Shrimp subsidies in Mexico boosted harvests by up to 110 tons per farm, causal analysis finds

Government subsidies that paid for shrimp larvae in Mexico appear to have raised harvests on the farms that received them, according to a new econometric study that claims to be the first causal evaluation of biological-input subsidies in the country’s aquaculture sector. Using an instrumental-variables design, researchers estimated that farms with semi-intensive production systems produced about 110 more tons of shrimp over 2020 to 2025 than they would have without the subsidies they received in 2018 and 2019, while farms using intensive systems showed an estimated boost of roughly 83 tons.

The study, published in Frontiers in Marine Science and led by Gerzaín Avilés-Polanco of the Centro de Investigaciones Biológicas del Noroeste in La Paz, analyzed the Program to Promote Fisheries and Aquaculture Productivity, known by its Spanish acronym PFPPyA. Run by the agriculture ministry through the National Fisheries Commission, the program provided public funding for the purchase of biological inputs, chiefly shrimp fry and postlarvae, for producers registered in the national fisheries registry. Small producers were prioritized as beneficiaries.

The program’s operating guidelines were published in 2014, but it only received a budget allocation in 2016 and remained active until 2019. During that window, 546 of 862 applications from aquaculture economic units were approved, receiving subsidies totaling 484.9 million pesos. Over the same period, national shrimp production rose 33.18 percent, from 129,049 tons in 2016 to 171,862 tons in 2019, and reached 214,546 tons in 2020, the year after the program ended. From 2021 to 2023, the authors report, production declined noticeably relative to the program years.

The central methodological challenge is that shrimp farmers who received subsidies were not randomly selected. Farms that obtained funding already reported higher average production, more years of experience and larger cultivated areas than unsubsidized farms, so a simple comparison of the two groups would conflate the subsidy’s effect with pre-existing differences. Ordinary least squares regression, the authors note, would yield biased and inconsistent estimates because unobserved factors such as managerial skill could drive both subsidy receipt and output.

To get around this, the team used a two-stage least squares design in the tradition of the local average treatment effect framework. The instrument was a binary variable indicating whether a farmer requested a subsidy amount above the average granted by the program during 2018 and 2019. Because that average was unknown to applicants when they filed, the researchers argue the requested amount behaves like a random nudge into or out of treatment. The first stage confirmed relevance: requesting above-average amounts raised the probability of receiving a subsidy by 44.2 percent across all farms, with the effect strongest among extensive producers. First-stage F-statistics exceeded the conventional threshold of 10 for the semi-intensive, intensive and pooled samples.

The analysis drew on pooled cross-sectional data covering 5,630 shrimp farms from 2018 to 2025, of which 1,542 received biological-input subsidies during the program’s final two years. Each of the 862 applications could cover more than one farm with differing crop areas, technologies and experience levels. The models controlled for cultivated area, years of experience and type of cultivation technology to reduce omitted-variable bias, and the authors report that tests supported pooling over panel specifications with fixed or random effects.

The headline results are heterogeneous by farming system. For semi-intensive farms, the local average treatment effect was 109.95 tons, equivalent to a coefficient of about 4.7 on the log-scale outcome, with a 95 percent confidence interval from 85.1 to 141.9 tons. For intensive farms, the estimate was 82.56 tons, equivalent to about 4.4 in log terms, but with a much wider interval running from 25.1 to 271.6 tons, reflecting greater uncertainty. In both cases the subsidy coefficients were positive and statistically significant at the 99 percent level. For extensive farms, no causal estimate was reported: the subsample of 54 operations produced a first-stage F-statistic of 3.43, below the threshold of 10, meaning the instrument was too weak to identify an effect for that segment.

A robustness exercise using instrumental quantile regression, based on the inverse quantile regression estimator of Chernozhukov and Hansen, addressed whether the instrument might secretly proxy for farm size, productive capacity or managerial quality. When endogeneity was ignored, the median-production effect converted to 1.80 tons; with the instrumental-variables correction it rose to 10.80 tons, both statistically significant at the 99 percent level. The authors interpret this pattern as validating that the instrument captures exogenous variation in subsidy allocation rather than underlying farm characteristics.

Why subsidize larvae at all? The authors build their case on a credit-market failure documented in earlier Mexican research. Most shrimp producers lack access to formal financing, in part because farming occurs on federal concessions that cannot serve as loan collateral, and in part because of volatile international prices and recurrent disease outbreaks. Acute hepatopancreatic necrosis disease, which remains endemic in both intensive and semi-intensive systems, once collapsed national yields to roughly 40 percent of historical levels. Postlarvae and feed are the heaviest components of variable production costs, with compound feed alone representing up to 60 percent of total variable capital, and prices for fishmeal and fish oil hit historic peaks of 2,600 and 6,000 dollars per ton respectively after pandemic-era supply shocks.

The authors are careful to scope what their estimates do and do not show. The local average treatment effect applies specifically to compliant farmers, those who requested above-average amounts and benefited from the program’s decisions, and the authors caution that higher physical production does not guarantee financial sustainability, particularly in an international market they describe as prone to oversupply and price collapse. They also note that a previous Mexican evaluation, which found aggregate production expanded even as subsidies contracted by about 40 percent between 2010 and 2021, used a Cobb-Douglas framework without explicit causal inference and so cannot separate policy effects from technology adoption and market growth. On the strength of their own design, the authors argue that targeted input subsidies differ from generalized transfers such as diesel credits, and they recommend that any redesigned support program pair funding for disease-free postlarvae with technical assistance, aquaculture insurance and evaluation against food-security as well as economic indicators.

Subject of Research: Causal estimation of the effect of Mexican government subsidies for shrimp postlarvae on shrimp aquaculture production

Article Title: Analysis of the impact of subsidies on biological inputs for shrimp production in Mexico using instrumental variables

Article References: Avilés-Polanco, G., Beltrán-Morales, L. F., Sour, L., Magallón-Servín, P., & Salgado-Beltrán, L. (2026). Analysis of the impact of subsidies on biological inputs for shrimp production in Mexico using instrumental variables. Frontiers in Marine Science, 13, Article 1958858. https://doi.org/10.3389/fmars.2026.1958858

Image Credits: AI Generated

DOI: 10.3389/fmars.2026.1958858

Keywords: shrimp aquaculture, Mexico, subsidies, postlarvae, instrumental variables, local average treatment effect, PFPPyA, econometrics, fisheries policy, production, credit constraints, Frontiers in Marine Science