Source: UNIVERSITY OF CHICAGO submitted to NRP
THE ECONOMIC DISAMENITIES AND HEALTH IMPACTS OF ANIMAL FEEDING OPERATIONS’ PRACTICES IN THE UNITED STATES
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
Reporting Frequency
Annual
Accession No.
1028032
Grant No.
2022-67023-36392
Cumulative Award Amt.
$596,065.00
Proposal No.
2021-10808
Multistate No.
(N/A)
Project Start Date
Jun 1, 2022
Project End Date
May 31, 2026
Grant Year
2022
Program Code
[A1641]- Agriculture Economics and Rural Communities: Markets and Trade
Recipient Organization
UNIVERSITY OF CHICAGO
5801 SOUTH ELLIS AVE.
CHICAGO,IL 60637
Performing Department
Harris Sch Pub Policy
Non Technical Summary
This project aims to investigate the social and environmental sustainability of industrial animal farming practices by quantifying the externalities they generate through environmental pathways, and the resulting harm to rural communities and threat to public health. Our empirical strategy captures the environmental channels of exposure to contaminants from animal farming operations using fine-resolution data with a large spatiotemporal coverage to estimate their causal impact on: the local environment; the health, quality of life, and value of assets of surrounding communities; and antimicrobial resistance.Over the last century, U.S. livestock production has undergone a structural transformation towards increasingly large and intensive confined operations, called Animal Feeding Operations (AFOs). The high animal densities concentrated inside these farms have two important implications: (1) they generate, at an industrial scale, dust and animal waste that contain various pollutants known to be detrimental to human health and well-being; and (2) they are sustained by the routine application of large volumes of low-dose antibiotics, fostering the selection of antimicrobial resistance genes in the local pathogen pool. The pollutants generated in the high-density environment (including particulate matter, pathogens and malodor) and the biochemical products of antibiotic use (antibiotic resistance genes and drug residues) are then transported by air and water to surrounding communities.
Animal Health Component
(N/A)
Research Effort Categories
Basic
100%
Applied
(N/A)
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
6053999301020%
6055320301020%
7235370301020%
4030210301020%
4015330301010%
4030210205010%
Goals / Objectives
The major goals of this project are to analyze the social and environmental sustainability of dominant animal farming practices by building the most comprehensive database of animal farms and using a scalable quasi-experimental design to quantify the externalities generated through environmental pathways. U.S. livestock farming has greatly intensified over recent decades and is now predominantly composed of industrial ``animal feeding operations'' (AFOs). The confinement of animals in great densities, sustained by the routine application of low-dose antibiotics, generates pollution sources that, through air- and waterborne transmission, expose an increasing share of the population to multiple health threats, decrease quality of life and property values, and promote antimicrobial resistance. While nuisance lawsuits against AFOs are increasing across states, we lack a systematic and causal understanding of the impacts of AFO practices on the rural environment. Our study assembles the most comprehensive geo-coded AFO database, covering major producing states since 1984, and a causal inference design capturing exposure to farm pollutants from weather variation, in order to estimate the impact of environmental exposure to AFO practices on: antimicrobial resistance; key human health outcomes obtained from hospital records, medicine purchase data, and birth certificates; water quality; and neighboring property values. We leverage wind and precipitation variation and model the position of all units in the hydrological network to capture exposure mechanisms.This project addresses key elements of the U.S. animal agricultural system that determine its sustainability, on the local, national, and global scale. It will generate comprehensive and detailed knowledge on a system-wide level of significant externalities, in order to inform policy to address them and improve the sustainability of the system.This project addresses an important and understudied topic of environmental and natural resource economics (ENRE): the impacts of exposure to air and water pollution from animal feeding operations.It aims to estimate the causal relationship between the by-products of AFO practices and measures of environmental health, human health, and rural prosperity, caused by two channels of exposure: water and air transmission.This project combines (i) an identification strategy that captures \textit{causal} effects, (ii) granular data for all variables, which also enable us to capture potential heterogeneity--e.g., across number, size, and type of AFOs--in these relationships, and (iii) an extended spatiotemporal coverage (multi-state, 1984-present), and thereby will generate knowledge on these externalities that are both finely estimated and relevant on a national scale.
