Abstract
Elected politicians are instrumental in providing public goods to their constituencies. On one hand, dishonest politicians can expropriate public funds for personal use, thus reducing the funds available for public goods provision. On the other hand, through clientelistic and patronage politics, dishonest politicians can give appropriated funds back to citizens.
In the first chapter, I study how dishonest politicians affect public goods provision in their constituencies. To identify dishonest politicians, I determine whether they evaded income taxes conditional on their minimum earnings and occupation, using a unique dataset based on the asset disclosure and tax forms candidates submit prior to elections in Bangladesh. Directly comparing constituencies represented by dishonest and honest politicians will not reveal the causal effects on public goods provision and economic development, since constituencies that elect dishonest politicians may systematically differ from the ones that do not. I rely on close elections, in which a dishonest politician narrowly defeats an honest one, and a regression discontinuity design to examine the effects of dishonest politicians on public goods provision and economic progress. Between 2009-10 and 2014-15, I find that in sub-districts that narrowly elected dishonest politicians, 27.3% fewer households received social safety net benefits compared with sub-districts that narrowly elected honest politicians. To analyze the effect between 2014-15 and 2019-20, I use a number of health and infrastructure variables at the sub-district level to develop an index of public goods provision using principal component analysis. I find that constituencies with dishonest leaders have a 0.74 standard deviation lower index value than constituencies with honest leaders. Results are quantitatively similar under the choice of different bandwidths and are robust to various specifications. Furthermore, categorizing sub-districts by wealthy dishonest versus wealthy honest leaders, measured by politicians’ total assets above the mean value in 2014 - compare long-term economic development, proxied by growth of nighttime light brightness. I find that sub-districts with wealthy dishonest leaders have 5.75% point lower growth in night-time light brightness. Using my own estimates to convert night-time light brightness to GDP growth, I find 0.94% lower yearly GDP growth per sub-district under wealthy dishonest leaders.
In the second chapter, to understand the mechanism, I provide evidences from sub-district-level budgets showing statistically significant lower constituency-wide expenditure by dishonest politicians. I found local-level tax revenues do not differ significantly between upazilas with honest versus dishonest politicians. I also show empirically that public goods provision does not depend on the ability of politicians, as measured by years of education, and politicians’ honesty and ability are independent of each other. Furthermore, I provide evidence that the central government resources allocation also does not differ significantly, it suggests that honest politicians are not favored by the central government. I provide further evidence that dishonest and honest politicians' occupation choices are not statistically different. Lastly, I show that re-election probability of dishonest politicians also does not influence public goods provision differentials.
Third chapter is a conjoint work with Abu S. Shonchoy (Florida International University) and Towhid I. Mahmood (Texas Tech University) on public health concerns in Bangladesh arising from infectious diseases. Bangladesh is among the top fifteen SARS-CoV-2 (COVID-19) affected countries in the world. However, the country has the lowest testing capacity per million in that group. Faced with growing pressure to continue livelihoods, Bangladesh government lifted the lockdown abruptly in 2020, costing an immediate surge in the virus caseload. Against this backdrop, there is a dire need to derive data-driven planning, for mitigation and management of COVID-19 cases in Bangladesh – prioritizing the efficient allocation of limited resources. Utilizing publicly available and administrative data, this paper introduces a contagion risk (CR) index, which can work as a credible proxy to detect potential virus hotspots – aiding policymakers with proper planning. Grounded on disease spreadability vectors, we derived the CR-Index at the district level, based on nine variables across five domains: socio-economy, demography, occupation, migration, and health infrastructure. The CR-Index is validated against the district wise COVID-19 cases across the study period. CR-Index is positively correlated with district-wise COVID-19 cases across the pandemic period (average correlation is 0.65, p-value 0.001). We found that one percent increase in the CR-index predicts a 3.8 percent daily change in the increase of COVID-19 cases across districts. The proposed CR-Index can predict seven out of the top ten COVID-19 caseload districts of Bangladesh.
