Labor shortages and farmer adaptation strategies

The production of fruits, vegetables, and horticultural commodities (FVH) is highly labor-intensive, with labor accounting for 40% of the total production costs (USDA, 2022). The U.S. agricultural sector heavily relies on immigrant workers from Mexico (Fan et al., 2015), yet the supply of these workers has been declining (Charlton & Taylor, 2016; Richards, 2018; Taylor et al., 2012; Zahniser et al., 2018), and labor shortages are becoming a major challenge for agricultural producers (CFBF & UC Davis, 2019; Hertz & Zahniser, 2013; Martin, 2007; Rutledge & Taylor, 2019a, 2019b).

Fisher and Knutson (2013) explain that agricultural labor markets are unique and particularly prone to labor shortages even during periods of high national unemployment. Several factors contribute to this, including the local and seasonal nature of agricultural labor markets leading to short-term spikes in labor demand, the limited mobility of labor causing localized shortages, the physical demands and working conditions that deter potential workers, and a declining domestic labor supply (Charlton, 2022; Charlton, 2024). Economic theory suggests that a reduced labor supply leads to wage increases, changes in production practices, and adoption of new production technologies. However, the impacts of labor shortages on specific farming practices remain unclear. This study aims to fill this knowledge gap by providing empirical estimates that quantify the statistical relationship between farm labor shortages and key production and labor management practices among crop producers.

We build on recent literature examining labor shortages in U.S. agriculture. This literature focuses on three distinct research objectives. One set of studies attempts to identify whether labor shortages are prevalent in U.S. agriculture. For example, Martin (2007) reports that increased immigration enforcement will likely lead to a reduced supply of labor, more mechanization, or switching to less labor-intensive crops. A subsequent study by Charlton and Taylor (2016) finds that the supply of farm labor is declining at a rate of 1% per year. Additionally, Hertz and Zahniser (2013) find evidence of declining farm employment and rising wages, consistent with labor shortages and a tightening labor market. Moreover, Richards (2018) finds evidence of chronic labor shortages among harvest workers in California, covering the same geographic region as ours.

A second set of studies identifies key drivers of labor supply shocks. These studies note that structural changes in the Mexican economy (Fan et al., 2015), increased education, and lower birth rates in Mexico (Charlton & Taylor, 2016), and increased non-farm job opportunities for low-skilled workers both in Mexico and the U.S. are contributing to the decline in labor supply (Richards & Patterson, 1998). The impact of specific policies on farm labor supply is also highlighted in the literature. For instance, Luo et al. (2023) examine the effects of E-Verify adoption, finding that states with stricter enforcement experience a decline in the total number of farm workers. In contrast, Boucher et al. (2007) show the positive relationship between the North American Free Trade Agreement and the Immigration Reform and Control Act, and an increase in migrant farm labor supply from Mexico.

A third set of studies attempts to identify mitigation strategies that could potentially allay the labor shortage issue. For instance, Charlton et al. (2019) quantify the extent to which rising wages might keep workers employed in agriculture while Hamilton et al. (2022) find that farm employers will tend to mechanize when technology and labor are substitutes. Charlton and Kostandini (2021) find evidence that supports Hamilton et al. (2022)'s findings when focusing on the dairy sector. While these studies argue that technology and labor-augmenting mechanization might improve labor efficiency and productivity, they also find that they might not be sufficient to offset negative labor supply shocks.

These prior studies document a declining farm labor supply and rising wages, and they identify drivers of labor shortages and mitigation strategies but have not extensively explored how farmers change their practices. Our research offers novel insights into these adaptations and their prevalence. To the best of our knowledge, only one study quantifies the extent to which farmers are changing farming and labor management practices in the United States. Santos et al. (2009) find that adjusting wages and nonwage benefits are the most effective adaptation strategies for farmers in the Southeastern United States, which is consistent with our research findings. Moreover, Santos et al. (2009) argue that, in addition to offering better compensation for employees, changing certain farming practices is likely a more effective long-run strategy. We find evidence consistent with this long-run strategy in our empirical setting. Our findings provide novel empirical evidence that can be used to inform the farm labor policy discussion by shining light on the types of policy interventions that are better suited to maintain a stable source of domestically produced fresh fruits and vegetables.¹

