Non Technical Summary
After agricultural producers adopt environmentally friendly practices, how likely are the producers to persist in using these practices? What environmental and economic factors affect persistence? The answers to these questions are critical for the design and evaluation of sustainable agri-environmental programs. Yet despite myriad studies on the adoption of agri-environmental practices, studies on their persistence are rare. A recent review of 35 years of research on the adoption of agri-environmental practices reported that "[t]here is...little to no focus on adoption over time, a phenomenon that is referred to as maintenance and persistence" (Prokopy et al. 2019).To help fill this gap in the evidence base, we will engage in an interdisciplinary, mixed-methods research agenda that relies on developing quantitative and qualitative data and methods to shed light on the persistence of cover cropping and the factors that shape its persistence. In particular, we seek to elucidate the patterns and drivers of post-adoptionpersistenceof cover cropping at a national scale, thereby extending the rich literature on the patterns and drivers of theadoptionof cover crops (Knowler and Bradshaw 2007; Prokopy et al. 2019; Prokopy et al. 2008; Ranjan et al. 2019; Wauters and Mathijs 2014). In other words, we aim to assess the degree to which producers continue to use cover crops after adopting them and to understand the factors that affect whether producers continue to use them. Cover crops are a focus of government and private sector programs, and, unlike other working lands conservation practices, they require sustained attention and investment to deliver their benefits. Building on our prior work to develop large, field-level longitudinal data sets, we aim to identify causal drivers of the persistence of cover cropping. In particular, we expect to estimate the effects of short-term financial support for cover crop adoption, a variable under the control of USDA and its partners, as well the effects of crop rotations and precipitation, variables that will be affected by climate change. Understanding what takes place after adoption, and why, is crucial not only for designing programs to improve soil health, reduce nutrient runoff, and sequester carbon on American farms, but also for modelling the ecosystem services generated by agri-environmental practices, for putting economic values on those services, and for doing cost-benefit analyses of agri-environmental programs."Sustainability" implies the persistence of behaviors and technologies that ensure the U.S. agricultural system is productive now and well into the future. To promote this persistence, scientists and program administrators need a deeper understanding of it. What does it look like under naturally occurring conditions and why does it look that way? The answer to this question can help improve agri-environmental and climate-smart program design in two ways. First, quantifying how long the targeted practices persist, and under what conditions they will persist, is critical for efforts to quantify the returns to public and private sector investments in these programs (i.e., cost-benefit analyses). Second, knowing the conditions that facilitate persistence can improve the design and targeting of agri-environmental programs to yield the most benefits per dollar invested. Efficient use of limited budgets is important because, over the last 20 years, government agencies and nonprofit organizations have provided billions to producers in financial incentives to encourage practices that yield environmental and climate-smart benefits at local, state, national, and global levels. Importantly, the working lands sustainability programs in the U.S. and elsewhere rely on short-term contracts to induce long-term change. They are thus premised on an assumption that temporary assistance can lead to persistent changes in soil health and land use practices. The validity of that assumption, however, has not been rigorously assessed.In addition to developing new datasets and new insights into what drives the persistence of sustainable land use practices among U.S. producers, our project will provide a more behaviorally informed foundation on which natural and physical scientists can improve their models of the ecosystem and agricultural effects of agri-environmental practices. With behaviorally informed models of anthropogenic land use, scientists will be able to provide more accurate advice to improve the sustainability of agricultural production.

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
605	7410	3010	100%

Knowledge Area
605 - Natural Resource and Environmental Economics;

Subject Of Investigation
7410 - General technology;

Field Of Science
3010 - Economics;

Goals / Objectives
Goal: Understand the factors that are associated with spatial and temporal variations in the persistence of agri-environmental practices.Supporting objectives:Quantify the effect of different elements of agricultural systems on the persistence of cover crop use.Quantify the effect of climate on the persistence of cover crop use.Quantify the effect of short-term assistance from agri-environmental programs for cover crop use on the persistence of cover crop use.Develop hypotheses about the mechanisms that mediate the effects on persistence from agricultural systems, climate, and agri-environmental assistance programs and develop a behavioral theory of persistence for agri-environmental practices.

