on
Blog Post 7
Shocks and Pooled Models?
In today’s blog I will briefly go over the idea of shocks as well as the final changes to my model including the potential for using a pooled model. Shocks are sudden events that may come up unexpected in an election season. In 2022, the unquestionably most impactful shock was the Supreme Court decision in Dobbs v Jackson, which overturned the legal right to an abortion. I was unable to find any meaningful way to use shocks, including the abortion decision, to predict elections because this is obviously not something which happens year to year or in previous elections. I also opted not to look into other shocks from this election cycle because I did not personally find any other shocks that I thought expanded over a long enough period of time to impact the election.
Before I get into the updates to the model for this week, I want to go into the other key component of this week’s analysis. I will be talking about the potential of a pooled model and why I will not be using it for my model. While I think turnout by demographics is obviously very important, from what I have looked into, I think even the effects that the largest of demographic changes may have on the final result of a prediction are overstated. For example, I looked at doubling various demographic groups and saw smaller than expected changes on the final predictions. Part of this is because within one district, the impact of a demographic may be hard to understand. While we may understand the national or state impact of a demographic group, within one district that group may happen to be represented differently and may not be the same as a demographic group in a different district. People are diverse and models are hard and even harder for such small samples of by district. Additionally, because the 2022 election is the first election with the 2020 census congressional maps, I think it also would not be good to incorporate a model that uses demographic data or predictions from potentially different districts.
Updates to model
This week, there are no new signifcant updates to the model. However, I have gone and used the district model which I showed in detail last week and used it for all of the districts that should be close and of interest. One small change is that I have updated expert predictions as of 10-27-2022. In the table below, you can see the list of predicted winners and Democratic 2-Party voteshare for these close districts. The weights and formula is the same as the MI-10 I showed last week where Cook PVI of a district is 0.3, the expert prediction model is 0.3, the polling is 0.2, and the economic model is 0.2 However, even though our model is already not great for district polling, it is made even worse in that some districts either had no polling (or the only polling seemed very bad to me because it wasn’t even polling the actual nominee or was very old). In these cases, I had to drop polling and chose to add this weight to the expert predictions, which I have found to be the best of our models, so expert predictions becomes 0.5 in that case. I have included a table representing MI-10 and a table reprsenting TX-15 with the detailed breakdowns to show one example of each of the two weights. One other thing of interest is that I have noticed in some districts which have a PVI indicating a not so close race, Democratic incumbents might be hurt by this in my model. In other words, some Democratic incumbents may be able to win in Trump or Republican districts, but my model doesn’t really account for that when considering PVI. Therefore, I am acknowleding this as one possible drawback which I haven’t been able to avoid. I still think on the whole PVI is important to consider the landscape of a district and account for voters who vote the same year to year. In the final prediction for the 2022 election, I will be using these close district predictions and again revisting my national model which has not changed in the past several weeks.
##
## District Predictions of Close Races
## ==============================================
## District Winning Party Dem Voteshare (2-Party)
## ----------------------------------------------
## AZ-2 R 45.7470851582591
## CA-22 R 48.7924297143008
## CA-27 R 48.7547837526603
## CO-8 R 47.7627564350919
## IA-3 R 47.5058427211595
## IL-17 R 48.7917663721384
## KS-3 R 48.7787975251267
## MD-6 D 50.468004448844
## ME-2 R 49.0432976863614
## MI-10 R 46.4641210999723
## MI-3 D 50.2224333767329
## NC-13 R 47.6695632939094
## NJ-7 R 47.5439896178719
## NM-2 R 48.3163292900566
## NV-3 R 48.9933306670497
## NY-18 R 49.78956801633
## NY-19 R 47.9841852168202
## NY-1 R 45.8281489154832
## NY-22 R 47.5349576438108
## NY-3 D 50.4847897120962
## OR-5 R 48.5908127123744
## OR-6 R 49.4188689896569
## PA-17 R 48.5598612823508
## PA-7 R 47.7408442976346
## PA-8 R 47.9598612823508
## RI-2 R 48.9277947086825
## TX-15 R 46.4970851582591
## TX-34 D 50.3149388120414
## VA-2 R 47.7242743580655
## ----------------------------------------------##
## Michigan 10th Congressional (District Democratic Predicted Vote Share)
## ===========================================================
## Prediction R-Squared
## -----------------------------------------------------------
## Overall Prediction 46.4641210999723
## PVI (0.3) 48.5 NA
## Expert Prediction(0.3) 43.863209901485 0.702252438403214
## Polling Prediction (0.2) 50.1175395137293 Low
## Economy Prediction (0.2) 43.6582511339047 0.447206543351126
## -----------------------------------------------------------##
## Texas 15th Congressional District (Democratic Predicted Vote Share)
## ===========================================================
## Prediction R-Squared
## -----------------------------------------------------------
## Overall Prediction 46.4970851582591
## PVI (0.3) 48.5 NA
## Expert Prediction(0.5) 45.8308698629563 0.702252438403214
## Polling Prediction (0.0) NA Low
## Economy Prediction (0.2) 43.6582511339047 0.447206543351126
## -----------------------------------------------------------