Our analysis uses a retrospective panel of data obtained from a 2019 survey of California farmers. California is the largest FVH producer and agricultural employer in the United States. It accounts for nearly one-third of the nation's total labor expenses (NASS, 2021). Our dataset contains 671 crop farmers who report experiencing (or not) labor shortages and farm practices used over the period 2014 to 2018. To investigate the hypothesis that farmers who produce labor-intensive crops are impacted by labor shortages more than the average farmer in the state, we also examine the impacts of labor shortages on the set of crop farmers who report producing a labor-intensive crop as their main revenue-generating commodity.²

Our primary research objective is to gain insight into how labor supply-driven workforce shortages impact the adoption of specific farming and labor management practices. Our outcomes of interest include the following adaptation strategies: wage increases, changes in cultivation practices, adoption of labor-saving technologies, and use of farm labor contractors. Our empirical strategy deploys fixed-effects panel regression models at the individual-year level of aggregation. The outcome variables and the main explanatory variable of interest are binary indicator variables that identify farmers' experiences with adaptation strategies and labor shortages, respectively. We also provide supplemental analyses that use a lagged labor shortage variable in place of the contemporaneous labor shortage variable to assess the extent to which there is a delay in the change of production practices resulting from a labor shortage.

Our empirical results provide strong evidence that farmers are changing production and labor management practices to cope with labor shortages. Our results show that labor shortages are associated with an increase in the probability of raising wages by 21 percentage points, changing cultivation practices by 9 percentage points, and using farm labor contractors by 3 percentage points. The impact on technology adoption is not statistically significant in the current period. In the following period, labor shortages are associated with an increase in the probability of wage hikes by 9 percentage points, changing cultivation practices by 5 percentage points, adopting labor-saving technologies by 5 percentage points, and using farm labor contractors by 4 percentage points. These effects are generally larger for labor-intensive crop farmers.

Our study makes three main contributions. First, we advance the field of farm labor research by examining the impacts of reduced labor availability on production and labor management practices. While previous studies explore the effects of labor supply shocks on crop production and market dynamics (e.g., Brady et al., 2016; Cassey et al., 2018; Zahniser et al., 2012), our research provides new insights into how labor shortages specifically influence farming practices and management strategies.

Second, we are the first to provide empirical estimates of how farm labor shortages specifically impact farming practices in the United States. Unlike Santos et al. (2009), who analyze the impact of labor fluctuations on general production practices (i.e., downsizing, changing commodities, or adjusting machinery) with a small sample (N = 72), our study uses a larger dataset and focuses on individual farming practices. We also document whether farmers were able to hire the needed employees each season.

Third, our findings contribute to the policy discussion on U.S. food stability. The White House's National Security Memorandum on Strengthening the Security and Resilience of United States Food and Agriculture emphasizes an essential workforce and preparing for a national crisis (White House, 2022). Our study offers insights into prevalent adaptation strategies among crop producers, informing policy-makers of effective options to strengthen our food supply chain and ensure a steady supply of fresh, nutritious foods.

In the next section, we start with a discussion about the data and follow that with a description of the empirical model. We subsequently explain the results from our empirical model and conclude with a discussion of policy implications.

DATA

In this study, we use proprietary, individual-level panel data from a survey of California farmers collected in 2019. This survey collected information on farmers' experiences with labor shortages, and their production and labor management practices between 2014 and 2018. The voluntary online survey was sent to approximately 30,000 California Farm Bureau Federation members and various affiliated grower groups in January 2019 (Rutledge & Taylor, 2019a). 1071 California crop and dairy farmers participated in the survey. We focus our analysis on 671 crop farmers located across 37 of California's (CA) 58 counties in our data (see Figure S1 for sample distribution). Using this sample, we construct a retrospective balanced panel, which contains approximately 3350 observations over the period 2014 to 2018.

In addition to our main analysis, we also provide separate estimates for a subsample that focuses on farmers who produce “labor-intensive” crops to test the hypothesis that farmers who produce the major labor-intensive crops are impacted by labor shortages more than the average farmer in the state. This restricted sample includes approximately 1320 observations. In Appendix B, we also test the robustness of our results to two alternative definitions of “labor-intensive” crops, each of which is progressively less restrictive. Table 1 shows the list of crops that are in each definition. Our preferred definition is the most restrictive set of crops (i.e., definition 3), which removes all crops that had a mechanical harvest option available during our survey period.