Project Methods
We aim to quantify the changes in water quality caused by AFOs, in particular the concentrations of pollutants known to pose a public health risk. We focus first on pathogens that originate exclusively from food animals, and can cause severe gastrointestinal illness: fecal coliforms. These bacteria, which reside in the digestive tracts of food animals and end up in their manure, are pathogens that, under short-term exposure, can induce severe gastrointestinal illness, e.g., diarrhea, vomiting and cramps. The presence of these bacteria in surface waters is an indicator of an extremely likely contamination by AFOs.We extract all existing measurements of fecal coliforms, from each water monitoring station with such measurements, from the Water Quality Portal.This platform comprises all water quality data collected by the U.S. EPA and the USGS since the early 20th century. We then tailor the drainage area and period considered to our data. The outcomes are at the level of water stations, and we know the day of measure. We thus delineate the drainage basin of each water station, and use the exogenous variation from intense precipitation events to estimate the impact of upstream AFOs on counts of fecal coliforms in surface waters.We also analyze the changes in levels of other pollutants related to AFOs that carry a public health threat, notably excess nutrients and general indicators of water quality. Because crop agriculture in the vicinity of feedlots might be an important confounder, we will extract the effect of AFOs by capturing extreme precipitation on the exact locations of (i) the main AFO facility and (ii) its sprayfields, whose coordinates we have for a sample of farms, and control for the presence of different types of crop agriculture using the USDA Cropland Data LayerThe State Inpatient Database (SID) of the Healthcare Cost and Utilization Project (HCUP) is an annual dataset of inpatient discharge records from most, if not all, community hospitals in a state.For each inpatient stay, it provides the patient's month of admission, ZCTA of residence, and the list of diagnoses with ICD-9/10 codes.We identify all ICM-9/10 codes related to antimicrobial resistance, and flag all patient records that present such a diagnosis, at the month-by-ZCTA level. We also extract the monthly total number of inpatient visits to run model specifications with case counts or rates as the dependent variable.We have already acquired data for Iowa (1990-2017) and North Carolina (2000-2016). Through the acquisition of additional data, we will use data in an area responsible for more than 50% of U.S. swine production.The HCUP SID data provide spatio-temporal variation and close to full coverage of cases of antimicrobial resistance in a given state, but do not identify the specific bacteria and antibiotics concerned. Data at the level of drug-bacteria pairs can provide useful information in our context, as certain classes of antimicrobials are used almost exclusively in distinct types of animal production. We use data from antibiograms, aggregated in a dataset available to researchers based on tests performed in hospitals and laboratories. This dataset provides spatial, temporal, bacterial and antibiotic variation down to the hospital-level, for the period 2013-2017. We will compare changes in the prevalence of specific resistant strains, e.g., distinguishing livestock-associated methicillin-resistant S. aureus (MRSA) from community-associated MRSA.Infants exposed to contaminants from crop agriculture pollutants in-utero are at risk of developing severe health conditions; we are concerned that animal farming pollutants may have a similar effect, which has so far been overlooked by the literature. We have applied for and obtained access to the CDC Linked Birth - Infant Death restricted data files for all states in our analysis, for the period 1998-2017. The birth certificates contain measures of perinatal health and background information about the mother; the restricted data files provide the mother's county of residence; and for each infant who died under 1 year of age, the information from the death certificate is linked to that from the birth certificate.We will measure exposure during the period of gestation, as the exposure of the mother may have an effect on fetal development. As detailed above, we measure the mother's cumulative exposure to the variations in weather experienced by upwind/upstream AFOs during that period, and study the effects of that exposure on infant health. Because the outcome is at the county-level, the drainage area considered will be that of the county.Residential property values capitalize the value of long-term outcomes and other disamenities such as odor nuisances. We use CoreLogic Tax and Deeds Data on residential transactions and tax assessments, covering the period 1976-2020, to quantify the change in local property values after the openings or expansions of AFOs in their vicinity.The database contains not only the sale price and location of the property, but also a set of variables on the property characteristics--such as the number of rooms, bathrooms, size of the property--such that we can control for potential changes in the composition of houses built or sold following the change in the local intensity of AFOs. Using the parcel identifiers, we will also conduct an analysis using only properties that were sold multiple times, notably before and after a change in animal production. With transaction-level data we will be able to exclude intra-family property transfers, and to test whether farms buy houses that might otherwise trigger setback