How I’m making an impact on the 2018 Congressional race

Like a lot of Americans, I’m disturbed by President Trump’s behavior. He doesn’t seem to respect the Constitution, the People of the United States, or human rights more generally. I’d like to see Congress place checks on the Executive Branch, but Congressional Republicans have made it clear that they’re not going to take any concrete action to do this.

Because of this, I want to see the Democrats win the majority in the House in 2018. Having some access to power will allow them to take real steps to check the Trump administration. But what can I do to help make that happen? I live in a relatively liberal district, so while I plan to vote, it is unlikely to make an incremental impact on the system.

This cycle, I’ve chosen to focus on political giving. Especially for challengers and in close races, having cash on hand is essential to winning races.

Because I can’t afford to be a major donor, this article is about my attempt to make sure that my limited funds go as far as possible in making an incremental impact in the 2018 House race.

In deciding who to support, I had three priorities (in order):

  1. Make sure I’m giving in a race that is truly competitive — if a race isn’t particularly close one way or another, my money would be better deployed elsewhere
  2. Make my money go as far as possible by looking for low-cash competitive races — once I’m reasonably assured that a race is competitive, I want to look for races that, all else equal, have less cash in them; my meager donation is worth 10x more in a $1M race vs. a $10M race.
  3. Screen out any obviously bad investments — I can’t do exhaustive research on every race, but if, for example, there’s a pending indictment on one of the candidates that’s likely to change the answer one way or another, I want to know about it.

I’ll go into how I went about this in detail below, I identified 102 races that had some likelihood of being competitive. From this list, I picked 30 races to research based on an assessment of how likely they were to be competitive and the amount of money in the race.

In the end, I settled on 8 candidates to support now, 4 honorable mentions, 7 high value long shots, 1 stay away, and 10 races to check back in on in September.

Candidates to support now

I selected these candidates because:

  1. They are in races that are likely to be close — these races were all competitive in more than one of the scenarios I considered
  2. They’re a good value — each of these races is at or below the median of competitive races for cash on hand based on the data I was able to find
  3. No red flags: Nothing stood out as a red flag in some light online research

The candidates:

Honorable mentions

These candidates are all in races that are likely to be close and have candidates without any obvious red flags, however they have more than the median cash-on-hand compared to other competitive races.

  • Colin Allred, TX-32: Colin is a former NFL player and current civil rights attorney.
  • Gina Ortiz Jones, TX-23: Gina is a former Air Force officer and civil defense servant.
  • Brendan Kelly, IL-12: Brendan is a former Naval officer and current State’s Attorney.
  • Perry Gershon, NY-1: Perry is a businessman and an active member of his community

High value long shots

These candidates are unlikely to be in competitive races unless the national environment really tilts towards the Democrats, however there is relatively little money in their races, so it provides an opportunity to run up the score.

  • Clint Koble, NV-02: Clint is a long time public servant, focused on rural and agricultural development
  • Francys Johnson, GA-12: Francys is a pastor and an attorney.
  • Mike Siegel, TX-10: Mike is a former teacher and city attorney
  • George Scott, PA-10: George is a former Army officer and current pastor
  • Ryan Watts, NC-06: Ryan is a management consultant
  • Rob Davidson, MI-02: Rob is an emergency physician
  • Linda Coleman, NC-02: Linda is a former teacher and state legislator. It should be noted that Linda hasn’t filed a personal financial disclosure. Given that the President refuses to release his tax returns, I have trouble getting upset about this, but it may impact the race.

Stay away

I only had one of these: Susan Moran Palmer, OH-16. Susan seems like a perfectly good candidate, but her opponent is former Ohio State star and 1st round NFL draft pick Anthony Gonzalez. Given the name recognition that he has in the area, I felt like this was unlikely to be a close race (although I hope she proves me wrong).

Check back in September

Last but not least, I ran into 10 races that don’t have final candidates yet because their states hold primaries until later this month, but would’ve otherwise qualified for at least an honorable mention, assuming the candidates selected don’t have any red flags.

