Thursday, May 28, 2015

TR Elections 2015

Applied a variation of the TfC model to the upcoming TR elections; the data I have is


The data is for popular vote percentage for the incumbent, as it was for the US case. The column gdp2 is a rough average of GDP growth of the election year and the previous year. This was done because of the turbulent nature of TR economics, for example year 2002 could have growth of 6.2 but previous year there was a crisis with growth of -5. I felt that this had to be reflected somehow in the numbers. Column popularity is a rough 0/1 value of leader popularity - Ozal had it, Demirel, Ecevit, Ozal caretaker Yilmaz didn't (for Demirel and Ecevit, their post 90s versions didn't, at least).

In case of a coalition, the vote percentage for the largest piece in coalition is counted. 

The two_terms column for TR has no effect - because this country experienced so many coups, anyone who can remain in power for more than 2 terms is probably pretty successful, so they don't get punished for this.

The incumbent for 1983 election is taken to be MDP because this is the party the generals of the 1980 coup "recommended" the public to vote for. Since the generals were in power, the MDP vote is considered to be a vote on them (the incumbent). They were ignored, Ozal won the election.

[geek] The final model has Adjusted R^2=80%. [/geek]. For predictions; For 2014/2015 GDP growth has been lukewarm, let's take this as 3.0. Leader popularity for AK, this is iffy, ex PM's popularity is not what it used to be, and then there is the new PM. Let's take this as 0.6, then the absolute best case for the 95% confidence interval shows 40% for AK. This is the _best_ it can be according to the model, chances are, it will be worse. Even .1 decrease in the popularity causes dramatic decrease in the popular vote.

Work, etc

(New way is too slow) An interesting article: says "simply having the option to decline a task has been shown to boost productivity ...