The US meat processing sector has recently experienced considerable volatility along a number of dimensions. Much of this volatility was associated with the COVID-19 pandemic, which disproportionately impacted labor in the meat processing sector due to close working conditions. An understanding of the factors that influence employment and wages in meat processing is important to industry participants and policymakers, who have raised concerns about the stability of the sector and its ability to respond to unforeseen shocks. As part of the American Rescue Plan, President Biden announced grants of $100 million in 2024 to improve resilience of the meat processing industry following the volatility realized during the pandemic. Uncertainty and policy responses impacted meat processing workers and led to meat processing plant shutdowns. This analysis considers factors affecting employment and wage levels in the meat processing sector. Particular attention is given to the effects of the COVID-19 pandemic. The overarching objectives of this analysis are to identify economic factors affecting labor market conditions in the meat processing sector and to measure the impacts of the COVID-19 pandemic on employment and wage relationships.
The COVID-19 pandemic was an unprecedented event that affected economies across the world. Impacts of the pandemic were especially acute in meat processing industries, where workers often operate in close proximity to one another, thus facilitating transmission of the virus. Processing plants were forced to reduce production and, in some cases, to close completely. These closures started in March of 2020. Muth and Read (Reference Muth and Read2020) note that the loss in production ranged up to 25 percent for beef slaughter plants, 43 percent for pork slaughter plants, and 15 percent for chicken slaughter plants. Absentee rates increased as workers either became ill or avoided work to mitigate their chances of infection. However, the impacts appear to have been relatively short-lived in nature. Social distancing protocols were generally lifted by June of 2020 and the industry returned to normal levels of production in short order. Cooper et al. (Reference Cooper, Breneman, Ma, Lusk, Maples and Arita2023) report that capacity utilization in pork processing plants rebounded to pre-COVID-19 levels (approximately 95%) by June of 2020. Plant capacity utilization is an important determinant of the derived demand for labor to work in the plants. The physical plant structure is relatively fixed in the short run but can be used at varying levels of capacity. Thus, an important avenue for adjusting labor demand levels is through utilization of the physical capital of a plant. Figure 1 illustrates pork processing capacity utilization and pork production during the months of the pandemic.
Figure 1. Daily Pork Plant Capacity Utilization and Output 2020 (Compiled by Mildred Haley of the Economic Research Service from unpublished Agricultural Marketing Service data). Excludes Saturday slaughter. Source: https://www.ers.usda.gov/data-products/charts-of-note/chart-detail?chartId=98682.
Capacity utilization fell to below 40% in April of 2020, but had recovered to over 90% by June of that year (Vaiknoras et al. Reference Vaiknoras, Hahn, Padilla, Valcu-Lisman and Grossen2022). Likewise, pork production dropped off in April but recovered to pre-COVID levels by early June. This demonstrates both the substantial impacts of the pandemic on plant capacity utilization and pork production as well as the relatively speedy recovery of the industry to pre-COVID-19 levels of production and plant capacity utilization. Although a plethora of studies have examined these short-run impacts of COVID-19 on food systems and meat processing, little attention has been given to the longer-run structural effects that the pandemic may have triggered.
The purpose of this paper is twofold. First, we identify and quantify economic factors related to employment and wages in the meat processing industry, focusing on the relationship between the historical volatility of employment and wages and current employment and wages. Second, we consider the longer-run impacts of the COVID-19 pandemic on employment and wages in the sector. Understanding this relationship is particularly important for improving resilience—the ability to and speed with which the meat processing sector is able to respond to shocks. We utilize reduced-form, dynamic panel data models of employment and wages in the sector. Our analysis faces a significant hurdle in terms of the available data for assessing such relationships. In particular, county-level data, which form the basis of our analysis, are often censored due to non-disclosure considerations of proprietary data. We account for this censoring by estimating a first-stage selection equation and then including the appropriate correction terms in the reduced-form, dynamic models.
