RESEARCH

Farm consolidation has profoundly transformed agriculture in developed countries, yet its economic and environmental consequences remain insufficiently understood. This project aims to estimate the causal effects of increases in farm size on agricultural performance and environmental outcomes in France. The analysis relies on an instrumental-variable strategy exploiting plausibly exogenous sources of variation in local farm consolidation. Combining panel data on agricultural, economic, and environmental indicators, we investigate how changes in farm structure affect productivity and pesticide use.

Identifying Climate Damages on Crop Yields with Endogenous Land Reallocation

with François Bareille

Standard empirical estimates of climate impacts on agricultural productivity typically regress observed crop yields on weather shocks or long-run climate changes. This paper shows that such estimates are biased when farmers reallocate crops across plots of heterogeneous quality. Because crop yields are subject to decreasing returns to land, changes in cultivated area mechanically affect average observed yields, confounding intensive-margin productivity effects with extensive-margin land-allocation responses. We develop a simple theoretical framework to characterize this bias and propose an instrumental-variable strategy exploiting exogenous changes in land-allocation incentives to correct it. Monte Carlo simulations provide a sensitivity analysis of the bias with respect to key parameters. We apply the approach to exhaustive regional data on European wheat yields, exploiting reforms of the European Common Agricultural Policy to identify climate effects net of endogenous land reallocation. Comparing standard long-difference regressions with specifications controlling for instrumented land reallocation allows us to assess whether conventional climate–yield regressions understate the damages of climate change on agricultural productivity, with implications for the measurement of climate damages and the calibration of economic models.

The Cross-Category Effects of Food Labels: A Canteen Experiment

with Olivier Allais and Hugo Molina

Dietary imbalances constitute a significant global health challenge, to which health authorities increasingly respond by adopting nutrition labels that encourage healthier food choices. Because these labels affect many food categories, cross-category synergies such as compensatory behaviors across foods may mitigate their efficacy. To evaluate the importance of this issue, we leverage a field experiment in which a nutritional food label (Nutri-Score) was introduced in 3 out of 6 canteens to assess its effect on diet quality. Using individual-level panel data on meal choices, we estimate a random coefficient logit model of demand for bundles and find evidence of cross-category synergies. Our counterfactual experiments show that these synergies altered the impact of Nutri-Score on calorie consumption.

Do Ecolabels Change What We Know, Want, and Choose? Evidence from a Canteen Experiment

with Olivier Allais

Labels informing consumers about the environmental impact of food products are a promising policy tool to encourage dietary change, yet existing evidence suggests their behavioral effects remain modest. This paper studies the impact of a point-of-purchase environmental label on food choices using data from a pre-registered field experiment conducted in French workplace canteens. Combining survey responses and revealed-preference purchase data, the study follows a three-step design: (i) a pre-intervention survey measuring baseline environmental knowledge and willingness to reduce red meat consumption; (ii) a two-month introduction of the ecolabel in three of six canteens, covering more than 10,000 treated individuals; and (iii) a post-intervention survey assessing changes in knowledge and willingness among initial respondents. By linking stated measures to observed food choices, the paper investigates the channels through which ecolabels may influence dietary behavior and inform more effective labeling policies. Preliminary results indicate precise null effects on both knowledge and choices, while the survey design allows the analysis of heterogeneous treatment effects across individuals.

Nutrition Labels and Consumer Welfare

with Johannes Kandelhardt, Simon Martin, and Joel Stiebale

This paper studies the supply- and demand-side effects of the 2020 update of the US Nutrition Facts Label, which introduced a mandatory line for added sugars and increased the prominence of calorie information. Focusing on the breakfast cereal market, we combine detailed nutritional information with retail scanner data covering US purchases between 2018 and 2022. Exploiting staggered compliance with the new labeling requirements, we estimate the effects of compliance on product reformulation and consumer demand. Preliminary results show that compliance reduced sugar content by about 0.5g per 100g while increasing demand, particularly for products without front-of-pack labels, low-sugar cereals, and products sold in lower-income markets. The findings highlight how nutrition labeling policies affect both product composition and consumer choices.

Junk Food, Leisure, and Income Inequalities

with Ondine Berland, Eve Colson-Sihra, and Hugo Molina

Income-related inequalities in diet quality are a major public health concern, yet the mechanisms underlying these disparities remain poorly understood. This project aims to disentangle causal mechanisms from confounding factors in the relationship between income and diet quality using harmonized data from France and the United States. Combining dietary surveys, household purchase panels, and econometric decompositions, the analysis examines how the income–diet quality gradient varies across indicators, demographic groups, and measurement approaches. We consider both supply-side mechanisms, such as relative price differences across foods, and demand-side factors including risk preferences, patience, and time constraints. On the demand side, our project highlights the role of nonseparable preferences between food and non-food consumption, whereby unhealthy but palatable foods may substitute for other sources of well-being under tight budget constraints. The project investigates this mechanism theoretically and empirically, and assesses its implications for the design and distributional effects of food and nutrition policies.

Organic Food and the Poor

with Pierre Mérel and Richard Sexton

Many countries actively promote the expansion of organic farming through dedicated public policies. Yet organic production typically requires more land, labor, and costly inputs than conventional agriculture, while organic food is disproportionately consumed by higher-income households. This paper develops a model of crop supply, food production, and food demand with heterogeneous consumers to quantify the distributional effects of policies specifically designed to expand organic acreage. We develop a comprehensive calibration to US data to show that, from the consumer perspective alone (i.e., excluding taxpayer costs), such policies would hurt low-income households through higher food prices, while benefiting high-income households. This project aims at highlighting the potentially regressive nature of policies aimed at promoting environmental objectives when they increase the cost of food.

Structural Preference Measurement under Survey Constraints: Evidence from a Representative Panel

with Antoine Nebout

This paper develops a framework for structurally estimating risk and time preferences in representative surveys under binding time and cognitive constraints. Our approach combines adaptive binary choices with a hierarchical structural model to jointly estimate utility curvature, discounting, and individual-level response noise using a limited number of comparisons. Implemented in a panel representative of the French adult population, the method recovers economically interpretable distributions of risk aversion and impatience while documenting substantial heterogeneity in preferences. Our estimated preference measures reproduce standard sociodemographic gradients and are strongly associated with real-world behaviors, supporting the validity of the structural approach in representative survey settings.