This dissertation studies the economics of food regulation, with applications to food safety, third-party certification, and litigation. It contains 2 main parts. The first part of the dissertation contains two essays: the first essay studies the value of certification in signaling food safety; the second highlights the effectiveness of the common auditing system on firms' behavior change. In the second part, I study the impacts of food safety lawsuits on governmental regulatory behaviors and firms' food safety practices.

Third-party certification has been widely adopted in the food industry, but its credibility and effectiveness are controversial. In Part I, I assemble a comprehensive data set using plant inspection records from the U.S. Food Safety and Inspection Service (FSIS) and food safety certification results from both the Safe Quality Food (SQF) and the British Retail Consortium (BRC) global standards program. The first essay examines the role of third-party certification on certified plants' food safety practices and outcomes in the meat and poultry industry. I take the compliance rate of FSIS sanitary tasks as a measure of the level of food safety practices and the FSIS pathogen sampling test results as a measure of food safety outcomes. I find that certification ratings are informative of plants' food safety practices, but not their food safety outcomes. Leveraging the timing of initial and re-certification audits, the second essay finds SQF certified plants have a gradual increase in their average food safety practice level after initial audits, but not BRC certified plants. I do not find a significant impact of re-certification audits on food safety practices.

In Part II, I study the net impact of food safety lawsuits brought by those who claim harm on public food safety regulation intensity and defendant plants' food safety practices by combining a food safety lawsuit dataset that I developed with FSIS inspection records and exploiting information on the timing of lawsuit filing dates. I find that food safety lawsuits have a crowd-out effect on the regulatory intensity in inspection tasks that are directly relevant to the food safety disputed issues in the specific lawsuit. There is no evidence that food safety lawsuits have a statistically significant impact on defendant plants' food safety practices. The empirical result expands the knowledge of the interactions of plaintiffs' claims in litigation, public regulation, and plant behaviors in the context of food safety issues in the US meat and poultry industry.

Farmworkers are vulnerable to ambient environmental conditions, and an emerging health hazard is smoke from wildfires and agricultural burning. This smoke poses an immediate threat to the health and wellbeing of farmworkers by increasing the risk of injuries. Additionally, smoke can influence farmers' and farmworkers' decisions about when, where, and how much to work, potentially exacerbating ongoing labor shortages in the agricultural sector. My dissertation quantifies the effect of smoke from 1) wildfires and 2) agricultural burning on farmworker labor and health outcomes.

In the first chapter, we study the effect of wildfire smoke on farmworker labor outcomes in California. Using high-frequency individual-location data, we find that labor declines at both the extensive and intensive margins on days when fields are affected by wildfire smoke. On smoky days, the number of workers in a field is reduced by 17.51% and working hours are reduced by 23.12%, relative to days without smoke. Estimated effects are largest for labor-intensive crops. Farmworkers are more likely to be observed in a field immediately before smoke events and less likely to be observed after. They are also more likely to work in other fields when their primary worksite is treated. Results highlight the significant effects of wildfire smoke on farmworker labor outcomes, showing reductions in work activities and the adoption of substitution behaviors among a marginalized and hard-to-survey group.

In the second chapter, we study the impact of wildfire smoke on workplace injuries among agricultural workers, using workers'-compensation claims between 2007 and 2021. We find a substantial increase in agricultural-worker injuries attributable to wildfire smoke and smoke-induced PM2.5. Specifically, 10 micrograms per cubic meter increase in daily PM2.5 exposure from wildfire smoke increases traumatic injuries by 2.3 percent, and exposure to high levels of PM2.5 (above 20 micrograms per cubic meter) increases traumatic injuries by 14.29 percent relative to days without smoke. The effects of smoke are higher for young workers than for old workers. We find that injuries occur even at levels of PM2.5 considered safe. Our back-of-the-envelope calculation suggests that in 2020 alone, wildfire smoke was responsible for approximately 282 additional agricultural-worker injuries in California compared to a hypothetical scenario without smoke.

In the third chapter, using data from the major seven air districts in California engaged in agricultural burning and worker's compensation claims from 2000 to 2021, we investigate the impact of exposure to smoke from agricultural burning on agricultural workers' injuries. Agricultural burning has long been used for various purposes in U.S. agriculture, such as removing crop residue and controlling pests, resulting in significant emissions of air pollutants. Prior research has mainly focused on its effect on the general population particularly in developing countries. By leveraging daily changes in fire location and wind direction for identification, our findings show that on days with agricultural burnings, there is a 2.6 percent increase in injuries within the downwind zip code. The impact on farmworker injuries is larger when agricultural burnings occur over consecutive days.

