I want to predict the election results in a country. Firstly I want to check and specify probabilty. For doing this, I intend to use the election results of the past years and the estimation results of the survey companies. For example, in the election held in 2012, the survey firm x made a prediction of 30 percent for party a, part b, etc... In the election that took place that year, party a came in first with a share of 32 percent. Here, using the difference between the 30 percent forecast and the 32 percent rate for verification, how can I predict the election results in 2023 with using the survey forecasts of the companies.

  • 2
    $\begingroup$ You have looked at what fivethirtyeight.com do (they have some posts describing their approach to some extent), right? They are kind of the poster child for doing this in a largely reasonable way. There's of course efforts in other countries (e.g. learnbayesstats.com/episode/…). Also note that survey firms will to some extent try to fix their past mistakes (unless they have a [deliberate??] partisan bias), so the mistakes they make in one election may be different from another election (538 discussed this a bit, too). $\endgroup$
    – Björn
    Oct 5, 2022 at 13:02
  • $\begingroup$ The Economist has also published information about (some of) their forecasts. For example: Forecasting the US elections. Which country, what election and how much relevant data you have (in addition to past results and polling) is going to be important. $\endgroup$
    – dipetkov
    Oct 5, 2022 at 18:43


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.