distance requirements if occupied.In addition to distance to an AFO, being located downwind from the farm is an important determinant in the level of exposure. Using wind vector data, we will compute prevailing wind patterns and identify each property as downwind or upwind to the AFOs in a given radius, and characterize the extent of airborne pollution by comparing the effects upwind and downwind from a farm and how they dissipate with distance.We also want to analyze the changes in rates of multiple health conditions, identified in prior environmental health research as plausibly associated with AFO exposure, such as respiratory problems and gastrointestinal illness. We identify corresponding ICD-9/10 diagnosis codes, and count the number of total inpatient visits, recorded by HCUP SID, for each health condition, at the ZCTA-month level. We use the same research strategy as for AMR case rates.A category of aforementioned contaminants poses immediate health threats: fecal coliforms. These pathogens come from the digestive tracts of food animals, and can induce severe gastrointestinal illness when ingested through water contamination, including diarrhea, vomiting and cramps.Individuals afflicted with gastrointestinal illness may first turn to purchasing over-the-counter medicine to alleviate their symptoms. OTC sales of anti-diarrheal drugs can therefore inform us about the exposure to bacteria from upstream livestock operations.We use data from the NielsenIQ Retail Scanner and Consumer Panel Data Sets, provided by the Kilts-Nielsen Data Center at the University of Chicago Booth School of Business. The Retail Scanner data contain weekly prices and total sales for all items sold in a network of retail chains across the contiguous U.S. (accounting for 53% of all sales in grocery stores and 55% in drug stores) starting in 2006. We extract the total sales of anti-diarrheal medicine and bottled water for each store in all states for which we have AFO information, and match the store to customers' ZCTAs using the Consumer Panel Data Set. We analyze changes in purchases of anti-diarrheal drugs at the ZCTA level, following extreme precipitation events that hit upstream AFOs, and control for potential averting behavior proxied by bottled water sales.

Progress 06/01/23 to 05/31/24

Outputs
Target Audience:In the project year 2024-2025, preliminary results have been presented at the following academic conferences, and research seminars of federal agencies: (1) Social Cost of Water Pollution Workshop (Wash. DC; Oct 2023) (2) U.S. EPA National Center for Environmental Economics Environmental Economics seminar (Wash. DC; Dec 2023) (3) USDA Economic Research Service seminar (online; Apr 2024) (4) 2024 European Association of Environmental and Resource Economists Annual Conference (Leuven, Belgium; Jul 2024) Changes/Problems:The key challenge we had to deal with over the past year was a change of procurement systems and providers at theHealthcare Cost and Utilization Project (HCUP). This is the source for the hospital-level records that are the foundation of the antimicrobial resistance analysis. We had to restart the entire application process after waiting for several months without hearing back from HCUP. This delayed the purchasing of the data by over a year, but did not delay the analysis as the bottleneck is the creation of the panel database on animal feedlots--where we have concentrated most ofour efforts on in the past two years. What opportunities for training and professional development has the project provided?We have supervised two master's level students here at the university enrolled in an individual research course to work on this project over the course of a full academic semester. They worked on the labeling data task as well as the replication of papers on which this research is building on. Using additional resources available at the University of Chicago, we have offered four full-time 10-to-15 weeks long paid research assistantships to undergraduate students, who commenced work on June 3rd. During the summer, they will each work on separate components of the data and analysis, and meet twice a week with the post-doctoral researcher and the program director. How have the results been disseminated to communities of interest?Social Cost of Water Pollution Workshop (Wash. DC; Oct 2023) U.S. EPA National Center for Environmental Economics Environmental Economics seminar (Wash. DC; Dec 2023) USDA Economic Research Service seminar (online; Apr 2024) 2024 European Association of Environmental and Resource Economists Annual Conference (Leuven, Belgium; Jul 2024) What do you plan to do during the next reporting period to accomplish the goals?The biggest challenge to the empirical analysis has been to construct a panel database on the location, operation, and the number of animal units of animal feedlots. We have finished collecting the administrative records. We will finish collecting the training data set for the machine learning algorithm by the end of the summer. In fall 2024, we will have the full panel database. We hare prepared the housing transactions data, the infant morbidity and mortaltity data, the weather shocks (rain and wind) data, as well as the antimicrobial resistance (AMR) data. We have partnered with a PhD student who is taking lead on the housing values analysis, and they are already making progress. We expect a working paper to be ready by the end of 2024. We have finished the analysis of the water quality and will complete the writing of the paper in fall 2024. We will complete the analysis of the AMR and infant health oucomes in winter and spring of 2025.