Background
The meat processing sector is a key component of the wider food supply chain that links producers and consumers. This entire food supply chain was impacted by the pandemic as consumers shifted away from food consumed away from home toward at-home consumption (McLaughlin et al. (Reference McLaughlin, Stevens, Dong, Chelius, Marchesi and MacLachlan2022), Marchesi and McLaughlin (Reference Marchesi and McLaughlin2023)). Meat price spreads increased dramatically in the first few weeks of this period, but generally returned to normal levels by June, having tracked plant capacity utilization. Food supply chains were somewhat slower to recover as disruptions lingered for several months. The impacts of COVID-19 may have been especially acute in the meat processing sector due to the highly concentrated nature of the industry. At the same time, the concentrated nature of the industry may have allowed a faster recovery from the disruptive impacts of the pandemic.
Supply chain risks arise in many forms and can generally be characterized as operational or disruption risks. Tang (Reference Tang2006) defines operational risks as those that are inherent to a particular business, such as uncertain market conditions or input supply risks. Disruption risks are those that occur outside of the normal operation of a sector and pertain to issues that are unanticipated, such as disasters or other extreme events. COVID-19 is certainly an example of a disruption risk. It was unprecedented and therefore unanticipated. Further, the likelihood of such an unforeseen event was unclear and therefore was not subject to the day-to-day planning of businesses operating under conditions of uncertainty. Firms operate under conditions of uncertainty, but typically form priors regarding the outcomes of uncertain events, such as market conditions. Various risk management mechanisms such as contracts are used to manage such operational risk. However, since the probability of such events is unclear, a priori management of such risks may be impossible.
Labor market disruptions are more impactful when jobs require specific skills, such as those required for meat processing. In such cases, affected workers cannot be immediately replaced from an unskilled labor pool. These risks are especially acute in the meat processing sector, where consolidation and increasingly larger plant sizes have led to a greater degree of specialization in workers’ skills and abilities (see MacDonald et al. (Reference MacDonald, Ollinger, Nelson and Handy2000) and MacDonald et al. (2014)). Along with plant capital, labor and raw materials are the main inputs into meat processing and thus the sector is highly dependent upon a readily available supply of skilled labor. Without such, plants may be forced to scale back production, as was the case with COVID-19.
The effects of COVID-19 on markets for food and, in particular, meat products, have been the focus of a considerable amount of recent research. Hobbs (Reference Hobbs2021) discussed the potential for supply-side disruption due to labor shortages in downstream food processing and transportation. She notes that the nature of the COVID-19 pandemic afforded manufacturing facilities a period of time to make adjustments to manufacturing processes and working environments. Larue (Reference Larue2020) considered the impacts of COVID-19 on labor markets throughout the supply chain. He notes that food industry firms may have the ability to reallocate productive capacity if shutdowns are localized but that industry-wide shutdowns are difficult to manage.
Balagtas and Cooper (Reference Balagtas and Cooper2020) find that COVID dramatically decreased retail demand for food, including meat products. They also note that, beginning in March of 2020, meat processing plants were forced to scale back production by closing plants or by slowing production lines. This was driven by shocks to the labor supply. However, as noted, these effects were relatively short-run in nature. Muth and Read (Reference Muth and Read2020) note that the euthanasia of hogs occurred when plants could not process deliveries. As detailed in Balagtas and Cooper (Reference Balagtas and Cooper2020), on April 28, 2020, President Trump issued an executive order invoking the Defense Production Act to keep meat-packing plants open. This executive order exempted plants from state and local orders to close nonessential businesses.
Lusk et al. (Reference Lusk, Tonsor and Schulz2021) proposed a model of the impacts of COVID-19 on meat and livestock markets. They found that the pandemic affected meat processing plants primarily by raising their cost of production. This was driven by labor market instabilities. They also found that higher processing costs decreased processors’ demand for livestock and in turn decreased the supply of meat products at the retail level. This raised the price of meat products in retail markets as supplies were constrained. Cooper et al. (Reference Cooper, Breneman, Ma, Lusk, Maples and Arita2023) directly quantified the effect of the COVID-19 pandemic on US meatpacking production. They found a larger under-utilization rate of processing capacity for larger-sized beef and pork plants during the peak of plant slowdowns in April–May 2020. Meatpacking plants located in counties with significant COVID-19 infection rates had greater disruptions, but these plants also tended to recover quickly.
