Bureaucrats implement policy. How important are they for a state's productivity? And do the tradeoffs between policies depend on their effectiveness? Using data on 16 million public purchases in Russia, we show that 39 percent of the variation in prices paid for narrowly defined items is due to the individual bureaucrats and organizations who manage procurement. Low-price buyers also display higher spending quality. Theory suggests that such differences in effectiveness can be pivotal for policy design. To illustrate, we show that a common one-bid preferences for domestic suppliers-substantially improves procurement performance, but only when implemented by ineffective bureaucrats.

This paper investigates the role that sanitation plays in upholding the social contract, whereby citizens pay taxes in exchange for governments providing goods and services. We study the case of Manaus, Brazil, where sewer connections vary considerably across the city and property taxes are calculated in a presumptive manner that does not account for a household's access to sanitation. We find that households with access to the city sewer system are significantly more likely to pay their property tax, relative to households that only have access to latrines or lack access to improved sanitation entirely. Our evidence is consistent with a role for the social contract in this decision, as households with sewer systems are more likely to have positive attitudes towards the municipal government.

We design a field experiment to study how the allocation of authority between frontline procurement officers and their monitors affects performance both directly and through the response to incentives. In collaboration with the government of Punjab, Pakistan, we shift authority from monitors to procurement officers and introduce financial incentives in a sample of 600 procurement officers in 26 districts. We find that autonomy alone reduces prices by 9% without reducing quality and that the effect is stronger when the monitor tends to delay approvals for purchases until the end of the fiscal year. In contrast, the effect of performance pay is muted, except when agents face a monitor who does not delay approvals. Time use data reveal agents' responses vary along the same margin: autonomy increases the time devoted to procurement and this leads to lower prices only when monitors cause delays. By contrast, incentives work when monitors do not cause delays. The results illustrate that organizational design and anti-corruption policies must balance agency issues at different levels of the hierarchy.

Using a novel source of quasi-experimental variation in interest rates, we develop a new approach to estimating the Elasticity of Intertemporal Substitution (EIS). In the U.K., the mortgage interest rate features discrete jumps-notches-at thresholds for the loan-to-value (LTV) ratio. These notches generate large bunching below the critical LTV thresholds and missing mass above them. We develop a dynamic model that links these empirical moments to the underlying structural EIS. The average EIS is small, around 0.1, and quite homogeneous in the population. This finding is robust to structural assumptions and can allow for uncertainty, a wide range of risk preferences, portfolio reallocation, liquidity constraints, present bias, and optimization frictions. Our findings have implications for the numerous calibration studies that rely on larger values of the EIS.

We investigate housing market responses to transaction taxes using administrative data on all property transactions in the UK from 2004-2012 combined with quasi-experimental variation from tax notches and tax stimulus. We present two main findings. First, transaction taxes are highly distortionary across a range of margins, causing large distortions to the price, volume and timing of property transactions. Second, temporary transaction tax cuts are an enormously effective form of fiscal stimulus. A temporary elimination of a 1% transaction tax increased housing market activity by 20% in the short run (due to both timing and extensive responses) and less than half of the stimulus effect was reversed after the tax was reintroduced (due to re-timing). Because of the complementarities between moving house and consumer spending, these stimulus effects translate into extra spending per dollar of tax cut equal to about 1. We interpret our empirical findings in the context of a housing model with downpayment constraints in which leverage amplifies the effects of transaction taxes.

To fight evasion, many developing countries use production-inefficient tax policies. This includes minimum tax schemes whereby firms are taxed on either profits or turnover, depending on which tax liability is larger. Such schemes create nonstandard kink points, which allow for eliciting evasion responses to switches between profit and turnover taxes using a bunching approach. Using administrative data on corporations in Pakistan, we estimate that turnover taxes reduce evasion by up to 60-70 percent of corporate income. Incorporating this in a calibrated optimal tax model, we find that switching from profit to turnover taxation increases revenue by 74 percent without reducing aggregate profits.