This dissertation studies how participants and food retailers involved in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) respond to the incentives created by the program. WIC's fixed quantity structure, stocking requirements, and clinic visits reflect costs and benefits for participants and retailers. These structures reflect broader incentives in the provision of safety net programs through private firms as in SNAP and Medicaid. This dissertation applies current econometric methods for estimating staggered adoption designs with panel data and partial identification under selection to administrative data on WIC in different policy settings.The first essay examines how participants' use of their benefits changes if food retailers become unauthorized. US food assistance programs including WIC require that participants redeem benefits at authorized food retailers. Using a novel natural experiment in the WIC authorization of retailers, I find that a participant that loses access to an authorized retailer is less likely to participate in WIC and redeems a smaller share of their WIC benefits. These changes provide insight into the unintended consequences of food assistance program structure and how low-income individuals respond to changes in their food retail choice set. Essay two evaluates the effect of the switch to Electronic Benefit Transfer (EBT) on the authorization outcomes of WIC vendors. WIC EBT aims to reduce stigma and transaction costs for participants but may increase costs for some vendors. This paper combines novel administrative data from The Integrity Profile and the Store Tracking and Redemptions System with a new nationwide policy data on WIC EBT implementation. Using a staggered adoption difference-in-differences approach, we find the effects of the WIC EBT transition varied across different states. In sum, independent retailers are more likely to become unauthorized following WIC EBT implementation. The experience of the financial services provider contracted to implement WIC EBT by state may mediate the magnitude of the effect of EBT on vendor authorization outcomes. The third essay returns to the question of how participants in food assistance programs respond to changes in the set of authorized food retailers, in this case examining whether participants' health outcomes and program health objectives are affected by changes to the set of authorized WIC vendors. This chapter examines birthweight as a primary health outcome for WIC participants. Birthweight affects children's later life outcomes and represents WIC's mission to support the health of pregnant people, infants, and children. One challenge with observing birthweight data through WIC administrative records is that a child's birthweight is only observed in the administrative data if the parent enrolls the child in WIC at birth. This implies substantial selection into treatment. Many infants' birthweight is not observed because the infant was not enrolled in WIC at birth. To account for this selection into the sample, I estimate bounds on the local average treatment effect of a vendor DQ on birthweight using Lee (2009) bounds. The paper finds minimal effects of changes in the vendor choice set from disqualifications on birthweight measures including low birth weight (LBW) and very low birthweight (VLBW). Results show no large effect even after adjusting for gestational age - whether or not the pregnancy went to term. This evidence can be interpreted in two ways. One possibility is that vendor DQ effects on participation and benefit redemption of pregnant people are not substantial enough to cause changes in birthweight. An alternative explanation is that participation effects are may cause changes in birthweight, but that in leaving the program themselves parents also do not enroll children, meaning that the effect on birthweight cannot be detected. Each of these three papers contribute to our understanding of how economic agents involved in the food retail environment and WIC respond to the incentives created by WIC's program rules. This research demonstrates that individuals (and to some extent firms) alter their behavior in response to the incentives of the program. These incentives can oppose each other, leading to ambiguous responses by agents, particularly firms that may face constraints set by broader chain policy. Taken together, the research in this dissertation provides evidence on when policy changes and structures in safety net programs lead to unintended consequences for economic agents, demonstrating that common program designs can affect households and firms in ways that are not expected by program administration.