Impacts
What was accomplished under these goals? We have constructed a harmonized database on the location, beginning and end of operation, changes in size (by animal category), of permitted animal feeding operations in six key states (selected for their coverage of total U.S. animal production, the availability of their permit data, and their forming a contained drainage basin enabling deeper insights into the accumulation of the water externality): Illinois, Iowa, Michigan, Minnesota, North Carolina, and Wisconsin. We constructed it by (i) combining a set of historical snapshots of permit databases maintained by state agencies, together with data extracted from the permit files themselves, (ii) downscaling operation-level information to facility-level information (i.e., correctly identifying them in space), and (iii) distinguishing actual size increases from spurious size increases due to administrative mergers. However, the overall periods and sizes of operations covered differ across the states (due to the different state-level regulatory settings). This first 'permits' version of the multi-state AFO database leads to the two following avenues of work that we are pursuing now in parallel: a. The training of a machine learning model to detect (and give an estimate of the size of) operations, beyond the period covered in our 'permits' database. The model is a convolutional neural network performing a semantic segmentation task. We are currently finishing to produce training data (polygons delineating the AFOs) out of our 'permits' database, augmented with labeled data constructed from two additional sources (a barn-level dataset of poultry in the northeastern U.S., and beef feedlots in Nebraska, to adjust for the lower representation of beef operations - relative to other operations - in our 6 states), in order to train and get the first outputs of the model by the end of the summer of 2024. b. The analysis of outcomes of the impacts of AFO practices. Specific progress, by goal, as outlined in the initiation report: - 1) Research paper on the impacts of AFO practices on water quality: We have conducted the planned analysis - the results of which we have presented at conferences - for a subset of states, and are now extending it to the full set of the 6 states. - 2) Research paper on the impacts of AFO practices on non-AMR morbidity and infant mortality We are currently running the analyses of infant mortality and morbidity, with the current 'permits' version of the AFO database. We have identified the specific perinatal conditions of concern, and our using our knowieldge of the exact location of AFOs to more precisely characterize the exposure to AFOs to explain outcomes - themselves measured at the county-by-month level. - 3) Research paper on the impacts of AFO practices on AMR outcomes: We have managed to complete the purchase of the remaining hospital records to be used in the AMR analysis. - 4) Research paper on the impacts of AFO practices on property values: We have processed the wind data, identifying prevailing wind directions from AFOs at different temporal scales, which will enable in the analysis to estimate the differential impacts of exposure to AFOs for downwind and upwind properties. The analysis is planned to start during the summer of 2024. - 5) White papers, by state: These will be completed as we finalize the research products from item (1) to (4).

Publications


    Progress 06/01/22 to 05/31/23

    Outputs
    Target Audience:The full-time postdoctoral researcher who was hired (start date, 08/01/2022) has presented results from the project at the following conferences: 1) Canadian Economics Association Conference 2023 (online) May 2023 2) AERE 2023 Summer Conference (Portland, ME) May 2023 3) LSE/Imperial/King's Environmental Economics Workshop (London, UK) Jun 2023 Changes/Problems:1) We have delayed the purchasing of the additional hospital records in order to potentially benefit from lower costs for each data file. Unfortunately, the prices have not changed. However, one more year of data became available. 2) We have not finalized the database on animal feeding operations as correspondences with state agencies took longer than anticipated. We have recruited two additional research assistants for this summer (using non-grant resources) to assist us with closing this gap. What opportunities for training and professional development has the project provided?A key component of the grant is the hiring of a post-doctoral researcher to assist in the data collection and analysis. I am happy to report that I was able to hire Claire Palandri, Ph.D. Claire obtained her doctoral degree in Sustainable Development from the University of Columbia. That degree program trains scientists with a rigorous understanding of economics and econometric techniques, as well asa rigorous training in natural science. Claire's focus has been on animal farming, hydrology, and water pollution. As such, she is an incredible fit for the purposes of this project. Claire is presenting finding from the analysis this summer at three different academic conferences. In addition, we have used the project in combination withadditional resources at the University of Chicago to offer two full-time 10 weeks long paid research assistantships to one undergraduate student, and one master's level studenthere at the university. Theycommencedwork on June 19th. During the summer, they will each work on separate components of the data and analysis, and will meet twice a week with the post-doctoral researcher and the program director. How have the results been disseminated to communities of interest?Thus far, we are planned to communicate findings in the following academic conferences this summer: Canadian Economics Association Conference 2023 (online) May 2023 AERE 2023 Summer Conference (Portland, ME) May 2023 LSE/Imperial/King's Environmental Economics Workshop (London, UK) Jun 2023 What do you plan to do during the next reporting period to accomplish the goals?For goal (1), we plan to finalize the database on animal feeding operations, re-run the analysis on water quality, incorporate the feedback from the summer conferences, and submit a manuscript to a peer-reviewed journal. For goals (2) and (3), we plan to run the analysis during fall 2023, present the findings and incorporate the feedback during spring 2024, and submit two separate manuscripts during summer 2024. Goals (4) and (5) will be the focus of the final report period.

