one state where the governor’s office is the only identifiable route to attaining fairer districting. (For other close governor’s races, Larry Sabato’s Crystal Ball estimates include Colorado, Florida, Minnesota, and Ohio).

Still thinking about the House? That’s good – this is no year for giving money and leaving it at that. This is a year for getting out there and doing everything you can. I suggest a two-prong strategy: (1) give money to national Senate/governor races (key races are listed at this ActBlue site for Democrats; Republicans can give directly to the NRSC), and (2) engage in personal get-out-the-vote activity in a district near you. Find a competitive district near you using this tool!

As I wrote above, I don’t recommend donating to House races. I guess an exception is close races in Florida, where increasing turnout in a Congressional district also moves the statewide Senate vote. So…Florida districts 18, 26, and 27.

*Making this calculation accurately requires using probabilities. Many considerations could go into it , such as estimating probability distributions and advertising costs.

Tags: 2018 Election · governors · House · Redistricting · Senate

24 Comments so far ↓

  • Paul Marino

    Sam,
    I went to your site to determine to whom to donate. I see you are suggesting that I should focus on senate and governorship races. You mention 6 key senate races, yet I don’t see these on your website? Which are they and other than Wisconsin, what other governorship races would you suggest I focus on ($ is going to Democrats)?
    6 yrs ago you suggested North Dakota (Heitcamp) and I donated a decent sum to her race and she won – so I am ready to take your advice once again. I also donated to Jones and Lamb this cycle – my batting average is pretty darn good.

    • Sam Wang

      Thank you! See the ActBlue at left?

      Be forewarned, part of the point of finding these knife-edge races is that sometimes the candidates don’t win…

    • Gelatinous_Cube

      Sam, the ActBlue link to the left is a graphic that doesn’t say “ActBlue” anywhere on it. (I’m on mobile.) This is confusing and a little frustrating. Can you make it clearer?

    • dawolf

      Hi Paul, in addition to the 6 races listed, there is also Arizona. Texas is also down to Lean Republican as well, with O’Rourke closing in on Cruz fast.

  • Liz

    Sam, why is your model so much more optimistic (again) than the 538 model? In 2016 you told us Hillary was going to win, and that sending her money was a waste. 538 said she was 0.8 to win, and she lost.

    Now you say the House is 0.9 to flip. 538 has them at 0.7-0.8. If they’re right, donating to House races (especially underfunded close races) is a good use of funds.

    Is it possible you’re overconfident again? If so, why not post the most effective House races to give to *given that one is determined to give to a House race*?

    • Sam Wang

      Sure, it’s possible. But it doesn’t really affect the calculation – the large number of House races dilutes your contribution quite a lot.

      For House races, I already made my recommendation: get out the vote in a key race in your area. I suppose you could give that person money. But in all sincerity, this is not a normal year where you can write a check and stay at home.

      I guess if you are committed to giving to House races, give to Florida swing districts – you will also be affecting the Florida Senate race between Bill Nelson (D) and Rick Scott (R).

    • 538 Refugee

      The indications were that Hillary may have miscalculated and spent money to run up the popular vote to get a ‘mandate’ and miscalculated the electoral vote,.

  • mediaglyphic

    Dr. Wang.
    It might be interesting to know what inputs you are using to arrive at your .9 probability?

    • Izzy Eiss

      On the left, towards the top, under “High Impact Races” you’ll see “Donations for Democrats”. Clicking that will take you to the PEC High Impact Races page at ACT Blue.

      Izzy Eiss

    • Sam Wang

      I have modified the post. But no, I do not plan to give an extended discussion of probabilities this year.

  • Amitabh Lath

    I agree, if Democrats do not gain the house given this environment, there is little any donation could do.

    However, there are second order effects to giving to specific house candidates. The local get-out-the-vote effort directly helps state level candidates.

    I am giving to Andy Kim in NJ-03 because incumbent Republican Tom MacArthur needs to go. He is the one that brought Trumpcare back from the dead and if not for the late John McCain’s thumbs-down, it would be law.

    I hope Kim’s ground game makes this bit of Burlington and Ocean county a little more D at the local level.

