Beginning farmer status and financial performance differentials

Abstract

Purpose

Beginning Farmer and Rancher programs are available for operators with ten years of experience or less on any farm. These programs support farmers who are starting operations, often without an initial asset allocation. However, some beginning farmers acquire operations that are already established, with substantial assets in place. The authors investigate whether a profitability gap exists between beginning farmers entering the industry ex novo and those operating a preexisting operation and if so, what factors contribute to the gap.

Design/methodology/approach

The authors utilize the Blinder-Oaxaca decomposition to determine what drives financial differences between first-generation beginning farmers, second-generation beginning farmers and established farmers using a unique farm-level panel dataset from 1997 to 2021.

Findings

Results indicate that first- and second-generation beginning farmers have similar operating profit margins, but first-generation beginning farmers have a statistically higher rate of return on assets than second-generation beginning farmers. Established farmers outperform second-generation beginning farmers on both the operating profit margin and rate of return on assets. These results suggest that economic viability for beginning farmers differs depending upon the initial status of their operation, suggesting that heterogenous policies may be more impactful in supporting various pathways to enter agriculture.

Originality/value

This analysis is the first to identify beginning farmers that enter the industry without an asset base and those that take over a principal operator role on an established farm through an assumed farm transition. The authors quantify differences in financial performance using detailed accrual-based financial data that tracks farms over time in one dataset.

Citation

Weir, R., Hadrich, J., Bonanno, A. and Jablonski, B.B.R. (2023), "Beginning farmer status and financial performance differentials", Agricultural Finance Review, Vol. 83 No. 4/5, pp. 762-782. https://doi.org/10.1108/AFR-05-2023-0054

Introduction

US principal farm operators are, on average, 58.6 years old (USDA-NASS, 2017) and 36% of operators are over the age of 65 and expected to retire in the near future (USDA-NASS, 2017). Due to the illiquid nature of land markets, the slow process of land transfer may be a threat to the continuity of farming. According to a 2014 survey, ownership of 10% of all agricultural land was expected to be transferred within a five-year period, but only 2.25% was to be sold on the open market or to a non-relative (Bigelow et al., 2016). More recently, nearly 100 million acres of US farmland was predicted to transition to new ownership over the course of the current Farm Bill (2018–2023) (National Sustainable Agriculture Coalition, 2023). Some farmers transition their farms to the next generation of operators, while others consolidate, selling their operations to an existing operation, which creates a net decrease in the number of farm operations. Data from the Census of Agriculture show a 10% decline in the number of US farms during the 1997–2017 period, from approximately 2.2 to 2 million operations.

In order to facilitate entry into the industry, a number of state and federal programs are available that give beginning operators priority consideration (USDA, 2023). The Farm Service Agency administers the Beginning Farmers and Ranchers Loan program, which provides access to low-interest farm ownership, direct and guaranteed and operating loans to farmers and ranchers in their first ten years of operation (USDA-FSA, 2022) [1]. The National Institute of Food and Agriculture offers grants through the Beginning Farmer and Rancher Development Program, which provides education, mentoring and technical assistance to beginning farmers (USDA-NIFA, 2022). Additionally, many states offer financial support and technical assistance for Beginning Farmers (BF), with some requiring participation in Farm Business Management Programs to access support (Minnesota Department of Agriculture, 2022a, b).

In 2017, 18.8% of US principal operators were beginning farmers (USDA-NASS, 2017), and this group's performance has been studied in a variety of contexts in recent years. The most common source of national data to study BF performance is the USDA Agricultural Resource Management Surveys (ARMS) (Detre et al., 2011; Jablonski et al., 2022b; Katchova and Dinterman, 2018; Mishra et al., 2009; Stevens and Wu, 2022; Williamson, 2016). Other studies combine the Census of Agriculture with other government agency datasets (Ahearn and Newton, 2009; Ahrendsen et al., 2022; Hartarska et al., 2022). These national datasets are assumed to be representative of the farming population, but there are a number of limitations. The Census of Agriculture is surveyed every 5 years rather than every year, while the ARMS is surveyed annually, but it does not survey the same farmers each year, so farm performance cannot be measured over time in a panel data structure. Furthermore, ARMS uses probability weights to create a representative sample. If surveyed farms are not representative of the industry, then averages can be significantly skewed. Inconsistency in the BF definition across the USDA agency, along with the definition of a farm operator within the Census of Agriculture leads to further data challenges in matching datasets to obtain financial performance data over time [2].

