What Can Cows Tell Us About Climate Change?

It is well established that expectations over future risks can affect current decision-making (Knight, 1921; Bloom, 2009). As science continues to compile evidence on climate change’s impact, business leaders are likely to incorporate risks arising from climate change into their decision-making models (Weick, Sutcliffe, and Obstead, 2005). For example, insurance markets are adjusting to increased flood events (Bin and Landry, 2013) and timber markets are adjusting prices of forest lands to account for higher wildfire risks (Wang and Lewis, 2024). We examine whether increased acceptance of impacts from climate change could lead to a shift in the location of the US beef packing industry. Current locations of cows and processing plants were optimized in a period before decisions incorporated climate risks. With today’s awareness of those risks, Aghion et al. (2021) suggests that we should expect different decisions.

Farmers and agribusiness leaders should be cognizant of the potential for climate change to impact their operations. Schlenker and Roberts (2009) and Burke and Emerick (2016) both showed that adaptation could lessen the impact of climate change on agricultural production but not completely eliminate it. Thus, decision-makers should both be aware that climate change is going to have impacts and be looking for ways to adapt to mitigate the damage.

Here we explore the possibilities for adaptation within the beef packing industry, specifically whether we should expect to see changes in the number and location of beef processing plants. Given that beef packing plants are typically among the largest local employers, this is an important question. We use optimization models of risk-averse beef packing firms of various sizes, calibrated to match US beef industry characteristics, to show how perceptions of future climate change impacts can affect the locations in which firms will choose toplace their beef processing plants. We find a small but discernable shift northward in industry capacity, withrelocation designed to mitigate the impact of heat-related events.

How Firms Pick Where to Locate Plants

The beef packing industry being studied here refers to companies that take live cattle, slaughter them, and produce primal cuts for wholesale beef markets. Firms in this industry come in three sizes, which we call big four, large fringe, and small fringe. Each firm in one of these size classes produces on the scale of 4 million, 800,000, or 400,000 head slaughtered per year, respectively. These size categories match the US beef industry, which has four firms that dominate in market share (Cargill, Tyson, JBS, and National Beef Packing) and a number of medium-sized and smaller producers supplying the remainder, or the fringe, of production. Each firm decides the location, size, and number of plants (between one to 12) they will operate. The possible locations are 12 cities with currently operating beef processing plants. Plants can be scaled to process 100, 500, 1,000, 2,000, or 5,000 head per day; the big four firms only choose from among the three largest sizes of plants, while fringe firms only choose from among the three smallest.

In simple terms, our approach asks how a cautious, or risk-averse, business will respond as climate-related disruptions become more common. Beef packers must decide where to locate plants while balancing profitability against the possibility of shutdowns caused by extreme weather events. These shutdowns are costly as the firm will still have to pay for the fixed costs (e.g., site rent and machinery costs) that occur regardless of whether the plant is operating. Firms that are more concerned about this risk may prefer to spread operations across more locations or shift capacity toward locations that are less vulnerable to climate disruption. This would limit the possibility of having costly shutdown days across the majority of their production. More detailed model descriptions are available in the appendix.

With the production side of the model specified, the increased shutdown risk plants will face over time due to climate change must be specified. This involves assuming an expected value for how often a plant will be shut down for each year and location over our 50-year planning horizon, allowing the impacts of climate change to change over time. We use projections from the 2018 IPCC report for impacts from warming (Hoegh-Guldberg et al., 2018) and the 2012 IPCC report for increased risk of extreme weather and natural disasters (e.g., hurricanes and ice storms) (IPCC, 2012). We create three sets of climate change projections: a high heat, low cold scenario referred to as climate 1; a high heat scenario (climate 2); and a high heat, high cold scenario (climate 3). Climate 1 models a scenario where the climate change effects are felt by the increase of heat-related extreme weather incidents (high temperature, wildfires, etc.) and a decrease in cold-related extreme weather incidents (blizzards, ice storms, etc.). Climate 2 models a similar scenario to the climate 1 model except there is no decrease in cold weather-related incidents.Climate 3 is the worst of the three scenarios, with an increase in all extreme weather regardless of heat or cold affiliation.