Vaiknoras et al. (Reference Vaiknoras, Hahn, Padilla, Valcu-Lisman and Grossen2022) found that the characteristics of employment in meat processing industries gives rise to higher risks of contracting communicable diseases. These characteristics include close and prolonged contact for long shifts in close proximity to one another, shared transportation and housing, and absentee policies that encouraged workers to continue to work while sick. According to the Centers for Disease Control (2020), as of May 31, 2020, 16,233 workers contracted COVID-19 and 86 workers died from the virus across 239 meat (beef, pork, poultry, bison, and lamb) processing plants in 23 states. Saitone, Schaefer, and Scheitrum (Reference Saitone, Schaefer and Scheitrum2021) found that within 150 days of COVID-19 being found within a county, large beef, pork, and chicken packing plants increased county transmission rates by 110 percent, 160 percent, and 20 percent, respectively. However, rates of infection in such counties quickly became comparable to infection rates in those counties without processing plants. Meat processing plants quickly adopted measures to stem the spread of COVID. These measures included mandatory face masks, staggered shifts, widespread testing of workers, and other measures meant to mitigate disease transmission (Vaiknoras et al. (Reference Vaiknoras, Hahn, Padilla, Valcu-Lisman and Grossen2022)).
Conceptual framework and empirical methods
Annual, county-level labor demand and supply in the meat processing sector is conceptually relevant to a number of factors specific to sectoral and county characteristics. These factors consist of variables relevant to local labor market conditions, demographics, and sector-specific market conditions, including livestock prices. We consider a conventional wage-dependent labor demand equation and a quantity-dependent labor supply equation of the following forms
$${P_{it}} = f\left( {{Q_{it}},{Z_{it}}} \right)$$
and
$${Q_{it}} = f\left( {{P_{it}},{W_{it}}} \right),$$
where
${P_{it}}$
is the relative wage paid in county
$i$
in year
$t$
,
${Q_{it}}$
is the quantity of labor supplied (and demanded in equilibrium),
${Z_{it}}$
represents factors influencing labor demand, and
${W_{it}}$
represents factors affecting labor supply in the sector. These equations are jointly solved for reduced form equations for both wages and employment as follows
$${Y_{it}} = h\left( {{Z_{it}},{W_{it}},{\lambda _i}} \right){\rm{\;\;\;\;\;}}for{\rm{\;\;\;\;\;}}{Y_{it}} = {P_{it}},{Q_{it}},$$
where
${\lambda _i}$
represents county-level fixed effects, which are specific to each reduced form equation. These reduced form equations form the basis for our empirical analysis. Labor demand shifters include the historical coefficient of variation on employment, per-capita income, and livestock prices. Labor supply shifters include the historical coefficient of variation on wages, per-capita transfer payments, the proportion of the population in the labor force, population, the unemployment rate, and general levels of wages across all occupations.
Data
Our empirical analysis is focused on annual data collected through the Bureau of Labor Statistics quarterly census of employment and wages. The census publishes a quarterly tally of employment totals and wages reported by employers, though we use annual averages for employment and wages. The data are reported at the four- and six-digit NAICS industry code levels. The census covers 95% of US jobs and workers and is available at the county level. We focus on the animal slaughtering and processing sector (NAICS 3116).
These data are supplemented with data from the consumer price index and associated regional price parity data from the Bureau of Economic Analysis. The regional price parities are used to derive a local (state and/or metropolitan area) consumer price index. This price index is used to deflate all nominal price and value series, including wages. Data from the Regional Economic Information System was used to derive measures of local incomes and transfers. Finally, the county-level unemployment rate was taken from the BLS local area unemployment statistics. The data cover the period from 2000 to 2023 and were taken from 2,009 counties. Only those counties reporting at least one establishment in the meat slaughter and processing industry in a given year were included in the analysis. Note that our observational unit is the county, observed on an annual basis. A county may have multiple meat processing establishments. Of course, those counties with no establishments were not included in the analysis.
