Link to working paper

Registering land with the state may provide landowners with legal protections but may also expose them to expropriation. This paper studies how that tradeoff shaped land formalization in post-independence Tunisia. Using text-as-geospatial-data methods, I chart the universe of citizen-initiated requests for land registration between 1956 and 1980 and support my large-N data with extensive qualitative archival research. I argue that registration was prioritized if land held significant fixed capital investment and if there were threatening private neighbors.  Registration protects land from unscrupulous neighbors, but offers little protection from the state itself. Registration was more likely and occurred sooner for land with tree crops (a long-term investment) and for land with potentially ambiguous boundaries. Conversely, registration was delayed for properties adjacent to or owned by the state or by family members. Modeling diffusion between direct neighbors, I show that properties are especially likely to be registered following a neighboring property with similar land use. A stacked difference-in-difference event study provides further support for registration spillovers to properties with the same land use type. Finally, I leverage a temporal shock of a high profile state seizure of farmland. With the threat of state expropriation suddenly increased, land registration dropped precipitously and remained low given the prospect of forced collectivization of agriculture. Insecurity can drive “demand” for the state and public compliance, but only when the public trusts state authority.

Link to working paper

 Why are some populations more or less visible to the state? I argue that state discrimination reduces the legibility of disfavored communities through two channels: demand mechanisms, where mistrust discourages voluntary registration, and supply mechanisms, where the civil service is selectively inaccessible. This paper examines political sources of uneven population legibility, focusing on birth registration in Côte d'Ivoire before, during, and after the Ivorian Civil War (2002–2011). Drawing on five household survey datasets and postwar census data, I use a triple-difference design to show the emergence of significant variation in birth registration inequality across the wartime line of control, which I trace to discrimination and exclusion during the war. Migrant communities faced declining registration under the anti-migrant Gbagbo regime, while enjoying a registration advantage in rebel-controlled territory. Mechanism tests provide suggestive evidence that supply-side exclusion predominated in rebel zones, while both supply and demand dynamics drove inequality under regime regions. After the war, judicious retroactive enumeration corrected wartime inequalities. This paper emphasizes how the extent and unevenness of vital records registration are political choices.

Land insecurity is a persistent feature of agrarian economies across the Global South, yet formal land documentation remains rare even where conflicts over property are common.  In Côte d'Ivoire, the state has limited vision of the domain and landowners have few state protections faced with an intense recent history of land conflict. I argue that documentation enables calling on the state and will be most useful when non-state institutions are weak or unreliable. This paper leverages  household survey data collected before and after a period of more intense violence to estimate a difference-in-difference by proximity to violence. I support this with qualitative evidence from Ivorian government reports and media accounts of violence. I find that exposure to violence increases reported land insecurity and individual (rather than collective) ownership. Violence has an overall null effect on documentation, but with considerable heterogeneity. Violence exposure increases land documentation in the most diverse communities and among internal migrants. State arbitration of property disputes is more needed when customary authorities are weak or untrustworthy. For local minority groups (as well as households directly victimized), though, violence appears to reduce the rate of documentation.

With Daniel Markovits and Andrew Thompson

Under Review

Correcting false and negative beliefs about political opponents has shown promise in reducing anti-democratic attitudes and polarization. Despite the simple nature of such corrections, there is little extant evidence that they are effective beyond immediately administered survey outcomes and it is unclear which voters opt-in to such interventions. To test these mechanisms, we worked with a partner organizations to implement a depolarization initiative that bundled factual belief corrections with elites modeling civil disagreement. We recruited an online panel of 3,461 eligible respondents and then randomized an offer to attend a 30 minute depolarization event in which bipartisan elites defended democratic values and discussed polling information suggesting mass commitment to democracy across party lines. We report two main sets of findings. First, despite generous financial incentives, there was substantial differential compliance by partisanship, though not by pre-treatment attitudinal measures of affective polarization or anti-democratic attitudes. Second, our intervention achieved a durable reduction in beliefs that opponents were opposed to democracy (measured at 1 week and 2 months post-event) and in willingness to attend future depolarization events. However, we found no reduction in anti-democratic attitudes across many pre-registered outcomes. We use a follow-up survey experiment to explore the mechanisms underlying both of these findings.

The causal inference revolution in the social sciences has increased concerns over the external validity of findings.  While inference may be internally valid, it may be estimating a quantity that is difficult to interpret. This paper develops an approach for identifying heterogeneous effects under treatment non-compliance by estimating local average treatment effects for multiple types of compliers. Under a design with multiple gradated levels of encouragement, we can conceptualize complier type as a continuous latent variable representing the minimum amount of encouragement needed to take treatment. This approach requires a monotonicity assumption that increases in encouragement cannot decrease treatment uptake. If we make an even stronger monotonicity assumption that treatment effects are strictly monotonic with respect to compliance type, I show that we could place partial bounds on the treatment effects for the subsample of non-compliers. I use these methods to re-analyze outcomes in two public health interventions that randomize the level of subsidy provided for goods. I find strong evidence of heterogeneity by willingness-to-pay, but do not find support for the strong monotonicity assumption needed for bounding effects among non-compliers. The paper concludes discussing how heterogeneous effect estimation by compliance type can estimate the average effect of an intervention with different encouragement provided. If willingness-to-pay predicts stronger treatment effects, it suggests the market may be sufficient for a socially efficient distribution of the treatment. If the reverse is true, it suggests the necessity for continued subsidization.