Table 1. Weather Risk Changes over the 50-
Year Time Horizon Under Different Climate
Change Scenrios
 

Table 2. Number of Days a Large Plant Will
Be Open per Year; No Climate Change,
Climate Change Year 1 and Then Year 50 for
Climate 1, Climate 2, and Climate 3
Scenarios
 

Note: ^aAll climate change scenarios have the same open
operation days in year 1.

The IPCC midrange projections currently show heat-based phenomena—extreme heat and wildfires—increasing as climate change continues and cold based phenomena—extreme cold, blizzards, and ice storms—decreasing, similar to climate 1. Hurricanes are projected to decrease in number but increase in intensity and tornado incidence is unpredictable due to its localized nature. Wildfires are highly correlated with the occurrence of drought, which is affected by climate change, and so projections of drought are used as a proxy for wildfire projections. Table 1 reports the magnitude of changes in each of the included weather phenomena for each shock level. The shutdown risks represented in these scenarios vary by location depending on each location’s vulnerability to the different climate change risks displayed in Table 1. The changes in the expected number of days that a plant operates in year 1 and year 50 are shown in Table 2.

Table 3. Results for No Climate Change
Scenario
 

Notes: Plant sizes are small (S), medium (M), large (L), jumbo
(J), and titan (T), representing plants that slaughter 100, 500,
1,000, 2,000, or 5,000 head per day, respectively. Big four firms
only choose between large, jumbo, and titan plants, whereas
fringe firms choose only between small, medium, and large.
A plant size–location choice is denoted by an “x” in that box

Figure 1. Changes to Optimal Plant
Configuration from No Climate Change
Scenario for Big Four Firms(top) and Big
Fringe Firms (bottom) for All Three Climate
Change Scenarios
 

Will Plants Move, Shrink, or Otherwise Adapt?

Firms of each size (small fringe, big fringe, big four) choose from 12 possible locations, can place no more than one plant at each location, and choose from one of three plant sizes. The total expected output from the plants chosen must be within the acceptable window of the firm’s goal production, here set to plus or minus 10%. Each firm chooses the optimal number, size, and location of plants. We calculate results for a range of risk aversion levels and for the three climate change scenarios outlined above. The results of optimal location choices are presented in Table 3 for the no climate change scenario, and Figure 1 shows changes from this baseline when considering the three climate scenarios. Green (red) arrows show where climate change spurs firms to build larger (smaller) plants, with most of the plants that would grow being in the northern parts of the beef processing footprint. Other combinations of risk aversion and perception result in almost identical results (available from the authors on request).

Figure 2. Weighted Average Latitude of
Optimal Plant Configurations
 

Under all combinations of risk aversion and climate change scenarios, only minor changes occur to the optimal plant configurations. This holds across different levels of baseline risk used as a robustness check. Firms favor a larger number of smaller plants over a smaller number of larger plants since climate-caused shutdowns are likely to be localized events, not forcing all plants to shut at the same time. Generally, as risk aversion or climate change severity increases, there are occasional additions of a plant to maintain goal production with more anticipated climate-related plant shutdowns. Overall, the impact of considering climate change risks on the optimal industrial organization of beef packing plants is not large.However, there are discernible shifts in beef production. Rather than focusing on specific plant size and location choices, we calculate the average latitude of beef processing under different scenarios and find that the center of gravity for big four and large fringe beef production shifts to the north with climate change expectations for all three climate change scenarios. As shown in Figure 2, latitude shifts for big four firms are up to 4°, and big fringe firms move about 1° north. Small fringe firms maintain their latitude, only moving about a third of a degree as they have fewer plants and fewer options for small configuration changes. Figure 1 shows the movement north on a map of the US, making clear the geographical locations of any disruption to current plant locations.