We complement our farmer survey data with two other publicly available datasets. First, we obtain weather data from the National Oceanic and Atmospheric Administration Climate Data Online website (NOAA, 2022), which captures the average temperature and cumulative precipitation each month at the county level. We use these data to calculate county-level annual average monthly temperature and precipitation. Second, we use annual employment and weekly wage data from the Quarterly Census of Employment and Wages (QCEW) to construct a Bartik-style variable to proxy for agricultural sector labor demand shocks (Bartik, 1994; Basso & Peri, 2015; BLS, 2024). To calculate the labor demand proxy variable, we use data from crop production workers (NAICS 111) and those in agricultural and forestry support activities (NAICS 115), which capture the bulk of employment in the agricultural sector.

Table 2 displays summary statistics of the variables used in our analysis. Farmers in our sample cultivated an average of 972 acres, with 567 acres being dedicated to their main revenue-generating crop (see Panel A, Table 2). Panel B reveals that the most common crops grown in our sample were wine grapes (25%), tree nuts (20%), citrus fruits (9%), ornamental, floral, or nursery products (9%), avocados (7%), vegetables (7%), hay or haylage (6%), and deciduous tree fruits (5%). The 5-year averages of labor shortages and farming practices can be found in Panel C. On average, 29% of sampled farmers experienced labor shortages over the 5-year period. The most common production and labor management practices include wage increases (60%), use of farm labor contractors (57%), changes in cultivation practices (40%), and adoption of labor-saving technology (23%).

In Figure 1, we illustrate the pattern of labor shortages between 2014 and 2018. The percentage of farmers experiencing a labor shortage increased significantly over our sample period. In 2014, only 14% of farmers reported a labor shortage, but the share of farmers experiencing a labor shortage reached 40% in 2018. Consistent with the pattern of labor shortages, the shares of farmers increasing wages, using farm labor contractors, adopting labor-saving technology, and changing cultivation practices also increased significantly over the same period (see Figure 2). In 2014, about 33% of farmers reported increasing wages, and it reached 81% in 2018.

Similarly, the percentages of farmers who changed cultivation practices and adopted labor-saving technology were only 10% and 25%, respectively, in 2014 (see Figure 2). But in 2018, 33% implemented at least one change to their usual cultivation practices, and 56% of farmers used a labor-saving technology. A large share of farmers (49%) already used a farm labor contractor in 2014, and the percentage steadily grew to 63% in 2018. While these statistics provide some summary evidence that farmers may be changing production and labor management practices in response to labor shortages, they could be confounded by unobserved heterogeneity that could lead to omitted variable bias. In the next section, we explain our empirical model used to generate estimates of the interest and address this potential issue.

EMPIRICAL METHOD

The main explanatory variable is ${\mathit{LS}}_{\mathit{ijt}}$, a binary indicator variable that takes on the value of one if farmer $i$ in county $j$ in year $t$ experienced labor shortage, and zero otherwise. Labor shortages are based on farmers' self-reported answers to the following two questions: (1) “During the past 5 years, have you ever been unable to obtain all the workers you needed for the production of [your main crop] in [your main] county? (Yes/No)”; (2) If yes, “During which of the years listed below were you unable to obtain all the workers you needed to produce [your main crop] in [your main] county? (please check all that apply or click “I don't know”).” Respondents were allowed to select from a list of years spanning 2014 to 2018. While there is the potential for some variation in interpretation of labor shortages among respondents, the questions were designed to reflect labor challenges based on farmers' operational needs and expectations.

One identification challenge in our empirical setting is the omitted variable bias due to unobserved local labor demand shocks. Labor shortages result from market disequilibrium, so they may be caused by labor supply or demand shocks. For example, Fisher and Knutson (2013) explain that regional weather conditions may affect the timing of peak harvest labor demand, which may create local labor shortages at a time when an insufficient number of qualified harvest workers are present in the local labor market.³ Since we are primarily interested in labor-supply-induced labor shortages, identification requires that we sufficiently control for labor demand shocks that could confound our estimates.

