Original geospatial and business data analysis conducted by RAFI indicates that the four-firm concentration ratio in the broiler chicken industry may now be as high as 63.4%. However, focusing on the national level of industry concentration in the poultry industry completely misses the true level of monopsony power wielded by poultry integrators over growers. 

To understand why this level of regional analysis is necessary, it is vital to understand the practical constraints that limit farmers’ ability to negotiate between more than one or two buyers. To minimize the transportation costs and associated economic losses that may be incurred due to chicken injury or death during transit, 90% of broilers are raised on a farm no further than 60 miles from their processing plant (MacDonald, 2014). This means that the radius within which a poultry grower can practically seek other buyers or contracts is extremely limited. 

RAFI projects that nearly 60% of all growers in the United States only have access to one integrator. These regional constraints that farmers and ranchers face by virtue of their reliance on long-term land ownership and their production of perishable commodities make them particularly vulnerable to regionalized monopsony power. (See more below map.)

In the map tool above, major poultry processing plants engaged in broiler slaughter are marked with white pins. The colored regions around those processing plants are “captive draw areas” – regions around each plant that are 60 miles or less by road travel. Any farm located in a red-shaded area only has one integrator processing plant within 60 miles of the farm. Farms located in yellow regions only have two integrators available, while farms in green regions have three or more. 

Using satellite imagery and a computer vision model, we’ve estimated the number of poultry barns in red, yellow, and green areas to arrive at an overall estimate of how much of the broiler chicken supply is captured in highly monopsonized regions. Market share and HHI estimates are calculated using historical sales data from the National Establishment Time Series database.

Here’s an example of barns flagged by the computer vision model (the barn locations are hidden from the public map to protect farmer privacy).

Users may filter the map by selecting only certain states and view an automatically recalculated HHI, market share table, and estimate of poultry supply in one, two, or three or more integrator captive draw areas.

Methodology

The plant data from this dashboard comes from the FSIS (Food Safety & Inspection Service) Meat, Poultry & Egg Product Inspection Directory. We join the FSIS inspection records with historical business data from the NETS (National Establishment Time Series) database. We first filter the FSIS inspection data to include only parent corporations that operate plants classified as “Large” (500 or more employees) that engage in poultry slaughter. We then filter out plants known to slaughter only turkey. We also filter out NETS records for plants that closed before 2022 (2022 is the most recent year for which we have data).

We then match FSIS plants to NETS records using a multi-stage matching process. We are using a subset of the NETS data filtered to include only NAICS codes associated with animal processing and slaughter. Both the FSIS and the NETS data include latitude and longitude, so we start by matching FSIS plants to any NETS records that are located within 1000m. We exclude any NETS records with obviously incorrect names (cattle companies, restaurant supply, etc.). We then check multiple fields in all spatially matched records for a fuzzy string match: establishment name, address, and DBAs. We also check for a match on DUNS number. If we have multiple geospatial matches, we select the record that matches on the most fields. In the case of a tie, we select the NETS record with the highest estimated sales. (A NETS record is created for each unique DUNS number, so there are often multiple colocated records for large plants.) In practice, this works well and does not yield any spurious matches.

We use the best-matched record NETS record for each FSIS plant to estimate that plant’s sales. NETS uses reported sales at the establishment level when available — otherwise, they use firm-level or industry-level sales per employee to estimate establishment sales. We set a lower threshold of $50,000,000 for plant revenue. We replace any missing sales records or records with less than $50,000,000 with the median sales for plants with the same parent corporation. For parent corporations that have no match, we use the median of all plants. For plants classified as “small”, we use the median of all small plants.

We then calculate a 60-mile driving distance around each plant using the Mapbox API. 90% of all chickens are raised within 60 miles of the plant where they are slaughtered (MacDonald, 2014). We then group plants by parent corporation and calculate intersections between the 60-mile driving radius around each plant to find areas with access to one integrator, two integrators, and three or more integrators.

We also estimate the percentage of poultry barns that are in a captured area. We use data from Microsoft’s computer vision model, which classifies poultry barns in aerial photography from NAIP (National Agriculture Imagery Program). The raw data from Microsoft’s model includes many false positives, so we filter out any barns that do not have a nearest neighbor within 50m (most poultry barns are in clusters of 2-4). We exclude any points located within geospatial datasets for airports, hospitals, parks, railroads, coastlines, etc. We then calculate the percentage of barns in a selected area with access to one, two, and three-plus integrators.

Citations
MacDonald, James M. Technology, Organization, and Financial Performance in U.S. Broiler Production, EIB-126, U.S. Department of Agriculture, Economic Research Service, June 2014.