A week ago, the Economist published a study on Brazil’s presidential election, in which they weighted the candidates’ vote shares by the GDP of (27) states and came up with an interesting result.
The Economist has recalculated the result, weighting it by the 27 states’ GDP rather than their population of eligible voters. If reais of output went to the polls instead of citizens (which they thankfully do not in a democracy), Mr Neves would beat Ms Rousseff by 53% to 47% (see pie charts).
Inspired by this study, I calculated regional Gross Value Added (per head) weighted vote shares for Turkish local elections, 2014. While the Economist’s work weights the vote shares of parties in 27 states by each state’s GDP; I weighted the vote counts (instead of vote shares) by regional GVA (per head of population), as this metric is provided by Turkish Institute of Statics and known to be a useful way of comparing regions of different sizes. To be more specific:
- First, I wrote a scraper and collected the election results from YSK (Federal Election Institute of Turkey)
- Extracted RGVA per head table from TUIK April 2014 report (Appendix 5, pp 57)
- Finally, recalculated RGVA (per head) weighted vote shares by
- Multiplying the vote counts of each party in each region by that region’s RGVA (per head)
- Aggregating the regionally weighted votes of each party separately
- Getting the new (weighted) party shares among the total weighted votes
So, how should we read the resulting charts?
One might infer that CHP voters are wealthier then others, or BDP voters are poorer overall; however, I cannot tell how statistically significant these interpretations are. See Ecological Fallacy. If we had a higher granularity RGVA (per head) data then I believe we would come up with a more concrete (and more divergent) results.
I believe that a major contribution of this study is scraping the ballot box level official election results from YSK’s website and making it available to the public in a single table. Anyone who wants to analyze Turkey’s 2014 local elections can now do so easily, which might otherwise be a very cumbersome task for social scientists to manually collect this data from YSK’s website. Along with the analysis files, on gitHub, I shared my scraper script as well as the organizer/merger script that I wrote to reformat and combine the election results into a single CSV file.