In this dissertation, I analyze the economic effects of weather, climate change, adaptation, and natural resource scarcity through the lens of agriculture in California. Quality is central to much of agriculture because of its role in price determination and contractual arrangements. Nevertheless, prior work on the effect of weather and climate change on agricultural production mostly focuses on staple crop yields. Ignoring quality will likely bias estimates of the impact of weather and climate change on productivity and farm income. In the first essay, we quantify the effect of temperature exposure and climate change on the revenue of specialty crop growers through two pathways: quality and yield. In contrast to earlier work on irrigated crops, we find extreme temperatures negatively affect both yield and quality leading to reduced grower revenue in a setting where irrigation is the norm. While the yield effect dominates, failing to account for quality significantly underestimates the true effect of temperature exposure on revenue by up to 20%. We predict climate change will significantly reduce yield, quality, and revenue by century's end absent additional adaptation. In the second essay, I investigate how natural resource scarcity affects farmers' decision making. Using a panel of 3,300 irrigated fields in California, I ask whether farmers respond to water scarcity by changing whether they plant, what they plant and how they produce. Novel data provide evidence of the mechanisms by which growers use water more efficiently - insights unattainable in publicly available or survey data. I find that during a water scarce year, growers are more likely to plant earlier, plant fast-maturing varieties, and preferentially plant fields equipped with drip irrigation rather than less water-efficient technologies. Water access drives how growers conserve water: growers with low-priority access rely on more costly margins, such as fallowing or changing growing practices in ways that reduce revenue per acre. This demonstrates that agricultural producers engage in water-saving practices more than what has been previously found and that these practices help growers avoid fallowing, a much more costly response. The first and second essays use proprietary data on California's $1 billion processing tomato industry, which produces nearly one-third of the world's processing tomatoes. Tomatoes are the second-most consumed vegetable in the United States and contain nutrients that are important to human health, but often under consumed. These novel data are collected by a large tomato processor for the purposes of contracting and payment, and include field-level quality, yield, price, and grower practices for many independent farmers operating in California. The third essay returns to the relationship between quality, weather, and climate for a different specialty crop product: wine. Wine is the most differentiated of all agricultural products, with much of the differentiation based on the combination of wine grape varieties and “terroir”: the natural environment in which the grapes are grown. The objective of this study is analyze the complex relationship between collective reputation, climate, weather, and price premia and quality for varietal wine in California. We find temperatures warmer than the regional norm had negative effects on both Cabernet Sauvignon and Chardonnay wine prices and scores. We also find that wines from premier regions are less influenced by deviations in temperature from regional climate. This supports the notion that producers of higher-value wines intervene more to mitigate the negative effects of weather. Quality matters for all agricultural producers but some more than others. Prices of processing tomatoes vary plus or minus 20% in my sample whereas wine grape prices vary by a factor of 50, even for the same variety and vintage produced in the same region. For processing tomatoes, we benefit from detailed and precisely measured data on quality attributes and how they affect grower revenue. Quality variation is arguably more important for wine grapes and wine but also more complex, which makes it more difficult to cleanly link weather effects to attributes that affect prices of grapes and wine. Rather than direct observations of quality attributes, we use expert ratings and recommended retail prices as indicators of wine and wine grape quality. Both papers contribute to our understanding of how weather and climate influence agricultural product quality. Each of these three papers contributes to our understanding of how economic agents are affected by and respond to weather, climate, and climate change. Most of what we know about agricultural production in the face of climate change comes from research on yields of staple crop. This dissertation contributes by analyzing specialty crops that are understudied despite making up 40% of the total value of U.S. crops. Taken together, the research in this dissertation highlights potential harm from a hotter and drier climate absent adaptation.

A great deal of research on farm and food policy and consumer choice focuses on the link between food consumption and nutrition. This paper presents and applies a new method to analyze the demand for food and nutrients, and consumer welfare. The foundation for this method is the recent extension of the Gorman class of exactly aggregable demand models to incomplete demand systems. The purposes of this approach are to derive and implement coherent, flexible empirical models of food and nutrient demand, to estimate the model parameters consistently with aggregate data, and to make inferences on the impacts of farm and food policy changes on food and nutrient demand and consumer welfare of those policies. We apply this framework to annual per capita U.S. demand for food and nutrients 1919–-2000, excluding 1942-1946 to account for World War II. The empirical model is an incomplete system of Gorman Engel curves that is nonlinear in income. This class of demand models generates theoretically consistent, exactly aggregable systems of demand equations for which only a small number of summary statistics for the distribution of income are required to estimate the model’s parameters consistently using aggregate data. This model is estimated with 0° homogeneity, monotonicity, symmetry, and a negative semidefinite Slutsky matrix imposed globally, so that the results can be used directly for economic analyses. A complete time series of price and income elasticities of food and nutrient demand are obtained for the sample period.

In this dissertation, I investigate how patients and providers respond to changes in healthcare technology with information and resource constraints. The following chapters are three diverse approaches to this core inquiry.

