Published on Development Impact

What do we know about tax-evading politicians and their effect on public goods provision and economic development? Guest post by Moogdho Mahzab

This page in:

This is the 16th in our series of post of students on the job market this year

The “honesty” of a politician standing for elected office is always a carefully scrutinized and debated issue in every election across the globe, including the recently concluded US elections. Despite this, dishonest politicians get re-elected and prior research attributes this to strong clientelism or patronage politics. However, there is relatively scarce evidence on how the election of a dishonest politician affects public goods provision and more broadly, economic development. This question is particularly salient in developing countries, where local elected representatives wield great discretion in provision of public goods, and institutions to keep them in check are weak. The impact of electing a dishonest politician on economic development is not obvious for two reasons: first, dishonest politicians may redistribute appropriated public funds to their constituents privately, and second, they may be more competent than honest one, thus more adept at managing and manipulating the system, and hence better at resource allocation despite appropriating funds. Moreover, identifying the causal effect of a dishonest politician is challenging, because constituencies that elect dishonest and honest politicians may be systematically different from each other in many ways. My job market paper examines these issues in the context of sub-district (“upazila”) elections in Bangladesh. These elections, held across all 491 upazilas in 2009 and 2014, elected a chairman for a period of five years.


Measuring “honesty”

An important reform in 2009 mandated every candidate running for public office to publicly disclose their income, assets, education, criminal background, and information on dependents, along with tax-return forms, prior to the election. This law therefore enables me with a direct way of measuring the “honesty” of politicians as follows: the disclosure of affidavits provides me with the income and assets reported by politicians, and hence I can calculate what the politician should be paying in taxes. Using tax-returns forms (returned earlier in the year), I then examine how much he actually pays. I categorize a politician as “dishonest” if he evaded income taxes than what he should be paying based on his income bracket and conditional on occupations (I also check for the accuracy of income and tax reports by examining under-reporting and evasion of taxes by income diversion to dependents). Based on this definition, 28% of upazila chairmen are tax-evaders in Bangladesh.


How different are dishonest politicians from honest ones?

I track 2009 (2014) election winners and runner ups in 2014 (2019), and find that on average the income of winners is Taka 1.52 million higher than that of runners-up, perhaps indicating that winners earn more through holding public office (as Fisman et al. 2014 find in the Indian context). More importantly, I find that dishonest winners have Taka 9.97 million more in assets, which grew 3.6% more than honest winners over the five-year tenure term. There is no statistical difference in education between honest and dishonest politicians (13.78 and 13.57 years of schooling respectively). Also, there is no statistical difference in terms of numbers of past criminal records; however, dishonest politicians have 10.8% more ongoing law-suits than honest politicians at the time of the elections.


Do dishonest politicians provide less public goods? Yes.

As discussed previously, causally identifying the effect of a dishonest politician is hard. To make progress, I rely on a strategy used by political scientists –using a regression discontinuity by focusing on elections where dishonest politicians win by a narrow margin in a “close election” (e.g. Lee 2008).  Strong competition in local government elections is indicated by the fact that 35.1%, 60%, and 82.2% upazila elections have under 5%, 10%, and 20% winning vote margins, respectively.


Social safety net programs (four main categories: special allowances to tackle poverty; employment generations through micro-credit initiatives; food security-based activities; health, education and training programs) serve as an excellent proxy for measuring public goods, since the upazila chairman is the front-line elected official directly responsible for their provision. To analyze the effect of dishonest politicians from 2009-10 to 2014-15, I use households’ inclusion to these programs as the main outcome measure.  Between 2015 and 2019, I use a combination of household surveys, phone surveys and administrative data; I construct a public good index for the provision of health and infrastructure services using number of community clinics, number of sanctioned beds in health complexes, average length of stay of a patient and number of self-reported Covid-19 patients, and growth of paved roads for each upazila to measure the effects of electing honest politicians; no information is available on social safety net programs after 2015.


Figure 1 shows the effect of electing an honest politician. To the right of the cutoff are upazilas where an honest politician wins, while to the left of the cutoffs are those where he narrowly loses to a dishonest one. The top panel shows that, on average, 12 fewer households (27.7% fewer households) yearly receive social safety net programs in upazilas with dishonest winners compared to upazilas with honest winners between 2010 and 2015. The bottom panel shows that upazilas with dishonest leaders have 0.94 lower index value, which is 0.74 standard deviation lower than upazilas with honest leaders. Comparing the means, this translates to 125% lower index value in upazilas with dishonest leaders.


For a smaller sample of upazilas, I use a survey on their revenues and expenditure and find that upazilas with dishonest politicians do not have difference in local revenue collection, but 10% lower development expenditure. 



graph shows better outcomes under honest politicans


Figure 1: Effect of electing honest politicians

Do dishonest politicians affect overall economic development? Not much.

As a measure of economic development, I used night-time light brightness, which is commonly used in the literature as a proxy for economic activity and development (Henderson et al. 2012).  Using the same strategy as above (comparing the narrow dishonest winners vs losers), I find no effect on the growth of night-time light. More generally however, when I compare sub-districts with above-median dishonest leaders and honest leaders (wealthy dishonest versus wealthy honest leaders)—I find that upazilas with wealthy dishonest leaders have 5.75pp lower growth of night-time light brightness. Using a back-of-the-envelope elasticity of night-time light growth in Bangladesh to yearly GDP growth, this translates into a 0.94 pp lower annual GDP growth per upazila.


What mechanisms could drive these results?

Are dishonest politicians also more able at their job? One mechanism that can rationalize the above results is that honest politicians are more able at their jobs i.e there is a correlation between ability and honesty. Using a politician’s education as a proxy for his ability (as is common in the political science literature), I show that there is no such correlation. Applying similar regression discontinuity design, with years of education as measure of ability, I find no effect on public goods provision. 


Do honest politicians get more funds from Central Govt? Using data from the Ministry of Finance, I show that central fund allocation and the use of it does not vary significantly across upazilas with dishonest and honest leaders. This implies that the results cannot be explained by differential treatment by the central government.


Concluding remarks

Honest politicians-led upazilas received better provision of social safety net programs from 2010 to 2014 and also had higher public goods index of health and infrastructural services between 2015 and 2019. There are no significant differences between honest and dishonest politicians in terms of ability (measured by education), occupations, and allocation of central development funds. The negative effects on public goods provision in dishonest politicians-led upazilas might be explained by the divergence in upazila-level spending. Upazilas with dishonest leaders, although not having any significant difference in revenue collection, had lower local-level development spending.



Moogdho Mahzab is a Ph.D. student in Economics at University of Virginia.

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000