Who Bears the Cost of Corruption?

Corruption is costly for developing nations’ economies, but scholars and policy makers do not fully understand who bears the cost of corruption. We conduct a field experiment in Malawi to determine the impacts of political connections, socioeconomic status, and shared ethnicity on the degree of corruption a citizen encounters across two contexts – police roadblocks and electricity service offices – that vary the risk borne by corrupt officials. By disentangling wealth and political power, we show that political connections reduce exposure to corruption, while wealth only insulates one from corruption when it is seen as a proxy for political power. Coethnicty with a corrupt official increases corruption, but only under high risk, since trust in such contexts is crucial. These findings indicate that officials make strategic decisions about when to engage in corruption, which leaves certain individuals, particularly the poor and politically powerless, disproportionately bearing the costs. Who Bears the Cost of Corruption? 1 Corruption, defined as “the use of public office for private gain,” is prevalent across societies worldwide. One of the most commonly studied manifestations of corruption is a public official accepting a bribe in exchange for providing a public good or service, or what Transparency International (2009) refers to as petty corruption: “everyday abuse of entrusted power by lowor mid-level public officials in their interactions with ordinary citizens, who are often trying to access basic goods or services in places like hospitals, schools, police departments, and other agencies.” It is well established that corruption is costly for developing nations’ economies by introducing inefficiency into an already disadvantaged system (Bardhan 1997). And while we also know that such corruption is more common in countries with greater inequality (Jong-Sung and Khagram 2005), we do not yet fully understand which sectors of society bear the costs of corruption and whether the costs are distributed equally. On the one hand, several observational studies demonstrate that wealthier individuals are more likely to encounter corruption, partially because they interact with officials more often than do the poor (Hunt 2007; Hunt and Laszlo 2012). On the other hand, there is also ample evidence that in unequal societies the rich and powerful are better able to insulate themselves and their wealth from corruption (Nielsen 2006). Using a field experimental approach in which socioeconomic status is randomly assigned to research assistants making illegal left turns, Fried, Lagunes, and Venkataramani (2010) find that poor drivers in Mexico paid bribes to Traffic Police more often than rich drivers, and that their total corruption burden was larger, even though the average bribe payment demanded from rich drivers was higher. One of the reasons it is so difficult to interpret corruption patterns across socioeconomic strata is that wealth in highly unequal societies sends two different signals to corrupt officials. Visible wealth suggests the ability to pay, making rich individuals particularly valuable targets for corruption by allowing corrupt officials to extract more money per interaction. However, wealth also serves as a strong indicator of political connections, making wealthy individuals particularly risky targets for corrupt officials who fear repercussions for targeting Who Bears the Cost of Corruption? 2 the powerful. In the Fried, Lagunes, and Venkataramani (2010) study, for example, the authors attribute the finding that the rich are less frequently asked for a bribe to the fact that “officers associate wealth with the capacity to exact retribution and therefore are more likely to demand bribes from poorer individuals.” The impact of wealth on vulnerability to corruption is difficult to determine because conspicuous wealth conflates socioeconomic status with political power. We attempt to disentangle these two characteristics experimentally. We also consider the role of ethnicity in shaping corruption patterns. Past research has shown that corruption is more prevalent in ethnically diverse societies (Alesina et al. 2003), but the mechanism behind this finding is not clear. Others show that shared ethnicity facilitates cooperation (Habyarimana et al. 2009), which could conceivably reduce bribery extracted from coethnics. On the other hand, coethnics tend to be trusted at higher rates than non-coethnics (Fershtman and Gneezy 2001; Robinson 2015), which would suggest corruption among coethnics is more likely, particularly in high-risk situations. To study the effects of wealth, political power, and coethnicity on corruption, we carry out a field experiment in which confederate researchers interact with two different types of public officials in Malawi, a country with widespread low-level corruption. First, research assistants pass through police roadblocks without displaying evidence of insurance coverageSecond, research assistants attempted to receive expedited service for a residential electricity connection at offices of the Electricity Supply Corporation of Malawi (ESCOM). The critical difference between the two contexts is that the risks of detection, retribution, and reneging are much higher for ESCOM officials than the Traffic Police. To estimate the impact of socioeconomic status, research assistants were randomly assigned to high or low socioeconomic conditions, conveyed to officials through attire, stated occupation and vehicle (Traffic Police context only). To estimate the impact of political connections separately from wealth, we independently assigned research assistants to be either political connected, conveyed by wearing a ruling party button and demeanor associated with power in the Malawian context, or not. Finally, the ethnicity of each official was coded Who Bears the Cost of Corruption? 3 by research assistants, and coethnicity was randomly assigned through driving route. We find that randomly assigned political connections result in lower rates of bribery, lower bribe amounts, and lower overall cost when interacting with the Traffic Police. In the ESCOM context, the politically connected were not only subject to lower rates of bribery, but were actually given preferential treatment without requiring a bribe. In contrast, we find that wealth has very little impact on exposure to corruption except when direct information about political connections is absent. Wealth did decrease the total amount paid per Traffic Police roadblock, but only because vehicles signaling wealth (and, thus, power) were less likely to be stopped. In the ESCOM context, wealth has no independent effect after controlling for political connections. These findings suggest that wealthy individuals’ ability to shield themselves from corruption may be due to the conflation of wealth and political power. Finally, we expected that shared ethnicity would have different effects across the two contexts given the differential risks to corrupt officials. In the lower-risk Traffic Police context, we expected in-group favoritism to reduce the corruption burden for officers’ coethnics. In contrast, in the higher-risk ESCOM context, we anticipated that reliance on coethnicitybased trust would counteract coethnic favoritism, resulting in higher rates of bribery among coethnics. While we find no impact of shared ethnicity on corrupt interactions with Traffic Police, coethnicity does indeed increase exposure to corruption in the ESCOM context. Together, these results indicate that there is an unequal distribution of the costs of corruption in Malawian society. This inequality is magnified by the fact that those marginalized by corrupt practices the poor and unconnected are the same as those marginalized in Malawian society more broadly. The findings offer three important contributions. First, we disentangle the effects of wealth and power on exposure to corruption, showing that wealth protects only when it serves as a proxy for political power. Second, we identify another dimension of discrimination for corruption; coethnics are targeted by corrupt officials in risky contexts. Third, we show corruption prevalence and patterns of discrimination vary across contexts based on corrupt officials’ strategic responses to the level of risk they face. Who Bears the Cost of Corruption? 4