Pharmaceutical innovations in the 20th century were transformative in the prevention and treatment of cardiovascular disease. Yet, the efficacy of medication may be lower than clinical expectations due to adjustments in perceived risk that cause changes in behaviors, a phenomenon known as risk compensation. In Chapter 1, the 1973 FDA approval of new classes of drugs to treat high blood pressure and high cholesterol is used to identify the effect of medication availability on nonsmoking, adherence to a diet, and body mass index. Results show that medication approval significantly decreases the probability of engaging in healthy behaviors, evidence of risk compensation. Once a diagnosis of cardiovascular disease (CVD) is received, a patient has updated information about the state of her health which may induce the adoption of healthy behaviors. For smoking, a diagnosis of CVD does partially attenuate the risk-compensation effect of medication. After medication is approved, individuals at high risk of CVD have increased take-up -- an indication that risk screening is implemented. Also, a CVD diagnosis prompts medication use as a complement to multiple healthy behaviors. The evidence demonstrates the importance of promoting healthy behaviors to a broad population and increasing risk-factor salience prior to diagnosis.

Chapter 2 provides new evidence on how technology affects healthcare markets by focusing on one area where adoption has been particularly rapid: surgery for prostate cancer. Over just six years, robotic surgery grew to become the dominant intensive prostate cancer treatment method. Changes in patient volume due to robot adoption are estimated using a difference-in-differences design. Results indicate that adopting a robot drives prostate cancer patients to the hospital. Estimating changes in patient volume at the market level tests whether this result reflects market expansion or business stealing. The findings here are significant but smaller, suggesting that adoption expands the market while also reallocating some patients across hospitals. Marginal patients are relatively young and healthy, inconsistent with the concern that adoption broadens the criteria for intervention to patients who would gain little from it. The chapter concludes by discussing implications for the social value of technology diffusion in healthcare markets.

Chapter 3 returns to the consideration of pharmaceutical drugs with a theoretical look at the way information, through advertising, impacts the decision of a generic pharmaceutical manufacturer to enter the market after patent expiration. Brand-name pharmaceutical firms with patent protection advertise in two ways: to physicians as detailing and direct-to-consumers through mass media. Prior research shows that each type of advertising uniquely influences markets. Physician advertising generates goodwill that the brand retains after patent expiration while direct-to-consumer advertising (DTCA) expands demand for the entire class of drugs. A potential generic competitor’s entry decision is based on its evaluation of future market conditions, which are affected by brand loyalties created by detailing and market size, as determined in part by DTCA. A two-stage vertical differentiation model is used to develop testable hypotheses that physician advertising is necessary for entry to avoid Bertrand competition, but in high levels acts as an entry deterrent. Conversely, DTCA promotes entry through market expansion. During the period of patent protection, the brand firm optimally chooses levels of detailing and DTCA to maximize expected profits over the patent-protection period and the period after patent expiration, anticipating the effects of its actions on generic entry. The analysis concludes by considering the impacts of pharmaceutical advertising and generic entry on consumer welfare.

Objective

Design

Setting

Subjects

Results

Conclusions

Gorman’s theory of demand is extended comprehensively to incomplete systems. The incomplete systems approach dramatically increases this class of models. The separate roles of symmetry and adding up are identified in the rank and the functional form of this class of models. We show that symmetry determines rank and the maximum rank is three. We show that adding up and 0º homogeneity determines the functional form and there is no functional form restriction for an incomplete system. We prove that every full rank system and reduced rank systems with a minimal level of degeneracy can be written as a polynomial in a single function of income. A complete set of closed form solutions for the indirect objective functions of this class of models is derived. A simple method to nest rank and functional form for incomplete systems is presented.

We measure the extent to which curbside access affects quantity recycled. We use novel data to distinguish between new recycling and material diverted from other recycling modes. We find that the marginal impact of expanding curbside programs on total recycled quantities is small, in part because curbside programs significantly cannibalize returns from drop-off recycling centers. Failure to account for cannibalization from other modes may substantially over-estimate the benefits of curbside programs. We conclude with simple cost effectiveness comparisons. Results suggest that incremental expansion of curbside access may not be cost-effective.

This paper proposes a straightforward, easy to implement approximate F-test which is useful for testing restrictions in multivariate regression models. We derive the asymptotics for our test statistic and investigate its finite sample properties through a series of Monte Carlo experiments. Both theory suggests and simulations confirm that our approach will result in strictly better inference than the leading alternative