Military dictatorship and the provision of public goods

ABSTRACT

Non-democracies, particularly dictatorships, provide local public goods differently when compared to democracies. We use the Partition of the Indian sub-continent in 1947 to examine how similar ethnic groups living in similar agro-climatic conditions obtain substantially different configurations of public goods when exposed to different governance regimes. Our methodology draws upon the shifts in the central regime in Pakistan, between popularly elected governments and military dictatorships while using India as a benchmark, which had democratic governments throughout. We create and utilize a novel dataset for our district-level analyses from various census rounds in India and Pakistan. Our regression results consistently show that there is a significant under-provision of various public goods under dictatorships, while controlling for a host of time-varying local factors. Our results survive a battery of robustness checks and are particularly, not driven by large cities, or specific provinces.

Keywords:

• Democracy
• military dictatorship
• Partition of India
• public goods

JEL CLASSIFICATION:

• H41
• D72
• H53

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. https://freedomhouse.org/report/freedom-world/2022/global-expansion-authoritarian-rule

2. There were some questions raised on why we did not use data from earlier Censuses of British India as controls. We looked at the Census of India and Pakistan in 1961, but the combined number of districts that were in the border provinces/states of India and Pakistan were less than 30. The districts were larger in those years, and in later years, they were divided into smaller districts for administrative efficiency. If we did a crosswalk of districts from 2008 (in case of Pakistan) or 2011 (in case of India) to 1961, we would have too few districts to elicit statistical power of our regression coefficients. On the other hand, a crosswalk of districts till 1971/72 gave us a large enough sample for obtaining valid regression estimates. Furthermore, the Censuses of 1951 and earlier (there was no Census of 1941 due to World War 2) did not have any information on the number of public goods at the district-level. The information collected in the earlier Censuses was mainly counting people by gender, race, religion and caste. While it would be beneficial to use the educational attainment or public goods of districts from the 1800s as controls for the current level public goods in the districts, unfortunately, such information is unavailable at the district level in those Censuses. Subsequently, we did not use the data from earlier Censuses, and restricted our sample from the 1970s to 2010s.

3. Data retrieved from World Bank’s World Development Indicators Database.