Last week, we launched the teaser version of our new website. In short, our intention is to 'bring a Moneyball approach to [primary challenges], and help progressive activists target their resources on the lowest-value members in the lowest-risk districts.'
One of the reasons we released a teaser instead of waiting to release our new website in full was because we were looking at this month as a type of beta test to see if our algorithm accurately captured who progressives thought deserved a primary in their own districts before the full-release of the site. Luckily, we had a great first week and heard from a lot of people. The response was overwhelmingly supportive, but we also had many smart people email us in our quest to further perfect our algorithm.
How Our Scoring System Works
We arrive at these primary scores through a two-step process: First, by weighting and averaging various partisan scores like Progressive Punch, DW-Nominate, and their party-line vote percentage, we get a clear picture of each member’s voting habits. Then we compare that value to other members representing similar districts in the current Congress. This is crucial, since members aren’t being judged against some woolly progressive ideal. For example, a Democrat representing a district with a D+4 partisan lean is compared to other Democrats in D+4 and similar districts — and the more conservative they are than those colleagues, the higher their primary score. This, along with the rest of our methodology, creates an algorithm which allows activists to find out where they can replace Democrats too conservative for their state or district with real progressives — with little to no fear of losing to said seat to Republicans.
Perfecting This Algorithm
As previously mentioned, we had conversations about this algorithm with progressives across the country looking to tweak any potential reoccurring issues people saw.
- Swing-District Democrats
The most common criticism was that our score was a bit harsh on Democrats in swing districts. Since our objective is to target Democrats too conservatives for their district along with a bit of reassurance that a Republican won't pick up the seat, we starting digging in a bit more.
Originally, our thought-process with regards to these conservative-voting members in D+4 and more Republican districts was (a) that their voting record was bad enough that a primary was worth the risk, because we could still get a Democrat in that seat who votes far more progressive and can stay safe based on their colleagues who are safe in similar PVIs. We still think this is right also -- which is why these swing members who were an 8/10 or a 9/10 primary score (must be primaried) last week didn't drop off completely, but dropped to a 4/10 or 5/10 (could be primaried). And (b) we were already giving candidates in R+ districts additional slack to populate our primary score, but it was a linear formula giving additional slack based on the x in an R+x. Expanding on this, we decided a polynomial formula was much more fitting to giving swing-district members additional slack based on the probability of a Republican picking up a seat with that PVI -- which we think made our scores even more accurate.
- New Members
The other less-regular but also understandable critique we heard were that these primary scores are a bit unfair to newer members of congress. We are open to more arguments, but this is our reasoning for treating new members the same as veteran members: First, our scores are updated weekly. While you're obviously not going to get a very clear picture of a MOC after just a few weeks of being in office, our regularly-updated scores will give a better sense of each representative's current trajectory -- which is the important thing early-on. And more specifically, the scores are heavily weighted on the current congress rather than lifetime scores -- regardless of whether you've been in office for one year or 20 years. So the only difference between new members and veteran members in this sense is that veteran members who voted progressively in previous years are given more slack in our scores if they're not voting as well in this current congress -- whereas, inversely, members that vote progressively in the current congress but have a history of voting more conservative, our scores are a little more tough on. Nobody should make a decision to primary somebody based on a voting record of a month or two, but by the time those decisions are getting made, each member would have cast enough votes to know whether they're likely to be a friend or foe of the progressive movement.
- Other thoughts?
Please let us know! We feel strongly that these scores accurately represent each member of congress that we attribute a score to. But ever since starting this project, we've been adamant about openness -- both with regards to explaining our algorithm along with our openness to tweak things about our methodology based on strong points from progressives who are looking to help.
Check Out The Teaser Site!
Check out our teaser site and sign up to see the 48 Democrats who are delivering the lowest value for progressives. We update the scores on our site every week based on new votes. And stay tuned for the full Primary Colors site, dropping at the end of the summer!