Variable definitions and summary statistics are presented in Table 1. The panel is unbalanced in that some counties only appear in the data for a small number of years. In light of the importance of the dynamic structure being modeled, only counties with at least four consecutive years of data were included in the analysis of wages and employment.Footnote 1 It should be noted that only 17% of the total number of counties having meat processing plants could be included in the employment and wage analysis. This is primarily due to disclosure limitations as well as data missing within the lags necessary for forming proper instruments from lagged values. However, these counties contained 46.3% of the total count of plants over time, reflecting the fact that many counties had multiple plants. Thus, our sample captures a significant proportion of the overall industry. We make appropriate corrections for selection bias due to the unobservable counties having plants, but it must be acknowledged that data availability poses significant limits on the sample size available for analysis. We also recognize that many dynamic adjustments of interest may occur within a year; our use of annual data may obscure some patterns of employment and wage adjustment. The short-run nature of adjustments to COVID-19 disruptions is an example. Thus, our analysis of annual data focuses on longer-run structural effects, which are likely to be realized across years.
Table 1. Variable definitions and summary statistics
We also consider the relative variability of employment and real wages. Figure 2 illustrates annual averages of the coefficients of variation (CVs) for employment and wages in the industry. The CVs for employment levels were calculated from the quarterly data over the previous two years of employment and wages. The CVs for wages, which are only reported on a quarterly basis, were taken from the preceding two years of quarterly real wages in the sector. Marked increases in the relative variability of employment and wages correspond to the period following COVID-19 outbreaks. As would be expected, wages are more highly variable in relative terms than is the case for employment levels. The increase in historical variability in employment decreased after peaking in 2001. The lag in impacts reflects the fact that the variabilities are based upon two years of lagged employment and wage levels. It should be noted that our use of a historical coefficient of variation necessarily reflects longer-run volatility patterns since the CVs are calculated over the preceding two years. COVID-19 shocks are reflected in the CVs, but other volatility factors are likewise relevant and a historical CV does not reflect contemporaneous shocks, but rather volatility of a historical nature.
Figure 2. Coefficients of variation for meat processing employment and wages (annual averages calculated from quarterly employment and wage data).
Methods
As noted, the BLS employment and wage data are highly censored to protect the confidential nature of the reported data. This raises important challenges for empirical modeling. If this censoring is ignored, the resulting estimates are likely to be biased and therefore will result in inaccurate inferences. To address this selection bias, we utilize a first-stage probit model of selection, which assesses the likelihood of disclosure.Footnote 2
Employment and wage models were specified as follows
$${y_{it}} = {\alpha _i} + {{\rm{\Sigma }}_j}{\gamma _j}{y_{it - j}} + {{\rm{\Sigma }}_k}{\beta _k}{X_{k,it}} + {\varepsilon _{it}}$$
where
${y_{it}}$
represents alternatively annual average employment and annual average wages in the meat processing sector in county
$i$
in year
$t$
,
${y_{it - j}}$
represents lagged employment and wages to account for dynamic patterns of adjustment,
${\gamma _j}$
and
${\beta _k}$
represent parameters to be estimated, and
${X_{k,it}}$
represents a set of
$k$
covariates suggested by the above reduced form model conceptually relevant to employment and wage conditions in the meat processing sector.
The dynamic panel data estimation methods of Arellano and Bond (Reference Arellano and Bond1991) were next used to evaluate the determinants of wages and employment in the meat processing sector at the county level. Because our focus is on volatility impacts on local labor markets, key variables in our analysis include the aforementioned coefficients of variation of employment and wages. Other key variables included in the employment and wage equations are per-capita transfers, which include the aggregate of social welfare transfer payments, the proportion of the population that is active in the labor force, local per-capita real income, an index of livestock prices, population, the unemployment rate, and local wages per job in the overall economy. The livestock price variable is the local annual average of real beef, pork, and poultry prices.Footnote 3 The panel data models also include county fixed effects. These empirical models are best thought of as dynamic reduced-form models. We utilize robust standard errors in the estimation.
Results and discussion
Estimates of the probit selection model are presented in Table 2. The count of establishments in a county plays a major role in the censoring of detailed employment and wage data. This correlation is expected as confidentiality/disclosure concerns are more likely to apply when there are few firms in a county. Disclosure is less likely in counties with significant relative livestock sales (in proportion to overall farm sales) and in counties with higher per-capita real incomes. Disclosure is more likely in counties with higher annual average unemployment rates. Other factors thought to be conceptually relevant to selection, including population and the size of the labor force, are not significant determinants of the disclosure of detailed employment and wage data. Disclosure occurs when a county has a sufficient number of firms so as to preclude disclosing individual firm data. Seventeen percent of counties had data that were disclosed, representing 46.3% of plants in the industry.