Rutledge and Mérel (2023) identify three types of shocks that can influence the demand for farm labor. First, nonlabor input supply shocks, such as urbanization reducing available land for crop production, can shift the local farm labor demand inward. Second, changes in consumer demand can affect the acreage devoted to labor-intensive crops and thus the demand for labor. Third, productivity shocks, which affect the marginal products of inputs, can change the optimal demand for labor. These productivity shocks may result from weather events, such as extreme heat exposure during the growing season or precipitation during the pollination period.

To help mitigate bias resulting from these shocks, we include a set of covariates that help control for a variety of omitted variables. First, we include a set of individual fixed effects ${\varphi }_i$ to control for unobserved farm/farmer heterogeneity such as farm size, education level, and farming knowledge. Next, we include a set of year fixed effects ${\varphi }_t$ to control for state-wide shocks common to all counties in a given year, such as changes in minimum wages, nonlabor input supply (i.e., availability of land and capital), and crop output demand due to changes in consumer preferences.⁴ To help control for trends in nonlabor input supply, local policies, and local economic conditions over time that are specific to county, we also include a set of county-specific time trends ${\varphi }_{j}t$, where ${\varphi }_j$ is a set of county fixed effects and $\mathrm{t}$ is a continuous time variable. To control for weather-driven productivity shocks, we include annual average monthly temperature ${\mathit{\text{Temp}}}_{\mathit{jt}}$ and cumulative precipitation ${\mathit{\text{Precip}}}_{\mathit{jt}}$ variables at the county level.

As Breuer (2022) notes, to reduce the number of confounding factors, the traditional Bartik instrument focuses on plausibly exogenous dimensions of the components that make up the treatment variable, which is comprised of a local predetermined industry-specific measure of employment and a nationwide component. The parallel in our empirical setting is the use of county-level industry-specific employment shares multiplied by a state-wide measure of growth in wages, similar to the one specified in Basso and Peri (2015). The empirical strategy employed by Basso and Peri (2015) includes the Bartik variable in the model to control for sector-driven labor demand growth and test whether the underlying research hypothesis is maintained (in their case, a positive correlation between immigration and the wage growth of native-born workers). Our rationale follows that of Basso and Peri (2015), and we find that our coefficients survive the inclusion of this variable (both in magnitude and statistical significance).

In a traditional setting where the econometrician uses a Bartik instrument in a two-stage least squares regression, there are several tests that can be performed to provide reassurance that the Bartik instrument is plausibly exogenous. As Goldsmith-Pinkham et al. (2020) explain, the Bartik instrument is equivalent to using the pre-period industry shares (a component of the Bartik instrument) as a series of weighted instruments. They provide several tests for researchers to probe whether their Bartik instrument meets the exclusion restriction in a two-stage least squares regression. These tests include pre-trend tests, examining the magnitudes of the weights for each of the share instruments to see which ones are most susceptible to misspecification, and determining the extent to which the share variables are correlated with control variables in the pre-period. However, our empirical setting does not fall within the purview of these tests as we are not using the Bartik instrument in a 2SLS regression but, instead, use a Bartik-like variable as a proxy for labor demand shocks.

While the control approach resembles the traditional instrumental variable setting (i.e., sufficiently controlling for labor demand shocks should help isolate a causal effect of labor supply driven labor shortages on our outcomes of interest) the requirement of exogeneity of the Bartik instrument is not required in our setting because we do not attempt to generate exogenous variation in the labor shortage variable driven by the Bartik instrument, but are instead simply using it to mitigate some sources of unobserved agricultural labor demand shocks.

We also construct the Bartik controls using different sets of industries to test the robustness of our results to various levels of industry aggregation. We provide regression results using the Bartik control based on QCEW data from crop production (NAICS 111) and farm labor contractors and crew leaders (NAICS 115115) in Online Appendix E. Since more disaggregated crop-level data may provide more variation across counties, we also construct the Bartik control using 5-digit and 6-digit NAICS data, with results reported in Appendices F and G, respectively. Overall, we find similar results across different Bartik instruments using different industries. As data suppression is more pronounced at the more disaggregated levels, we present these as robustness checks and report the Bartik control with NAICS 111 and NAICS 115 as the main analysis. A detailed discussion on the construction of alternative Bartik controls is provided in Online Appendices E–G.

