Хроники Попокалипсиса

https://wer-gewinnt-die-wahl.de/ (Black-Green Zhopa)

Marcus Groß about goldfish memory effect

The basic model idea follows a Bayesian state-space model, such as the model of Simon Jackman for the australian election. However, a major change is that the latent voter’s intention is not modeled by a random walk, but rather uses a “long-short term memory” extension. The concept behind this is that a specific event that influences the vote share of a specific party (scandals, nomination of candidates..) are first spread through the media which causes an overreaction or overshoot (short-term memory) with a maximum effect at around four weeks. Medium to long term, however, we see that this effect declines somewhat as other topics are covered by the media now and voters forget about the concerning event. As the remaining long term effect is smaller than the initial one, we see a regression to the mean behaviour. This is modeled by a time-series model structure resembling a mixture of autoregressive processes. This concept can be also confirmed empirically and improves the forecast quite a bit (between 5%-15% depending on the forecast horizon).

[...]

The 63% for CDU/CSU/Greens in the first graph do however not answer the question of getting a majority, but rather the probability of being in charge after the elections (i.e. constituting the government coalition). These probabilities therefore sum up to 100%! To derive these probabilities, the expert rankings are combined with the election outcome simulation results (see last section of the notebook.pdf [https://github.com/INWTlab/lsTerm-election-forecast] for a detailed explanation).

I tried a lot of other model formulations in stan, but the current one was the fastest I could do. The model takes only about 18 hours to fit currently, not a week; the web page is updated every day

Andrew Gelman about the response rate for polls

David Callaway writes:

I read elsewhere (Kevin Drum) that the response rate to telephone polling is around 5%. It seems to me that means you are no longer dealing with a random sample, what you have instead is a self selected pool. I understand that to an extent you can correct a model for data problems, but 5% response? How do you know what to correct for? Take another poll to determine what errors to correct for and how much? Use the time honored “listen to my gut” method?

My reply: I’ve heard that the response rate is closer to 1% than to 5%. As to your question of whether pollsters should just “go home” . . . I don’t think so! This is their business. And, hey, the polls were off by about 2.5 percentage points on average. For most applications, that’s not bad at all.