Table 2. Disclosure selection mechanism probit regression (
${d_{it}} = 1$
if disclosed, 0 otherwise)
Parameter estimates, short-run and long-run elasticities (evaluated at the means of the data), and summary statistics for the employment equation are presented in Table 3, while parameter estimates for the wage equation are presented in Table 4.Footnote 4 Sargan’s test of overidentifying restrictions is not rejected at the
$\alpha = 0.10$
level. Employment and wages are often bound by contracts and other labor agreements and thus may be slow to change in response to external shocks. Further, high fixed capital requirements may mean that labor is slow to adjust since plants require labor to efficiently employ the fixed plant capital.
Table 3. Parameter estimates and summary statistics for dynamic panel model of employment
Table 4. Parameter estimates and summary statistics for dynamic panel model of wages
Wages are likely to be quicker to adjust to shocks than employment. This is confirmed by the coefficients on lagged values of employment and wages. The coefficient on lagged employment levels has a value of 0.68, indicating significant lags in adjustment in the numbers of workers. In contrast, lagged wages have a coefficient of 0.23, indicating less autocorrelation and quicker adjustment to shocks. The employment equation was found to require three lags in order to adequately capture the dynamic patterns of labor market adjustments, while the wage equation needed two lags. This likely reflects the significant fixed assets involved in meat processing, which have substantial labor requirements to keep lines operating at efficient speeds and levels. That is, employment is slower to adjust and realizes significant lags in adjustment.
Economic factors affecting employment and wages in the meat processing sector
One focus of our analysis is on the impacts of historical volatility in employment and wages in the meat processing sector. To evaluate such volatility, we include the coefficients of variation of employment and wages, as defined in the Data section. In the case of employment (the number of workers), the coefficients of variation for employment and wages both exhibit a significant negative effect on employment (Table 3). This suggests that shocks to the historical volatility of the number of workers tend to exert a negative impact on the numbers of workers hired in the sector. In the case of employment, the coefficient of variation on wages has the greatest negative effect, as indicated by the elasticity values. Volatility in wages also exhibits a negative effect on employment in the meat processing sector. Workers may be less likely to pursue employment in firms that have significant volatility in wages and employment. In contrast, volatility in wages tends to have a positive impact on overall wage levels (Table 4). Workers appear to demand a higher wage in compensation for increased levels of volatility. Historical volatility in employment does not have a significant impact on overall wage levels.
In accordance with expectations, greater public transfers, which include unemployment compensation, tend to be associated with higher wages.Footnote 5 Higher wages may be required to induce workers to take jobs instead of collecting transfer payments. Transfer payments do not have a significant impact on overall employment levels. Higher per-capita incomes are associated with lower levels of employment in the meat processing sector. This likely reflects the fact that meat processing plants tend to be located in rural areas having lower incomes. Per-capita incomes do not have a significant effect on wages in the meat processing sector. As expected, higher wages per job tend to be associated with increased employment in the sector. In contrast, higher wages per job do not affect wages in the meat processing sector. This may reflect correlation between per-capita income and wages per job in the wage equation, neither of which appears to significantly affect wages.
Higher livestock prices tend to be associated with increased wages in the meat processing sector. This is in accordance with expectations that increased economic health in the industry, which would be implied by higher prices, tends to be associated with higher wage opportunities. Livestock prices do not have a significant effect on employment in the meat processing sector. In contrast to expectations, greater labor force participation tends to be associated with lower levels of employment in the meat processing sector. This may again reflect the fact that jobs in the sector, though specialized, are relatively low-skilled and often involve arduous employment activities. More economically vibrant areas with higher participation in the labor market may realize less employment in the meat processing sector.
Areas with greater population tend to have lower levels of employment, reflecting the fact that meat processing facilities tend to be located in more rural areas. Population does not have a significant effect on wages. A higher unemployment rate tends to raise employment in the meat processing sector. Again, this may reflect the fact that jobs in the sector tend to be low skilled. More workers may tend to seek employment in the sector as overall labor market conditions worsen. The unemployment rate does not significantly impact wages in the meat processing sector.