The explanation for these results appears to be the electric grid disruptions during heat waves plus larger Gulf hurricanes outweighing impacts from northern extreme cold. Interestingly, we tested whether moving the location of cattle for slaughter increased the magnitude of this shift; it did not. This was done by testing whether removing the three southernmost feedlot locations and making processors pay for transportation from more northern feedlots changed the optimal location of processing plants (simulating cattle producers moving north to avoid heat stress). Thus, we think that even without a full endogenous modeling of the location of the cattle themselves, we have a good estimate of the expected magnitude and direction of industry adjustment.

Implications

Beef packing plants are large employers, with upward of 5,000 workers at the largest plants. In many counties, such plants are the largest local employer. Thus, if climate change is going to cause relocation of the beef industry on a significant scale, it will be important for communities at risk of losing so many jobs (and so much tax base) all at once to prepare. The good news is that we find that for plausible shifts in climate-related plant shutdown risk, we can expect a noticeable but not overly large shift northward in the optimal location of the industry. Additional impacts such as a shift toward alarger number of smaller plants do not appear as likely. Robustness checks for possible shifts in feedlot locations, risk aversion, or baseline levels of shutdown risk do not change that basic result. Thus, it appears that beef cattle may soon be both a visible sign of climate change and an optimistic signal that adaptation in many areas may be manageable.




 

Apendix 

To measure the impact of climate change expectations on the beef packing industry, we use a standard mean-variance optimization model, where the firm is trying to maximize profit over a 50-year planning horizon while penalizing decision choices that result in a large amount of expected variability in profit per year. Climate risks enter the model through increasing the variability of expected profits, with this effect resulting from increased variability in the number of days a plant will be open or closed due to climate change related events (floods, ice storms, hurricanes, wildfires, etc.). The mean-variance formulation allows modeling of variable levels of risk aversion using λ as a firm-specific risk aversion parameter (Varian, 1992). The risk aversion parameter λ penalizes plant configurations with more variation in profit, and more risk-averse firms impose a larger weight on this term.

To compute the variance in profit, we leverage the assumed binomial nature of the shutdown risk (plants are either open or closed) for which given an expected value of φ, the variance will be φ(1 – φ). Thus, given an expected value of φ as the shutdown risk, the variance of profit will be φ(1 – φ) multiplied by the square of the variable part of the profit function. Fixed components like construction or equipment costs or loan payments are not included as they are not affected by random plant closures. The variable (or operating) profit accounts for revenues and costs including electricity, labor, water, and both assembly and distribution transportation costs. The optimization model is thus

(A1)           MaxΩj Eπ-λvarπ2 = Et=1Tβtφitp'q-w'x- λE{φit(1- φit)(φitp'q)2/2},

where p is a vector of prices for the output, beef; q is a vector of output and variable input quantities, such as live cattle, water, and electricity, with inputs expressed as negative values; w and x are vectors of prices and quantities for fixed inputs, such as rent or construction; λ is the risk aversion coefficient; β is the discount factor equal to 1 over 1 plus a discount rate, and φ is the probability the plant will be shutdown expressed as the percentage of capacity lost. Fixed costs like construction or equipment costs or loan payments are not included in the penalty for profit variability as they are not affected by random plant closures and must be paid regardless of whether the plant is open or closed. Our calculation of variable (or operating) profit accounts for revenues and costs including electricity, labor, water, and both assembly and distribution transportation costs. The decision set Ω includes the location, size (one of three), and number of plants (from one to 12) each firm chooses to operate. The possible locations are 12 cities with currently operating beef processing plants

We calibrate the model to match recent beef industry numbers. Price is set to the average prices for boxed beef and live cattle from November 2018 to August of 2021 as reported by the USDA-AMS, average state labor costs for slaughterers and meat packers is taken from BLS occupational employment and wage statistics, the national average diesel price over the past twenty years is used to compute transportation costs, and actual water and electricity costs are collected from the 12 possible locations. Cost shares for different inputs come from R. Maddock (2021, email correspondence with one of the authors), economies of scale are assumed to match Morrison Paul (2001a,b) such that increasing output by 10% only leads total variable costs to rise by 9.5%. Fixed costs come from Newlin (2020) and some specification information from Cargill (2022).

