Mapping Elections in Turkey

This page provides several examples from my work on mapping electoral results, spatial party positions, twitter data scraping and campaign terms in Turkey. For higher resolutions, please send me an email. I use Google Fusion Tables, Arcgis, Qgis and sp package for R to map electoral outcomes. You can download the necessary gadm format file for R from gadm.org. In 2012, administrative boundaries of districts changed, so for mapping district level results from 2014, old gadm file with 912 districts was used.

Here are some examples from various exercises I conducted so far. Please cite this web page address if you use these visuals in your work.

Center Left & Right Vote Changes Across Elections

2015-2018 General Elections Difference:

2014 Local Elections District Municipality Vote Shares

Inflation adjusted difference of Tax Income and Public Investment from 2007 to 2012

(Red – more public investment for tax income per capita)

Comparing Proxy for GDP (Satellite Night-time lights) and Real GDP values

Turkish Statistical Institute did not publish provincial GDP statistics from 2001 until this year. Today, I was able to compare these results with TEPAV’s work on measuring GDP through satellite imagery of night-time lights as a proxy for industrial output and productivity. From 2004 until 2013, for ten years I compared the data and the correlation between satellite and measured GDP is never below 0.70 which is promising for satellite image proxies as used by Brian Min and others in their work on measuring economic output in less developed countries. For those interested, see here for TUIK’s GDP across provinces and here for TEPAV’s satellite measurement.

Gender Difference in Educational Attainment (2012)

The only district where women had more education compared to men was Bodrum, Muğla.

Continuity and Change in Party Support from 1961 to 2011 (13 Elections)

Factor scores retained after a varimax factor analysis. From a longitudinal dataset on party support over 13 elections-district level. Two axes is hard to interpret but roughly, they are on Islamism vs. Secularism (x axis) and ethnicity (y axis).

Party Leader Tweets

I scrape twitter data using dmi-tcat on a Debian server I installed on an Amazon cloud computing instance. Similar to the post-coup attempt tweets, here is another example from twitter scraping I’ve been doing: the graph below is a daily plot charting locally weighted regressions for all the tweets (10,318) sent by the leaders of the four largest parties in Turkey.

It seems like Kılıçdaroğlu (CHP) and Demirtaş (HDP) converged to around two tweets per day since 2014. Bahçeli (MHP) uses twitter most actively among the leaders and the AKP’s leader-cum-president Erdoğan’s twitter usage decreased after he became the president (probably because he switched to the official presidential twitter account). I forgot to add the Prime Minister and current leader of the AKP, Binali Yıldırım (old habits die hard – still thinking of Erdoğan when I think of the AKP).

P.S. Bahçeli has a very curious twitter usage pattern, he floods his twitter page with long messages. His first tweet is generally “Hepinize iyi akşamlar (good evening to you all)” and then he rambles, sending around 20 to 30 tweets per session, referring to poets, writers and thinkers to make his point. Out of his 3214 tweets, he sent 104 tweets only to say “hepinize iyi akşamlar” to his followers.

Key words for 2002, 2007, 2011 and 2015 Campaign Periods

Clouds were created by worditout.com. Dataset contains titles of the 15 most read newspapers’ political news stories for three month campaign period before the election. Had to clean suffixes as much as possible because of the Turkish grammar (Agglutinative language with millions of suffixes!).

Page updated

Google Sites

Report abuse