Measuring the performance of government agencies is notoriously hard due to the lack of comparable data. At the same time, governments around the world generate an immense amount of data that detail their day-to-day operations. In this chapter we focus on three functions of government that represent the bulk of their operations and that are fairly standardized: social security programs, public procurement, and tax collection. We discuss how public sector organizations can use existing administrative case data and re-purpose them to construct objective measures of performance. We argue that it is paramount to compare cases that are homogeneous or construct a metric that captures the complexity of the case. We also argue that the metrics of government performance should capture both the volumes of services provided as well as their quality. With these considerations in mind, case data can be at the core of a diagnostic system with the potential to transform the speed and quality of public service delivery.

Governments frequently use proxies for deservingness - tags - to implement progressive tax and transfer policies. These proxies are often imperfect, leading to misclassification and inequities among equally deserving individuals. This paper studies the efficiency effects of such misclassification in the context of the property tax system in Manaus, Brazil. We leverage quasi-experimental variation in inequity generated by the boundaries of geographic sectors used to compute tax liabilities and a large tax reform in a series of augmented boundary discontinuity designs. We find that inequities significantly reduce compliance. The elasticity of compliance with respect to inequity is between 0.12 and 0.25, accounting for half of the observed change in compliance. A simple model of presumptive property taxation shows how mistagging affects the optimal tax schedule, highlighting the opposite implications of responses to the level of taxation and to inequity for optimal tax progressivity. Interpreting our findings through the lens of the model implies that optimal progressivity is around 50% lower than it would be absent inequity responses. These results highlight the importance of inequity for public policy design, especially in contexts with low fiscal capacity.

This paper exploits employee-employer matched administrative tax data on firms and salaried workers in Pakistan to explore the underappreciated role of firms in determining how workers' taxable earnings respond to taxation. I present evidence on three ways in which firms affect workers' earnings responses. First, third-party reporting of salaries by employers makes underreporting taxable income more costly for workers and reduces evasion of the income tax. Second, firms' equilibrium salary-hours offers respond endogenously to the presence of adjustment costs in the labour market by tailoring offers to aggregate worker preferences. Third, workers learn about the tax schedule from firms' salary offers, making them more responsive to taxation both contemporaneously (by 130%) and in subsequent years (by 100%). However, while third-party reporting makes misreporting more costly, it does not eliminate it in a low tax-capacity setting: 19% of workers still underreport their salaries, leading to a loss of about 5% of tax revenue, and indicating high returns to investments in improving enforcement capacity. The large role played by firms in determining workers' earnings implies that firms need to play a central role in our analysis of income taxation in lower income countries.

The creation of a tax system that raises revenues efficiently and equitably is one of the central challenges in economic development. Our project combines the new theory and a series of three Randomized Controlled Trials (RCTs) in Paraguay to provide new insights into tax evasion by firms, enforcement spillovers through production networks, and the optimal targeting of tax enforcement activities. We develop a simple, scalable tool to detect potential misreporting in the Value Added Tax (VAT). The first experiment (8/22–12/23) estimates the direct effects of notifying taxpayers of detected misreporting on their tax compliance. The second experiment (12/24) estimates the indirect effects on notified taxpayers' suppliers and clients. The third experiment (4/25) tests the optimal targeting rule, contrasting status quo targeting based on direct effects; theory-driven optimal targeting based on our network evasion model; data-driven optimal targeting based on the results of the first two experiments; and random targeting.

This project aims to trace out the contours of politically feasible property tax reforms through a series of two Randomized Control Trials (RCTs). The first experiment (10/2024-2/2025) measures citizens' preferences for the level and progressivity of property taxes and their determinants. A series of information experiments studies the impacts of tax knowledge, public goods, tax evasion, government waste, and elite capture on citizens' preferences. The second experiment (planned for 4/2025) measures political and bureaucratic decision makers' preferences and responsivenes to different subgroups of citizens' demands. Combining the results from the two experiments, the project will provide a roadmap for feasible property tax reform in Punjab, Pakistan.

