In my work, I rely heavily on a number of different data sources, including surveys and official statistics, within Turkey and cross-nationally. My aim is to make all of these available for others to use, whenever licensing permits.
Nearly all of these datsets are based on TUİK documents which I have OCRed, cleaned, and merged. Because these documents are freely available (the majority from here and here), I consider their contents to be in the public domain and make them available here. Whenever possible, I use data at the district (ilçe) level to reconstruct newer provinces (iller) in the years before their creation. For clarity, I remove Turkish characters from all district and province names. For a list of districts, nested within provinces, see here; and for a panel that indicates when districts were created (and from which parent units), see here.
Provincial Council Elections
Ethnicity & Language
Mother tongue in censuses (Across provinces, 1927-1965)
Ethnic self-identification from KONDA's monthly Barometer series (2010-2015)
Religion & Religiosity
Enrollment in imam hatip schools (Across provinces, by Sex, 1961-2019)
Enrollment in Qur'an courses (Across provinces, by Sex, 1960-2019)
Number of mosques (Across provinces, 1986-2020)
Self-reported religiosity from KONDA's monthly Barometer series (2010-2015)
CODE FOR MERGING Cross-National Surveys
For my dissertation and book, I merged responses from the World Values Survey and European Values Study together with those from several regional Barometer projects: Afrobarometer, Arab Barometer, AsiaBarometer, Caucasus Barometer, East Asian Barometer, and Latinobarometer. Although instructions are available for merging the WVS with the EVS, no such instructions are currently available for merging either together with the other regional surveys.
Below, I offer Stata .do files for merging the regional surveys together with the aggregate WVS-EVS dataset. Each file renames (and sometimes recodes) those variables from the each regional survey that appear in the WVS, discarding those that are not found in the WVS, even if they appear in other regional surveys. (Eventually, I hope to add code that merges all variables and also covers the AmericasBarometer, Eurobarometer, and South Asian Barometer surveys.) The one exception to this is the code for the Latinobarometer, which standardizes variable names across all waves before recoding them. Some variables with country-specific responses (e.g., region, party preference, religion, ethnicity) are converted into strings which will require users to install the vallab command.
In the few cases where the regional survey questions are similarly worded to those in WVS but the answers are coded differently, I create a new version of the variable. To make the best use of these recoded variables when using the merged surveys, it is important to recode the WVS variables to match before merging.
If you use any of these files and find an error or have suggestions for how they could be improved, please do not hesitate to get in touch.