  • MI-03
  • FL-15
  • FL-03
  • MN-08
  • AZ-06
  • OH-12
  • MN-01
  • FL-18
  • FL-06
  • WI-01

My approach in detail

As I stated, above, I had three priorities in my approach to deciding who I would support:

  1. Make sure I’m giving in a race that is truly competitive
  2. Make my money go as far as possible by looking for low-cash races
  3. Screen out any obviously bad investments

To do this, first I looked at recent political history and data on the national political environment to try and generate some likely scenarios for the 2018 House Race. Then, I combined these forecasts of the national environment with the Cook Political Report’s Partisan Voter Index to get a district level estimate of each race in my different scenarios. I also attempted to correct for the advantage that incumbents have over challengers on a by scenario basis.

Forecasting the National Environment

To try and identify races that are likely to be close, I started with three potential scenarios for the national environment. I came up with these scenarios by reviewing the congressional popular vote count of the last 25 congressional elections to get a sense for how common gaps of various sizes are and comparing that with the 538 polling chart for the 2018 Congressional election. My goal here is not to predict the outcome of the election, but come up with a range of plausible outcomes.

National vote differential over the last 25 Congressional Cycles by year (1968–2016)

It’s pretty rare for the party of the sitting President to gain ground in a Congressional election

National vote differential over the last 25 Congressional cycles aggregated (1968–2016)

The Challenging Party has performed the President’s party by a median of 5.7 points since 1968; this goes up to 7.9 points if you only look at non Presidential cycles.

One thing that stood out to me about this is that it’s pretty rare for a President’s Party to gain ground in a congressional election, especially in non-Presidential cycles. In the last 25 congressional cycles, the median value is a +5.7 point national vote advantage for the Challenging Party; this goes up to +7.9. The lesson here for me was that advantages of more than 5 or 6 points in the Congressional national vote have been happening with some frequency and so I shouldn’t be surprised to see that possibility .

In addition to forecasting the overall national environment, I tried to account for the advantage that incumbents have running against challengers. The best information I could find on this is from Vox/Fair Vote and it indicates that it’s somewhat related to the national environment — basically, the more there is a national consensus on the House race, the less any particular incumbent’s relationship with her district matters. From Vox:

In more-or-less neutral congressional years without a big wave for one party or the other, FairVote finds the incumbency advantage has recently been somewhere between 3 percent and 7 percent.

But in wave years, the parties’ incumbents’ fortunes sharply diverge. In 2010 and 2014, for instance, Democratic incumbents barely had any advantage at all (0.9 percent and 0.46 percent). And in 2006 and 2008 — two anti-GOP wave years — the Republican incumbent advantage shrank to 2 percent, while Democrats’ rose quite high.

Using this information, I generated 3 potential scenarios for the national vote and relative value of incumbency this fall. Note that I’m not saying that any one scenario is more likely than another, but instead trying to cover the likely range of acceptable outcomes.

Potential scenarios for the 2018 election

The scenarios I used to forecast the 2018 election

Scenario 1: No wave

My first scenario is a “low” scenario where the national environment favors the Democrats by about 4 points. This scenario would be below the median historically for the party that does not hold the White House and seems to be about the low end of the 538 chart. Because of the relatively neutral political national political environment, I’ve assumed a relatively high incumbent advantage of 5 points. This scenario results in some pretty significant Democratic gains, +6 seats before races within 2 points or less. While this scenario seems somewhat unlikely to me, I don’t want to rule it out entirely.

Scenario 2: A moderate wave

My second scenario is a “mid” scenario where the national environment favors the Democrats by about 7 points. This would be a little bit more than the median for the party that does not hold the White House in the past 25 congressional elections and a little more than the median for this party in non-Presidential cycles. It also seems to be closer to where the the 538 polling chart has been for the past month. Because of the more partisan national environment, I’ve assumed a smaller incumbency advantage of 4 points. This scenario results is very significant Democratic gains (+13 before races within 2 points or less) and a lot of very close races (45). In this scenario, it looks likely to me that Democrats take control of the House.