    Impacts
    What was accomplished under these goals? The key impact of our work thus far is the construction of a database on the location, timing of operation, and scale of operation for animal feeding operations. Our data collection efforts have focused on the following states: Illinois, Iowa, Michigan, Minnesota, North Carolina, and Wisconsin. This fills a large knowledge gap, one which prevents other research teams from conducting research in this space. We are putting together a harmonized version of the database. This will allow us, as well as other researchers, to compare how animal feeding operations have impacts on the environment and on human well-being--over time within each state, but also across states. To accomplish this, we have used a variety of tools. We have reached out to state agencies, submitted Freedom of Information Act requests, and use web scraping techniques to collect raw data on the location of farms, the start date for the operation, the end date (if applicable), and the number of animals or animal units permitted. We have used additional resources at the University of Chicago to have more research assistant work on preparing and harmonizing the data. Finally, we are learning how to utilize existing remote sensing data products and image detection tools to add farmsto the database, in the cases where they are not reported or recorded in the administrative records (mostly occurs when a farm opened, closed, and those records were not kept in the administrative records). In addition, we have collected data on water quality. See below for additional details. This is the process over which we have the least amount of control because it often requires the cooperation of different state agencies. This is also the backbone of the entire project because having the correct detailed information on animal feeding operations is the key input into the analysis of the health and property values outcomes. Specific progress, by goal, as outlined in the initiation report: 1) Research paper on the impacts of AFO practices on water quality: We have collected water quality data from X. Our goal here is to study how the openings of animal feeding operations affect water quality. To do so, we have examined which of the potential parameters that might be affected by the discharge is sufficiently monitored across states and years. We have converged on a set of parameters that responds to animal manure (fecal coliforms, nitrogen and phosphorous loads), and have a sufficiently high number of measurements. We have developed several models and estimation approaches to combine the data on the exact location of the animal feeding operation, as nested within a basin, to its water qualitymonitoring station. We will be presenting results from the analysis atthree academic conferencesthis summer. Once we incorporate the feedback, we will submit a manuscript to a peer-reviewed journal. 2) Research paper on the impacts of AFO practices on non-AMR morbidity and infant mortality: We have prepared the infant mortality data for analysis. We have the code ready to produce measures of morbidity from the hospital-level records. These are not being used now as our focus is on building the database on the location of the farms, and this goal is scheduled for the end of year 2 of the project. 3) Research paper on the impacts of AFO practices on AMR outcomes: We have prepared the hospital records for analysis. We are submitting the purchasing request of the additional data by the end of this month. This will potentially allow us to benefit from: (i) lower cost for each year of data (prices for more recent years decline over time), and (ii) the availability of additional years that were not made available when the grant was written and approved. This goal is on track to have results and research products by the end of year 2 of the project. 4) Research paper on the impacts of AFO practices on property values: We have obtained the property values data. We have an initial code base from a different project that will allow us to quickly measure the distances between the farms and the properties. This analysis is on track for research products by the end of year 3 of the project. 5) White papers, by state: These will be completed as we finalize the research products from item (1) to (4).

    Publications