Anti-corruption policy-making, discretionary power, and institutional quality: An experimental analysis

Abstract

We analyse policymakers' incentives to fight corruption under different institutional qualities. We find that 'public officials', even when non-corrupt, significantly distort anticorruption institutions by choosing a lower detection probability when this probability applies to their own actions (legal equality), compared to a setting where it does not (legal inequality). More surprising perhaps is the finding that policy-makers do not choose a zero level of detection on average, even when it applies to them too. Finally, corruption is significantly lower when the detection probability is exogenously set, suggesting that the institutional power to choose detection can itself be corruptive.

Punishment and corruption

“…Accordingly, when all punishment conditions were compared with the control condition, the interaction of bribe size with punishment was significantly negative, t(471.3) = -2.30, p = .022, b = -0.114, 95% CI [-0.212, -0.017] (see Figure 3). That is, the effect of punishment was present only for high bribes, t(488.2) = -2.25, p = .025, b = -0.099, 95% CI [-0.185, -0.013], 6 and there was no effect for low bribes, t(484.8) = -0.17, p = .865, b = -0.007, 95% CI [-0.082, 0.069]. 6 Note that the threshold for "high" bribes (130 and higher) was selected based on Figure 3 and the significance of the effect is not robust to the threshold selection even if the pattern of results is similar.…”

“…That is, the effect of punishment was present only for high bribes, t(488.2) = -2.25, p = .025, b = -0.099, 95% CI [-0.185, -0.013], 6 and there was no effect for low bribes, t(484.8) = -0.17, p = .865, b = -0.007, 95% CI [-0.082, 0.069]. 6 Note that the threshold for "high" bribes (130 and higher) was selected based on Figure 3 and the significance of the effect is not robust to the threshold selection even if the pattern of results is similar. The effect of punishment for high bribes is not significant when the threshold is 110 (p = .088) and 120 (p = .058), but it is significant for threshold 140 (p = .022) and 150 (p = .024).…”

“…Regardless of its specific nature, with increasing probability and size of punishment, the corrupt behavior is supposed to decline. While the general deterrence effect of punishment seems to be well established with observational (Goel & Rich, 1989) as well as with experimental data (Boly & Gillanders, 2018;Hanna, Bishop, Nadel, Scheffler, & Durlacher, 2011;Nagin, 2013), the interplay between the effects of increasing its severity and probability is less clear. Although Nagin (2013) argues that the probability of punishment 3 and not the severity of punishment serves as a deterrent, the prospect theory (Tversky & Kahneman, 1992) predicts that the relative effects of increasing the probability or size of punishment would depend on the starting point given that both the functions of subjective probability and subjective value of a loss are non-linear.…”

“…Although some of the explored approaches are less typical, such as staff-rotation (Ryvkin & Serra, 2012), competition among public officials (Shleifer & Vishny, 1993), anonymous public reporting (Ryvkin, Serra, & Tremewan, 2017) or ethics education (Banerjee & Mitra, 2018), most attempts at curbing corruption focus on the probability and size of a punishment (Abbink & Serra, 2012). Despite the amount of research interest in the effects of these two factors, the results remain mixed and inconclusive (Boly & Gillanders, 2018), possibly because many studies employ one-shot tasks and explore only a limited number of combinations of punishment probabilities and sizes.…”

Political Connections and Firm Pollution Behaviour: An Empirical Study

“…Therefore, increased costs for political connections inevitably result in remarkable changes in pollution discharges. In addition, many corrupt officials are caught during the anti-corruption campaign which inevitably influences the duration of political connections (Boly & Gillanders, 2018).…”