A growing body of literature finds that the election of criminal or corrupt politicians often leads to large negative economic costs. Less is known about what these politicians do to the government programs they are responsible for delivering. This paper examines this question with data from India’s largest rural workforce program. Using a regression discontinuity design, I find that in constituencies where a criminally accused politician won, the number of projects completed falls by about 61%, while the allocation of workdays increases by about 38%. Program funds in criminal constituencies are disproportionately allocated to labour, rather than materials. The results further show that the increase in work allocation is concentrated in constituencies where criminal politicians seek re-election. These findings suggest that criminally accused politicians redirect their efforts toward providing visible target benefits by distorting program delivery.

Can expanding women’s access to work compromise their safety? We study Indian states’ removal of factory night-work bans using data from 2008–2024. Female employment rises by 23 percent and recorded crimes against women by 22 percent, with no corresponding increase in unrelated crimes. Recorded domestic and non-domestic violence both increase. Survey-reported sexual violence by non-partners also increases, while the rise in recorded domestic abuse may reflect greater reporting. Reforms expanding women’s economic opportunities can thus have unintended safety consequences.

We investigate how political corruption affects citizens’ willingness to disclose tax evasion. We conducted a survey experiment with 1,200 respondents in Bangalore, India, combining corruption vignettes and list experiments. Respondents were randomly presented with hypothetical candidates whose attributes varied along three dimensions: (a) alleged honesty versus corruption; (b) prioritization of infrastructure versus other public spending; and (c) political party affiliation. Contrary to conventional expectations, exposure to corruption cues reduced respondents’ likelihood of disclosing tax evasion, both for themselves and for their community. We further find that the type of public spending does not affect tax evasion reporting, and that respondents with moderate preferences over different tax structures are less supportive of financing public spending through direct taxation. Corruption cues neither increase the willingness to justify evasion nor affect tax morale. Overall, our results suggest that respondents distance themselves from admitting illegal behaviour when corruption is made salient.

Policies expanding police discretionary authority are widespread and politically contested, yet their effects on police behaviour, institutional trust, and political preferences remain poorly understood. We study Britain's 2014 Anti-Social Behaviour, Crime and Policing Act, which authorised local councils to implement Public Space Protection Orders (PSPOs), prohibiting broadly defined anti-social activities and expanding officer discretion in designated areas. Using the first comprehensive dataset on PSPO implementation across London census tracts (2016–2024), linked to crime records, stop-and-search data, and electoral outcomes, we exploit PSPOs' staggered adoption to identify causal effects. These policies increase unproductive police stop-and-search activity, disproportionately targeting non-white residents. Reported anti-social behaviour and other crimes remain unaffected. These findings suggest that vaguely defined discretionary enforcement tools can generate racially unequal policing without delivering the public order benefits that justify their adoption.

Do dynastic politicians exert less political effort than their non-dynastic counterparts? Using a pre-registered field experiment with 4,608 national and state legislators in India and a regression discontinuity design based on close elections, this paper tests whether legislators’ political family connections affect their responsiveness to constituency service requests. We find that dynastic legislators are 12.4 percentage points more likely to respond than non-dynastic legislators. This difference does not depend on the partisan alignment of the sender or on the type of good requested. It is instead concentrated among dynasts whose families have not previously won the seats they represent, and among legislators facing an election within the next year. Using administrative data, we further show that a larger share of registered households receive work under India’s rural employment guarantee program in constituencies that elect dynastic legislators. However, dynastic legislators exhibit lower legislative engagement, attending a smaller share of legislative sittings. These findings suggest that dynastic legislators devote more effort to aspects of their representative role that are directly visible to voters and less to legislative activity.