A problem with political forecasts for the midterms is how much they rely on fundraising as one of the inputs. There is a wide gap right now between fundraising (which is off the charts for Democrats) and polling, which is not. The joys of curve fitting with 12 data points.
-
Show this thread
-
Even forecasts that claim to be 'polls only' still have fundraising as an implicit input. There's a feedback loop between polling and fundraising that makes safe districts look safer, since opponents don't get adequate funding to run a real race.
2 replies 2 retweets 5 likesShow this thread -
The upshot is, all the math in the world can't squeeze predictions out of this data, particularly when one of the indicators (fundraising) is now outside the realm of experience. No one knows what on earth is going to happen November 6
2 replies 2 retweets 5 likesShow this thread -
Replying to @Pinboard
I mean you can squeeze a good model out of ommited variable bias -- you just need to think of an instrument variable.
1 reply 0 retweets 0 likes -
-
Replying to @Pinboard
You want to find the effect of x on y but g affects both. Find z that affects only y through x and measure that for an unbiased statietical model. https://en.m.wikipedia.org/wiki/Instrumental_variables_estimation … The catch is you need to be clever with how the real world works to find a good z.
1 reply 0 retweets 0 likes
Thank you for explaining that. I think the 'clever with how the real world works' part is the shoal many grand ships of political forecasting run aground on
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.