Scenario 3: A big wave

My final scenario is the high scenario for Democrats where they have a 10 point national environment advantage. This would be well above the median for historical presidential cycles and is at the high end of the 538 polling chart. Because of the large national consensus in this scenario, I’ve assumed a very small incumbency advantage of 1 point. This results in a wave scenario, with the Democrats picking up 45 seats before races within 2 points either way, of which there are 62. In this scenario, the Democrats definitely pick up the House and are more concerned with running up the score.

Creating district-level estimates to identify potential close races

Next, I applied these three scenarios to individual races using the Cook Political Report’s Partisan Voter Index (PVI). PVI is an estimate of how much more liberal or conservative a given district is than the national environment. This allows me to use my scenarios from above identify races that are likely to be close in the given scenarios.

Scenarios and forecasted district-level results

Different National scenarios have the potential to shift the competitve ground quite a bit.

This gave me 102 districts that were estimated as competitive in at least one scenario, which I defined as being within 2 points in either direction. 20 of these races were estimated to be competitive in 2 of the scenarios outlined above. Next, I combined this with cash-on-hand data from the Center for Responsive Politics Open Secrets API to get an idea of which competitive districts had comparatively more or less cash in them.

This data definitely has some limitations. For one, because it doesn’t have the general election status of the candidates, I had to summarize it at the district level, rather than at the candidate level. This means that if a district has a competitive primary with lots of opponents that have a fair amount of cash on hand at the end of the primary, it might show up as artificially high (I’ve illustrated this below). I’ve tried to mitigate this by focusing on cash-on-hand rather than spending-to-date, but it does make the data less valuable. Second, at the time I pulled it, all of the Pennsylvania data was still using the old district definitions (from before the State Supreme Court did redistricting); I’ve done my best to correct for this where possible (more on this in the data section).

How a competitive primary could distort cash on hand data

Remianing cash on hand for primary candidates who have lost still counts towards the race overall.

In this scenario, District A would show up as having $400,000 in cash on hand left in the race while District B would show up as having $500,000, when in reality, the candidates still running in District B only have $300,000 in cash on hand.

Cash on hand by District that is competitive in at least 1 scenario

Not all districts labels appear due to space constraints.

Despite these limitations, I felt it was helpful as a directional guide as to where to invest. As you can see in the chart above, there’s a wide range in values, with the highest cash-on-hand race having ~8x the median. By focusing my giving on lower spending races, I can increase my impact by a significant amount.

In the end, because race competitiveness is my highest priority, I decided to investigate further all 21 of the races that were competitive in more than 1 scenario (the red districts above). In addition, I selected 9 districts that were only likely to be competitive in one scenario, but had relatively low cash in the race. This left me with 30 races to research. For each of the races in question, I did searches for both the Democratic and Republican candidates to try and get a sense for any major issues.

Risks to this approach

I think the greatest risk to this approach are changes in the national environment.There is still some time until November, and it is possible that something could happen which fundamentally changes the shape of the political landscape for this fall in a way that changes which races are likely to be competitive.

Another risk is the impact of new voter suppression acts, which would not be captured in prior district-level results, but could lead to me being systematically off in certain states. I believe that the effects of gerrymandering, on the other hand, should be already be priced in because the Partisan Voter Index looks at the past voting history of a district and gerrymandering affects who gets to vote where.

As I discussed above, the cash-on-hand data leaves a little to be desired. Last but not least, my candidate-level research is admittedly light weight.

Notes and data sources

  • My master spreadsheet is available here. It is a little messy, but feel free to take a look at it.
  • Here is the Cook Partisan Voter Index data
  • Incumbency data is from Wikipedia; this may be out of date as I pulled it in early July.
  • Cash on hand data is from the Open Secrets API; while this data is great, it can be a little noisey for a couple reasons. First, it pulls in all of the candidates from a district, not just the ones still running, so a wide-open primary or a special election can make a particular district look like it has more spending in it than it actually has. Second, all of the Pennsylvania data did not take into account redistricting at the time I pulled it, so I overwrote it as best as possible, occasionally pulling in data from Ballotpedia (and noting this where I did so).

operating a bit outside the lines