This will use the final polling average from RealClearPolitics for battleground states and compare them to the actual Democratic vs Republican spread when the voting totals are finalized.
Resolution will be a percentage based on the following formula using the Republican vs Democratic spread:
50% + (smaller error/higher error) * 50%
100% represents a perfectly positively correlated error (both were +3 points favorable to the Democratic candidate vs results)
0% represents a perfectly negative correlated error (one was +3 favorable to Dem, one was +3 favorable to Rep)
50% represents no correlated error (0 points favorable to Dem, +3 favorable to Dem)
In between will represent the correlation strength based on the higher error compared to the lower error
+3 favorable to Dem and +1.5 favorable to Dem would be 50% + (1.5/3) * 50% = 75%
+3 favorable to Dem and +1.5 favorable to Rep (-1.5 favorable to Dem) would be 50% + (-1.5/3) * 50% = 25%
The goal of this market is for traders to anticipate and model potential demographic overlap and provide a market idea of which states should vary together.
If Battleground states are added later (or are consistently aggregated in a similar way by RCP) I might add them. If Battleground states drop off (or see a significant reduction in polling) then I reserve the right to N/A all correlations for a state, although ideally I would find a different source for polling aggregation before that.
For pure Yes/No same party winners of states, I've also created this market: