Results of a quantitative analysis
One of my recent research projects focused on analyzing the causal effect of corruption on tax revenues using cross-country time-series analyses.
Corruption is likely to have an adverse effect on the willingness of citizens to pay taxes. Citizens might perceive government officials being corrupt, their hard-earned tax money being used for corrupt purposes rather than to finance development projects or social programs. Furthermore, many taxes, such as income taxes, are evadable particularly in developing countries, where the government has low capacity to observe citizens’ income.
Individuals are more likely to evade income taxes when they think their tax money will be used for corrupt purposes instead of being quid pro quo. The effect that corruption perception can have on tax revenue collection is interesting, since citizens might find a moral ground to evade taxes. In this paper, I attempt to analyze the effect of corruption perception on tax revenue using an instrumental variable: Press freedom index.
Although some studies have proposed the idea that corruption might have a positive effect on tax revenue as it motivates tax officers to work harder, there is a general consensus that the presence of corruption reduces tax revenues (Fjeldstad and Tungodden, 2001). Tanzi and Davoodi (2000) investigated the relationship between levels of corruption perception and GDP, and found that a one-point increase in corruption perception index is associated with a 2.7% point decline in tax-GDP ratio.
Ghura (1998) conducted a study on 39 Sub-Saharan countries and found strong evidence that an increase in the level of corruption lowers tax revenue to GDP ratio. Thornton (2008) included 53 Middle Eastern and African countries in his analyses, and found a strong negative relationship between corruption perception and tax revenues.
As we know, correlation is not causation -- in order to find the causal effect of corruption on tax revenues, I required a variable that affects tax revenues indirectly only through the channel of corruption -- such a variable is called instrumental variable in econometrics. Having an instrumental variable that satisfies the relevant statistical conditions enabled me to obtain an estimate of the causal effect of corruption on tax revenues.
As my instrumental variable, I used press freedom index, because it is strongly correlated with corruption and satisfies the statistical conditions, provided I account for co-varying variables such as the country and time specific trends, real GDP per capita, tax rates and tax structure, income distribution, etc.
Literature has several studies illustrating the inverse relationship between increased press freedom and corruption. Brunetti and Weder (2003) showed that there is a strong association between the level of press freedom and the level of corruption in country-level data. Suphacahlasai (2005) found mass media competition and press freedom significantly reduce corruption. Lederman, Loayza, and Soares (2005) found that political institutions namely political stability, freedom of press, and democracies have a significant influence on reducing corruption.
For quantitative analyzes, my dataset consisted of a sample of more than 150 countries worldwide, and covered the years 2001 to 2015. I created this dataset by pooling data from different sources such as World Bank, Transparency International, Reporters without Borders, and several country-specific websites.
My first stage results suggest that increase in press freedom index by 1 (on a scale of 0 to 100), lowers corruption perception by 0.35% on a scale (of 0 to 100) and this effect is statically significant. A simple correlation between tax revenue to GDP ratio and corruption perception suggests when corruption perception is lowered by 1 (on a scale of 0 to 100), tax to GDP percentage increases by 0.18%, and this effect is statistically significant.
From the preferred specification -- coming to the main results -- an indication that when corruption perception lowers by 1 (on a scale of 0 to 100) tax revenue as a percentage of GDP increases by 0.35%, and this effect is statically significant and robust across all specification tests. The results have very important policy implications especially for a country like Bangladesh, where 45-65% of income in the economy remains untaxed according to a recent Finance Ministry survey.
Only 1.6-1.7 million submit tax returns amongst 3.1million people holding Tax Identification Numbers, whereas according to the economic think tank Centre for Policy Dialogue, there are 8 million taxable people in the country.
Having sufficient tax revenues raised without distorting incentives for citizens is vital to economic growth. Reducing corruption and corruption perception might be an important component of a program designed to increase tax base or revenues.
Maliha Ahmed previously worked in research at BIDS and BIGD. She is currently pursuing her PhD in Economics at the University of Illinois at Chicago.