  • LondonYoung

    I don’t trust the methodology giving the dems a 13 point lead in the house because it excludes the non-special elections. On a vote-weighted , rather than election-weighted basis, it will be closer. Nate Cohn put it like this:

    “The relatively unimpressive Democratic performance in Virginia and New Jersey is particularly noteworthy: The two elections are usually entirely omitted from analysis of recent races, even though more votes were cast in those two elections than in all of the special elections combined.”

    https://www.nytimes.com/2018/03/20/upshot/special-elections-democratic-wave-midterms.html

    • Sam Wang

      I have to think how to score those. For good or bad, they aren’t special elections; they are normally scheduled elections.

      In the 2017 Virginia House of Delegates election, the statewide vote was D+9.4, 4 points ahead of the 2016 election. Note that Virginia elections are not exceptionally accurate at reflecting the national climate the following year. See this tabulation by Kyle Kondik and Geoffrey Skelley. 1993 missed the 1994 GOP wave, but 2009 caught the 2010 wave.

      The New Jersey Assembly statewide vote was D+16.7, 2 points ahead of the Presidential election.

      My inclination would be to score each as one election – and to lump other same-day, within-state special elections in the same manner.

    • LondonYoung

      So, the regularly scheduled and well attended governor and house races in NJ and VA were in the range of +0 to +4 movement towards the blue team.

      The irregularly scheduled and very poorly attended specials moved 11 points blue.

      The elections this fall will be regularly scheduled and well attended. Feels like dems +6?

  • Joseph Pleck

    Thanks for this useful analysis. Friendly question: in evaluating whether it is more useful to donate to Dems in Senate races vs. House races, in addition to your analysis of likelihood of flipping a chamber, should we also take into account the impact of one’s contribution relative to total spending in the race, where much more is typically spent in Senate races? Hypothetical example: if $20M is spent in an average Senate race, and $2M is spent in an average House race, wouldn’t a $1000 contribution have relatively more impact in the House race (albeit not a huge impact)?

  • Matt Brubeck

    What about the Senate races in Montana, Texas, and Arizona? Are there any likely scenarios where these races would be decisive for Senate control?

    • Sam Wang

      Montana, probably yes. Not enough data there.

      If Texas becomes competitive for Democrats, that’s probably super-bad for Republicans. If Arizona becomes competitive for Republicans, that’s super-mega-bad for Democrats.

    • Matt Brubeck

      What you say about Texas makes sense. I’m not so sure about Arizona. I know the primary isn’t even decided yet, but based on early polls it looks like it could be at least as competitive as Nevada. I guess we’ll know more when we have post-primary polling…

  • dawolf

    I think you need to consider House seats within tight Senate races. In tight races, money spent on House candidates has a double impact. Take Florida as an example. Scott is going to spend an absolute fortune. HHow much is an extra million worth if he is already spending 20? But within Florida, Cook Political lists the following currently Republican seats as Lean Democratic – Lean Republican: districts 15, 16, 18, 26, and 27. In those dual areas, it’s often about spending money on the better candidate. It’s like Dean’s 50-state strategy, but with an emphasis on House races where there are multiple races at stake.

  • James McDonald

    Sam,

    Your allocation makes an implicit assumption about the goal to be reached, apparently that we should be trying to flip both houses, with about equal significance.

    But if the goal is to flip at least one house (say to get subpoena power) the allocations could be different. For example, if that were the goal, then shifting the odds for the House from 70% to 80% might be better than shifting the odds for the Senate from 48% to 55%.

    Conversely, if the Senate is deemed to be more significant than the House (because it approves SCOTUS nominees, etc.) then moving the odds for the Senate from 48% to 52% might be better than moving those for the House from 60% to 90%.

    I’m not saying your model is wrong, but you might want to be more explicit about the goals, and perhaps consider the best strategy for other plausible goals.

  • Ebenezer Scrooge

    I think that there is a point to donating to House candidates. Donations to their campaigns buy them lowest-rate TV ads. More general donations only buy (much) more expensive market-rate TV ads. I’ve adopted one candidate to concentrate the impact of my puny checks.
    Agreed that it’s mostly about GOTV, and labor is probably more useful than cash. But cash is still worth it.

  • ArcticStones

    As someone said so succinctly:

    In November, give your “thoughts and prayers” to the Republicans – and your vote to the Democrats.

Leave a Comment