We summarize the relevant sources of omitted variables bias and the corresponding variables that are expected to mitigate them in Table 3. While it is highly likely that our models estimate a causal relationship between the labor shortages and farmers' adaptation strategies after we control for this lion set of covariates, we take caution and interpret our estimates as correlations in case we fail to fully control for unobserved labor demand shocks.

To provide insight into the extent to which there is a delay in the change of production practices resulting from a labor shortage, we also estimate models that use a lagged labor shortage in equation (1). The results between these two sets of models are qualitatively similar, but the coefficients with the lagged specification are slightly smaller in magnitude, suggesting that farmers are more likely to make changes during the period when a labor shortage occurs. Our results are also robust to the inclusion of contemporaneous and lagged labor shortage variables in the model.

REGRESSION RESULTS

The regression results from equation (1) are shown in Tables 4–7. Each table contains two panels of results that are estimated with different samples. The top panel displays the results from the full sample while the bottom panel shows the results from the set of farmers who produce crops that did not have a mechanical harvest option available during our survey period. Columns 1 through 6 present the results using the contemporaneous labor shortage variable, column 7 presents the coefficients from the specifications with the lagged labor shortage variable, and column 8 presents the coefficients with both contemporaneous and lagged labor shortage variables.

Moving from left to right (from column 1 to column 6), each estimate is produced from a model that has progressively more fixed effects and controls. For instance, the model used in column 1 does not contain any fixed effects or controls, and the model used in column 6 includes individual and year fixed effects, county trends, weather controls, and the labor demand proxy. The estimates in columns 6, 7, and 8 are from models that use the same control variables. The results show that our estimates are significantly attenuated after controlling for year fixed effects, county trends, and individual fixed effects. However, the estimates are relatively stable once the weather controls and labor demand proxy are included.

ROBUSTNESS TESTS

In this section, we highlight our findings from regressions that rely upon alternative definitions of labor-intensive crops (see Table 1 for definitions). For each alternative specification, we include an additional set of crops that are harvested both by hand and mechanically, where Definition 1 includes the most comprehensive set of crops. The results can be found in Appendix C (see Tables C1–C4). In each table, we show two panels: estimates from labor-intensive crop Definition 1 in the top panel and Definition 2 in the bottom panel.

Generally, the estimates from Definition 1 are both quantitatively and qualitatively similar to the results from the full sample. Our preferred specification in column 8 shows that labor shortages are associated with a 26 percentage point increase in the probability of raising wages in the current year, a 9 percentage point increase in the probability of changing cultivation practices, and a 4 percentage point increase in the probability of using farm labor contractors (see top panels of Tables C1–C4). In the subsequent year, labor shortages are associated with a 10 percentage point increase in the likelihood of wage hikes, a 5 percentage point increase in the likelihood of changing cultivation practices, and a 5 percentage point increase in the likelihood of using farm labor contractors. Similar to the full sample, we do not find statistically significant estimates of labor shortages on the adoption of labor-saving technology in the current year, but in the subsequent year, labor shortages are associated with a 5 percentage point increase in the probability of using a labor-saving technology.

To maintain brevity, we will not discuss the estimated results under Definition 2 in detail here. Overall, the effects of labor shortages on different adaptation strategies are similar both quantitatively and qualitatively to our main results for farmers producing crops without mechanical harvest options available during our survey period (i.e., labor-intensive crop Definition 3) (see bottom panels of Tables C1–C4).

Although the magnitudes of our estimates vary across different labor-intensive crop definitions, the overall empirical evidence from this analysis consistently shows that labor shortages are prompting farmers to alter their production and labor management practices in ways that align with our core hypotheses.

While the responses such as increasing wages and the use of farm labor contractors could be immediate, the adaptation strategies such as adoption of labor-saving technology and changes in cultivation practices could take some time to respond to labor shortages. Our main results (Tables 4–7) report one lag, but we also conduct robustness tests with longer lags to assess delayed farmer responses.