Farm Business Management (FBM) datasets, available from some states, are also used to analyze farm performance and they have advantages and limitations relative to the national data. Kuethe et al. (2014) compared statistical results between USDA-ARMS data and FBM data from Illinois, Kansas and Kentucky and found that FBM in these states tended to represent larger farms and a greater share of crop producers. FBM datasets often have detailed operator, farm, herd and farm financial information available, alleviating the need to combine and/or match multiple datasets to obtain a time series long enough to capture changes over time. Financial performance can then be analyzed as a farmer gains experience, rather than at a single point in time. An additional advantage of FBM data is that some states use accrual accounting practices to accurately represent asset inventories and dynamic decision making rather than cash accounting, which is common for farms seeking tax preparation services with a goal of minimizing tax liability. Finally, within the Minnesota FBM data, generational farm transfers can be identified.

Previous research documented that the financial performance of BF and their decision-making differ from established farmers (EF). Differences include technology adoption (Detre et al., 2011), farm specialization through value-added activities (Jablonski et al. 2022b; Mishra et al., 2009) and participation in direct payment and crop insurance programs (Jablonski et al., 2022a; Kropp and Katchova, 2011). Factors affecting these decisions and farm growth and survival include access to credit and land (Ahearn and Newton, 2009; Key, 2022; Stevens and Wu, 2022), the age distribution of BF (Williamson, 2016), the geographic area of the farm (Hartarska et al., 2022) and producer demographics (Ahrendsen et al., 2022), among other factors. While these studies have provided analysis of this important farmer group, the authors often relied on aggregate data or created panels across large time series gaps (e.g. 5 years), with limited information on the initial asset allocations of the farmer groups. Intuitively, it makes sense that operations with different starting points and entrepreneurial attitudes will have different performances. Williamson (2016) emphasized that to accurately study this question, a multi-year time series of annual data is needed to draw accurate comparisons between BF and EF.

The primary objectives of this research are to (1) document whether a financial performance gap exists between BF starting an operation ex-novo (First Generation Beginning Farmers – FGBF) and those taking over a principal operator role on a preexisting, established operation (Second Generation Beginning Farmers – SGBF) and (2) assess the factors contributing to such a gap. In addition to the primary comparison between FGBF and SGBF, we also compare the performance between FGBF and EF as well as SGBF and EF. We achieve these objectives by using twenty-five years (1997–2021) of accrual-based financial data from Minnesota dairy farmers, extracted from FINBIN. FINBIN is a farm-level dataset made available by the Center for Farm Financial Management (CFFM) at the University of Minnesota through a collaboration with the Minnesota State Colleges and Universities system's FBM program, and the data allow three exclusive groups to be identified: FGBF, SGBF and EF. Specifically, this research analyzes what factors drive financial performance gaps between FGBF and SGBF, FGBF and EF and SGBF and EF by using the Blinder-Oaxaca decomposition (Blinder, 1973; Kitagawa, 1955; Oaxaca, 1973) [3].

Following previous literature, we use two profitability measures as outcome variables to capture short-run and long-run profitability, operating profit margin (OPM) and the rate of return on assets (RROA), respectively (Detre et al., 2011; Gloy et al., 2002; Jablonski et al., 2022b; Katchova and Dinterman, 2018; Kropp and Katchova, 2011; Mishra et al., 2009). Because some characteristics may appear to have a different relationship with short-run (OPM) or long-run (RROA) performance metrics within BF, we use both measures to allow for a more holistic inference.