Finally, counties having higher overall wages, as indicated by the average annual wage per job, tend to have higher levels of employment. Average wages per job do not have a significant impact on wages paid in the meat processing sector. The inverse Mill’s ratio is statistically significant in each equation, indicating important selection biases that would be detrimental to inferences if not corrected by including the term.Footnote 6
Effects of COVID-19 on employment and wages in the meat processing sector
An important objective of our analysis included a consideration of the impacts of the COVID-19 pandemic on employment and wages in the meat processing sector. We compared actual wages and employment to predicted values obtained from the model estimates for 2020, the year in which COVID impacts were the most extreme. Figure 3 (a) illustrates actual and predicted levels of employment along with a
${45^o}$
line. Values beneath the line indicate that actual employment exceeded what the model would have predicted.
Figure 3. Predicted and actual annual employment and wages in the meat processing sector in 2020.
In the case of smaller plants with lower levels of employment, a significant number of plants had employment levels lower than what would have been expected based on the model estimates. This likely reflects the negative impacts of COVID on employment levels in the sector in 2020, especially for smaller plants. However, for larger firms, employment levels were actually higher than might have been expected, conditional on predictions from the model based on 2020 conditions. Larger firms appear to have been more robust in responding to COVID employment shocks, with many having higher levels of employment than what would be expected conditional on 2020 conditions. In the case of wages, a significant number of firms had wage levels higher than what might have been expected based on the model predictions. This is especially the case for firms with higher overall levels of wages. It appears to be the case that higher-than-expected levels of wages were necessary to induce employment in the sector during the pandemic.
Concluding remarks
Our analysis considered the impact of labor market volatility on employment and wages in the meat processing sector. We find that historical volatility had a significant negative impact on employment in the sector. This was true for both employment and wage volatility. In the case of wage volatility, we found that future wages are actually higher following periods of significant wage volatility. It is important to emphasize that volatility is historical, being measured by variations in employment and wages during the preceding two years. The fact that volatility exerts structural impacts for the two years following the shocks reflects the dynamic nature of employment and wages.
We find that employment and wages in the meat processing sector also tend to respond to other economic labor and market factors. Higher income areas tend to have lower levels of employment, perhaps reflecting the fact that meat processing facilities tend to be located in rural areas with lower incomes. However, conditional on total income, higher wages per job tend to increase employment in the sector. Stronger livestock markets, as reflected in higher livestock prices, tend to raise wages in the meat processing sector.
We also considered the impact of COVID-19 on employment and wages in the meat processing sector. We accomplished this by comparing actual employment and wages to predictions based upon the estimated models. Smaller meat processors had lower levels of employment, but a small number of large processors had significantly higher levels of employment. In contrast, wages were higher after COVID-19 for most counties included in the analysis. In an aggregate sense, COVID tended to largely reduce employment but increase wages in the meat processing sector.
The results have important implications for the sector and for policymakers. Recent policy initiatives have been directed at making the sector more robust in response to disruptions such as the pandemic. Our results reveal the relationships between economic factors, including labor market volatility, and employment and wages in the sector. These results may guide policymakers as they contemplate new policies to strengthen the responsiveness of the industry to disruptions. The industry realizes decreases in employment during periods following significant labor market volatility and must offer higher wages to induce workers to take jobs in the industry. These insights may guide the industry as it contemplates reactions to such disruptions in the future.
Future research will be needed to chart the long-run impacts of the COVID-19 pandemic on various industries in the US. As noted above, the immediate effects on meat processing were relatively short-lived, lasting only a few months. Output in the industry quickly rebounded. Our research suggests that volatility of the type brought about by COVID-19 may have important impacts on employment and wages in the months and perhaps years following the shocks causing the volatility. Future research could continue to examine structural aspects of the COVID-19 pandemic as additional experience is accumulated.
Data availability statement
Data and code are available on request.
Acknowledgements
We gratefully acknowledge the helpful comments of two anonymous reviewers and Jerry Cessna.
Funding statement
The findings and conclusions in this publication are those of the authors and should not be construed to represent any official USDA or U.S. Government determination or policy. This research was supported in part by the U.S. Department of Agriculture, Economic Research Service Cooperative Agreement Number 58-3000-0-0038: “Evaluating Pork Market Impacts of the COVID-19 Pandemic.”
Competing interests
The authors have no competing interests.
Open access