Each firm chooses the optimal number and location of plants by selecting the configuration that maximizes the expected utility of the discounted flow of profits over the 50-year planning horizon as represented by the objective function in equation (A1).

For More Information

Aghion, P., N. Bloom, B. Lucking, R. Sadun, and J. Van Reenen. 2021. “Turbulence, Firm Decentralization, and Growth in Bad Times.” American Economic Journal: Applied Economics 13(1):133–169. https://doi.org/10.1257/app.20180752

Bin, O., and L. Landry. 2013 “Changes in Implicit Flood Risk Premiums: Empirical Evidence from the Housing Market.” Journal of Environmental Economics and Management 65(3):361–376. https://doi.org/10.1016/j.jeem.2012.12.002

Bloom, N. 2009. “The Impact of Uncertainty Shocks.” Econometrica 77(3):623–685. https://doi.org/10.3982/ECTA6248

Burke, M., and K. Emerick 2016. “Adaptation to Climate Change: Evidence from US Agriculture.” American Economic Journal: Economic Policy 8(3):106–140. https://doi.org/10.1257/pol.20130025

Cargill. 2022. “US and Canada Protein Processing Locations” Available online: https://careers.cargill.com/locations-protein-processing

Hoegh-Guldberg, O., D. Jacob, M. Taylor, M. Bindi, S. Brown, I. Camilloni, et al. 2018. “Impacts of 1.5ºC Global Warming on Natural and Human Systems.” In: V. Masson-Delmotte, et al. (eds.) Global Warming of 1.5°C. Intergovernmental Panel on Climate Change.

Hultgren, A., T. Carleton, M. Delgado, D.R. Gergel, M. Greenstone, T. Houser, et al. 2025. “Impacts of Climate Change on Global Agriculture Accounting for Adaptation.” Nature 642:644–652. https://doi.org/10.1038/s41586-025-09085-w

Intergovernmental Panel on Climate Change (IPCC). 2012. Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. IPCC Special Report. Cambridge University Press.

Knight, F. H. 1921. Risk, Uncertainty and Profit. Sentry Press.

Morrison Paul, C. 2001a. “Market and Cost Structure in the U.S. Beef Packing Industry.” American Journal of Agricultural Economics 83(1):64–76. https://doi.org/10.1111/0002-9092.00137

———. 2001b. “Cost Economies and Market Power: The Case of the U.S. Meat Packing Industry.” Review of Economics and Statistics 83(3):531–540. https://doi.org/10.1162/00346530152480171

Newlin, L. 2020, June 12. “So You Want to Build a Slaughter Plant?” High Plains Journal. Available online: https://hpj.com/2020/06/12/so-you-want-to-build-a-slaughter-plant/

Schlenker, W., and M.J. Roberts. 2009. “Nonlinear Temperature Effects Indicate Severe Damages to U.S. Crop Yields Under Climate Change.” Proceedings of the National Academy of Sciences 106(37):15594–15598. https://doi.org/10.1073/pnas.0906865106

Varian, H.R. 1992. Microeconomic Analysis. Norton.

Wang, Y., and D. Lewis. 2024 “Wildfires and Climate Change Have Lowered the Economic Value of Western U.S. Forests by Altering Risk Expectations.” Journal of Environmental Economics and Management 123:102894. https://doi.org/10.1016/j.jeem.2023.102894

Weick, K.E., K.M. Sutcliffe, and D. Obstfeld. 2005. “Organizing and the Process of Sensemaking.” Organization Science 16(4):409–421. https://doi.org/10.1287/orsc.1050.0133