Project Methods
Through statistical and behavioral analyses, our project aims to expand generalizable knowledge about the persistence of producers' use of agri-environmental practices. We do this by advancing economic theories of behavioral persistence, developing data and methods that can be used to empirically assess persistence, and deepening our empirical understanding of the persistence of cover cropping and the factors that shape its persistence. The following steps will be taken for this project:Define Persistence: We plan to work with three definitions of persistence: (1) a positive definition based on year-to-year transitions (i.e., a multi-state extension of a survival analysis that assesses the effect of a variable on reducing the cessation of cover crop use); (2) a positive definition whereby persistence means conforming to a second-order Markov process; and (3) a normative definition whereby persistence means cover cropping in at least half of the post-adoption panel (i.e., after first time observed). For the normative analysis, we focus on the persistence patterns of fields with cover crops in the first year of the panel. We have panels of four years (ARMS) or six years (FSA, Indiana), creating a challenge for any definition based on orders of Markov processes higher than two. Moreover, the normative definition identified by four in five stakeholders in Indiana (i.e., cover cropping in three or more years out of six years) is well approximated by a second-order Markov process definition of persistence.Create Datasets: Merge the following three data sets on cover cropping with ProTracts data (containing longitudinal administrative data on USDA conservation incentive contracts with producers, specifically EQIP and CSP, including the estimated cost of cover cropping, the payment per year to the individual, the individual's start and expiration date in the program, the acres of cover crops planned, and the acres of cover crops they planted) and geo-referenced environmental (e.g., soil properties, climate, topography) and socio-economic (e.g., distance to roads and markets, population densities) data.USDA-FSA Crop Acreage Reporting Database (CARD) Form FSA-578 merged with CLU polygons for the period 2013-2019, yielding a merged dataset that provides geo-referenced, field-level, longitudinal dataset for the continental U.S. that can be used to study the persistence of cover cropping. The final data set will be a balanced panel of over 11,000,000 fields.ISDA Windshield Data for Indiana comprised of predetermined, geo-referenced points in which cover crop vegetation, summer cash crop, percent of residue on the field, and tillage method on the left and right side of a vehicle is recorded. The survey is merged with CLU shapefiles from the USDA and covers more than 31,000 fields in a balanced panel from 2014-2019.ARMS Phase II survey data for the period 2012-2018, which comes in the form of overlapping four-year panelsIdentify Causal Variables: To decide on which potential causal variables to focus our attention, we rely on two sources: (1) stakeholder expertise; and (2) the results from predictive modeling using machine learning. The stakeholders were the respondents to our 2022 online survey (see text related to Figures 1 and 6). These respondents identified six factors that they believed explained nearly 75% of the year-to-year variation in cover crop use on fields in Indiana: crop rotation (22% of this variation), whether cost-share payments were available to the farmer (15% of the variation), changes in precipitation (11% of the variation), "farmer beliefs about the profitability of cover crops" (13% of variation), commodity prices (10% of variation), and credit access (10% of variation).Estimate Causal Effects: We will develop novel strategies for estimating, without bias, the causal effects of our three variables (crop rotations, precipitation, and participation in cost-sharing programs) on persistence. Although we have done a substantial amount of preliminary work to assemble the relevant data and understand the patterns of persistence in our data, we still have more work to do to elaborate our strategies. Estimating the effect of program participation on persistence is more challenging than estimating the effects of crop rotation and precipitation because participation can play out over many years and both participation in the program and time-varying and spatially varying financial assistance levels can affect persistence.We will leverage the panel structure of our data to control for the most likely sources of confounding bias: time-invariant field and operator-level effects (i.e., field fixed effects) and time-varying county-level (or higher) effects (i.e., county-by-year fixed effects).Control for other potential time-varying sources of bias, such as local temperature and other attributes of weather, using a rich dataset of geo-referenced covariate dataset.Use the results of the case-control interviews to ensure that we are not missing important features of the decision environment that affects persistence and may be confounding our analyses. Our connections to program managers who have a deep understanding of selection into cost-share programs also helps us ensure that our empirical strategies are viable.Given the challenges of inferring causality from correlations, we will quantify the sensitivity (uncertainty) of our inferences to potential unobserved biases in our empirical design. In this vein, we will take two approaches:i.ii.If our inferences differ depending on the dataset used, we will seek to identify the sources of those differences (e.g., sources of biases, heterogeneous treatment effects across sample populations), which can shed light on the validity of our empirical strategies.We will also explore the heterogeneity of the estimated causal effects conditional on geography. In addition to exploring heterogeneity by sub-groups of states organized by USDA-ERS's nine farm resource regions, we will also explore heterogeneity for two important subgroups of states: (a) states in the Mississippi River Basin, a priority region for nutrient reductions; and (b) states with histosols (carbon-rich soils), a priority region for carbon sequestration. Understanding how the drivers of persistence may vary geographically provides actionable evidence for USDA program managers and fodder for further scientific inquiry .Conduct Case-Control Study using Semi-Structured In-Depth Interviews of Producers: We will conduct semi-structured, in-depth interviews with producers who have previously participated in cover crop incentive programs in Indiana, Illinois, Maryland, and Delaware. We will interview participants who have continued using cover crops after participating in a cost-share program, and those who discontinued after participating. Questions will include details on the farm, producer background, cover crop and conservation program experiences, and trusted information sources. Interviews will be recorded, then transcribed. Researchers will analyze the data using qualitative analysis software (NVivo). Qualitative data analysis will be undertaken by two researchers who will work together to interpret the data into common themes (codes).Elaborate Theory and Mechanisms: If we identify a causal effect from any of the three variables, we will aim to elucidate the mechanisms (mediators) through which they operate. Our primary focus for advancing our understanding of mechanisms is the cost-share programs.