Data for: Political Connections and Firms’ Tax Behavior: Evidence from Kazakhstan
Description
This dataset supports research on how political connections to regional parliamentarians influence firm tax behavior in Kazakhstan. The central hypothesis is that firms acquiring political connections through the March 2023 regional maslikhat elections will exhibit systematically different tax payment behavior compared to observationally similar non-connected firms, with effects potentially varying by connection level (regional vs. national MPs). The dataset covers 2017–2024 and includes tax payment records for 806 firms: 325 politically connected and 481 matched controls. Political connections were identified from biographical records of 643 newly elected MPs across all 20 administrative regions of Kazakhstan. A firm was classified as connected if an elected MP held a position, ownership stake, or managerial role in the firm for at least one year preceding 2023. Firms with pre-existing connections to MPs from 2016 or 2021 elections were excluded. MPs also holding national mandates were coded separately. Outcome variables are annual corporate tax payments (excluding VAT) and payroll taxes sourced from adata.kz, which publishes firm-level records from Kazakhstan's State Revenue Committee. Firm characteristics (size, age, industry, region) come from the Bureau of National Statistics Enterprise Statistical Register. Key findings show that firms connected only to regional MPs increased corporate tax payments by approximately 52% and payroll taxes by 79% post-election, consistent with the "grabbing hand" hypothesis. Firms connected to national MPs showed no significant tax changes, suggesting compensatory benefits at higher governance levels.
Files
Steps to reproduce
Step 1 – MP biographical data collection. MP data were collected manually from the official maslikhat portal (https://www.gov.kz/page/maslikhats?lang=ru) covering all 20 regional parliaments. Biographical profiles of 643 newly elected MPs from the March 2023 elections were reviewed to identify their last workplace prior to election. Only MPs with no prior parliamentary service in 2016 or 2021 were retained. Step 2 – Identification of politically connected firms. Firms were classified as politically connected if an elected MP held a position, ownership stake, or managerial role in the firm for at least one year immediately preceding 2023, yielding 325 connected firms. MPs also holding national mandates were flagged separately (nat=1). Firm characteristics (BIN, size, age, industry, region) were retrieved from the Bureau of National Statistics Enterprise Statistical Register, producing firms_with_polcon.dta. Step 3 – Control group matching (Stata). Using firms_with_polcon.dta as input, up to 5 control firms per treated firm were selected within each region, matched sequentially on industry, size tier, and nearest age. Output: bins_for_did_analysis.csv (806 matched firms) and matched_bins.txt. Step 4 – Tax data collection (Python). Using matched_bins.txt, annual corporate tax (Tax) and payroll tax (WTax) data for 2015–2024 were scraped from adata.kz (https://pk.adata.kz/company/{BIN}/taxes) using Selenium and BeautifulSoup, producing companies_tax_data_final.csv. Required libraries: selenium, beautifulsoup4, pandas, webdriver-manager. Step 5 – Final dataset assembly. Tax data from Step 4 were merged with matched firm characteristics from Step 3 on BIN. Four analysis-ready Excel files were produced: data.xlsx and data4.xlsx contain corporate tax (Tax_year columns), while data2.xlsx and data5.xlsx contain payroll tax (WTax_year columns). Files data4.xlsx and data5.xlsx additionally include the nat indicator distinguishing firms connected to national MPs, enabling three-group analyses.