We use administrative data and a triage experiment with bureaucrats in Peru to study disparate treatment in the handling of citizen whistleblower allegations of corruption. We show that institutional decisions feature only limited disparate treatment, despite bureaucrats' preferences exhibiting bias against defendants with indigenous names. Through our triage experiment we are able to observe downstream outcomes of cases while experimentally manipulating bureaucrats' available information to decompose the sources of biases. We find that bureaucrats exhibit substantial preference-based bias against indigenous defendants, and that they statistically discriminate in favor of indigenous defendants, though not sufficiently to offset their preference-based bias, leading to an overall bias against indigenous defendants.

This paper explores a unique market-wide view of public procurement using administrative data generated by the mandatory use of electronic invoicing in Brazil to study the São Paulo state procurement of pharmaceutical products. This data allows us to observe private sector transactions by government suppliers, as well as by firms that are not government suppliers but do transact goods purchased by the government. Thus, we can benchmark Business-to-Government (B2G) against Business-to-Business (B2B) transactions, identify the pool of potential suppliers, and determine if and when the government accesses suppliers with better prices. We begin by leveraging recent advances in Natural Language Processing to classify products based on the free-text descriptions in invoices. Then, we describe the circumstances under which governments pay more (or less) for goods in the pharmaceutical market. On average, we find that the government pays 13.6% more than the private sector for the same goods. However, when controlling for supplier fixed effects, this difference drops to -8%, indicating that while the government selects higher-priced suppliers, it successfully negotiates lower prices from those contracted. This suggests that obtaining better value-for-money depends largely on the government's access to more competitive sellers. We also find substantial heterogeneity across products and some evidence that a larger government presence in a market is associated with a smaller supplier selection issue and greater price advantages in public versus private purchases from the same supplier.

We exploit a program of randomized audits covering the entire population of VAT filers from Pakistan to study how much evasion audit detects and how much evasion it deters by changing behavior. We document substantial evasion at the baseline: almost one-third of firms engage in tax evasion, and conditional on some evasion, the average evasion rate exceeds 40 percent. We find remarkable heterogeneity in evasion by firm size with the evaded amount exceeding the reported liability in the bottom three quartiles but is merely 7 percent in the top. Despite detecting substantial liabilities, audit does not deter tax evasion. Examining more than ten outcomes, we detect no effect of audit on proximate or distant behavior. We offer an explanation of the detection-without-deterrence result and discuss its optimal policy implications.

The literature on optimal income taxation assumes that wage rates are generated exogenously by innate ability and therefore do not respond to behavior and taxation. This is in stark contrast to a large empirical literature documenting a strong effect of current work effort on future wage rates. We extend the canonical Mirrleesian optimal tax framework to incorporate such career effects and provide analytical characterizations that depend on estimable entities. Besides the standard static earnings elasticity with respect to the marginal tax rate, the optimal tax schedule also depends on the elasticity of future wages with respect to current work effort. We explore the empirical magnitude of this "career elasticity" in a meta-analysis of the literature on the returns to work experience and tenure, concluding that a reasonable value for this elasticity lies between 0.2 and 0.4. Calibrating the model to US micro data (under reasonable values of the career elasticity), we present numerical simulations of optimal nonlinear tax schedules that depend on per-period earnings and potentially on age. In the case of age-independent taxation, the presence of career effects make the tax schedule substantially less progressive than in standard models with exogenous wage rates. In the case of age-dependent taxation, career effects create a strong argument for lower taxes on the old, opposite the recommendation in the recent literature on age-dependent taxation. This result reflects both a career incentive effect and an equity effect, where the latter effect arises because increasing earnings over the career path for each ability level imply that, conditional on earnings, age and ability are negatively correlated.