Mean absolute error 176 votes per Place · correlation 0.9980
Akashic forecast laboratory
2028 Place vote forecast
Separate Democratic and Republican vote models for 30,999 Census Places. The geography model estimates where each party's votes are likely to be distributed; the selected scenario supplies the national turnout and party environment.
Active scenario
neutral 2028
Arithmetic mean of 2016, 2020, and 2024 matched-Place vote/CVAP environments
Projected Places30,999
Projected CVAP183,276,672
Democratic votes/CVAP34.6%
Republican votes/CVAP27.1%
Turnout-reconciled Places934
Deepest reconciliation54.5%
Locked 2024 holdout
Out-of-time validation
Mean absolute error 139 votes per Place · correlation 0.9948
Search any modeled Place
Use a city, town, village, CDP, or its seven-digit Census Place FIPS.
Method
What the forecast does—and does not—claim
The model fits Democratic and Republican votes separately. It combines a votes-per-CVAP rate model, a CVAP-exposure count model, and calibrated persistence. Ensemble weights were learned by Place-size decile on 2020 and then tested without retuning on 2024.
Census inputs come from the Census API. Population and CVAP are projected from recent ACS trends with small-Place shrinkage toward state medians. The neutral scenario is structural—not a poll or candidate forecast—and can be recalibrated when a real national environment is available.