Using Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) tests, we determine the optimal lag length, testing up to 3 years with the preferred specification in model 6. The results (see Online Appendix H) suggest that three lags provide the best fit for most dependent variables, except for farm labor contractor use, which indicates two lags. While AIC/BIC provides the best-fitting models, this does not necessarily determine that the models are appropriate in our settings, as additional lags reduce sample size, limit degrees of freedom, and may introduce multicollinearity between contemporary and lagged labor shortages.

Regression results using two lags (see Table H2) show similar coefficient magnitudes to the main results (see Tables 4–7), but the current labor shortage variable becomes insignificant in some cases, possibly due to smaller sample size. Table H2 also suggests that lagged effects diminish after one lag. In Table H3, most labor shortage variables are insignificant, and some lagged effects (e.g., wage increases and cultivation changes) switch signs from positive to negative, indicating potential multicollinearity. Given these findings and the lack of concrete theoretical guidance on appropriate lag length, we consider one lag to be most appropriate and present additional lags as robustness checks.

CONCLUSION AND POLICY DISCUSSION

A growing body of evidence indicates that labor shortages are becoming an increasingly serious problem for U.S. farmers. However, to our knowledge, no existing studies have assessed how farmers are adjusting their production and labor management practices in response to these shortages. This study addresses this gap by using novel retrospective panel data obtained from a 2019 survey of California farmers to estimate the impact of labor shortages on key farming practices.

Our analysis indicates that farm labor shortages are widespread and are prompting domestic farmers to alter their production and labor management practices. The most prevalent strategies among California's crop farmers include raising wages, changing cultivation practices (such as the timing and intensity of pruning, weeding, and harvesting), adopting labor-saving technologies, and using farm labor contractors. Industry sources warn that if these labor supply pressures persist, domestic labor-intensive crop production will need to be scaled back, likely shifting to Mexico where farm wages are between $2.00 and $3.00 per hour. Although the H-2A program appears to be a potential alternative to domestic labor, recent empirical evidence suggests that it is not fully compensating for the reduction in domestic workers (Kim et al., 2023). Consequently, American farmers may need to find ways to boost the productivity of their existing workforce, seek alternatives to labor inputs, or reduce the production of labor-intensive crops.

Our findings indicate that the most common strategy to mitigate labor shortages is raising wages. However, many farmers warn that they can only increase wages to a certain extent before risking the closure of their businesses. If sustaining domestic production is a priority for American stakeholders, policymakers should consider measures to help offset rising labor costs in the United States. Increasing the labor supply is one option, while tax credits or other financial offsets could also provide relief. Additionally, the adoption of technology is becoming more widespread. Many existing technologies, such as hydraulic platforms for fruit picking and conveyor belts in fields, enhance worker efficiency.

Labor shortages are an increasingly pressing issue in U.S. agriculture, particularly for labor-intensive crop production. While these strategies help mitigate immediate challenges, they highlight the need for targeted policy interventions to ensure the long-term sustainability of domestic agricultural production.

One critical policy avenue is the promotion of labor-saving technologies. Expanding incentives for research, development, and adoption of mechanized agricultural practices can ease labor pressures while maintaining productivity. Tax credits, subsidies, and financial assistance could help farmers, particularly smaller operations, invest in these innovations. However, technology alone is not a sufficient solution. Ensuring a stable agricultural workforce remains crucial, as wage increases alone are not a sustainable long-term strategy. Reforming the H-2A visa program, streamlining application processes, and expanding legal work authorization pathways for experienced agricultural workers could help address labor supply volatility.

Additionally, as mechanization and changing labor practices reshape the agricultural sector, investment in workforce development is essential. Training programs tailored to new agricultural technologies and modernized farm operations could help workers transition into higher-skilled roles. Collaboration between policymakers, educational institutions, and extension services can facilitate this shift. Furthermore, small and mid-sized farms—often most vulnerable to labor shortages—may require targeted financial support, such as low-interest loans, grants, or technical assistance to remain competitive.

Recognizing the regional variability in labor needs, policymakers should adopt localized approaches to agricultural labor policy. Different crops, climatic conditions, and farm structures influence labor demand, making one-size-fits-all policies ineffective. Strengthening supply chain resilience through investments in infrastructure and processing facilities could also mitigate some of the pressures exacerbated by labor shortages.