We focus on one commodity sector (dairy) and geographic area (Minnesota) for a number of reasons. First, by focusing on one sector and one state, we avoid unobserved heterogeneity in commodity-specific characteristics and drivers of profitability, as well as state-specific differences in policies and programs that may affect the decision to enter farming by either acquiring a farm ex-novo or operating on an existing one. Second, trends in the Minnesota dairy industry are exemplary of the declining farm numbers in the US (Figure 1). The number of licensed dairy farms in the US decreased by 47.4% from 53,132 to 27,932 from 2010 to 2022 (USDA-NASS, 2023). Similarly, the number of licensed dairy farms in Minnesota decreased by 52.5% (from 4,567 farms in January 2010 to 2,171 in January 2022; MDA, 2022c). Third, dairy is one of the most important agricultural sectors in Minnesota, with approximately $1.7bn in sales in 2017 and it makes up approximately 9.4% of total agricultural sales in the state (USDA-NASS, 2017). Finally, due to the large capital investment necessary to enter the dairy industry, it is unlikely that dairy farmers would begin farming as a hobby, allowing us to assume operators entering the dairy industry are driven by economic objectives.

This analysis is innovative and contributes to the literature in three ways. First, previous studies have not identified what contributes to the financial performance gap between BF and EF, second, we use detailed accrual-based financial data that tracks farms over time in one dataset and third, we are able to identify BF that enter the industry without an asset base and those that take over a principal operator role on an established farm through an assumed farm transition. Our novel contribution is that of studying whether, among BF, starting an operation ex-novo is more profitable than managing a preexisting operation.

The next section describes the empirical framework, and it is followed by the data description. The fourth section details the empirical results, the fifth section presents a discussion of the findings, and in the sixth section, the paper concludes.

Empirical framework

Econometric model

Reduced-form profitability equations are specified for FGBF, SGBF and EF. Let j indicate a group of farmers, either FGBF, SGBF, or EF.

(1)πitj=αj+∑hγhjOChitj+∑mλmjFHCmitj+∑kκkjFFCkitj+μtj+εitj=Xj′βj+εitj

where πitj is a profitability measure (OPM or RROA) for farm i in year t, belonging to group j; the OCs are h variables capturing operator characteristics; FHC_s are m variables capturing farm and herd characteristics; FFCs are k variables capturing farm financial characteristics; α,γs,λs and κs are parameters to be estimated for each farmer group; μtj represent year-level fixed-effects; and εitj is an error term. Because the analysis focuses on one geographic area (Minnesota), area-specific fixed effects are not included in the model [4].

For simplicity, all explanatory variables and the parameters to be estimated in (1) are collapsed in the matrixes Xj and βj. Using the average values of each profitability metric and explanatory variables for the two groups j and –j, respectively π¯j and π¯−j, X¯j and X¯−j and the estimated parameters β^j and β^−j. The Blinder-Oaxaca decomposition is performed by calculating:

(2)π¯ j−π¯ −j=X¯ j−X¯ −j′ β^ j+X¯ −j′β^ j−β^ −j

The first term on the right-hand side of (2), (X¯j−X¯−j)′β^j, is the portion of the difference in profitability between the two farmer groups explained by differences in the values of the covariates included the model. The second term on the right-hand side of (2), X¯−j′(β^j−β^−j), is the unexplained portion, capturing how differences in estimated coefficients contribute to the gap in profitability between the two groups. The “explained” component of the decomposition indicates how much the difference in the outcome would change had the base group (−j) been assigned the characteristics of the comparison group (j), which is referred to as the endowment effect. The “unexplained” portion captures the difference in profitability that is not explained by the differences in endowments, which is referred to as the coefficient effect. It measures the difference in profitability that a farm in the base group would experience if it showed the same returns (coefficients) as the comparison group. As illustrated by Fortin et al. (2011), the second portion of the decomposition can be interpreted as an average treatment effect on the treated (ATT). Specifically, if the two groups we are considering are FGBF and SGBF, with the latter being the “treated” group, each element of the coefficient (or unexplained) portion of equation (2) can be loosely interpreted as a “contribution” of a given covariate to the ATT [5].

Data

This study uses FINBIN data from Minnesota dairy farms and crop and dairy farms from 1997 to 2021 (finbin.umn.edu) [6]. FINBIN is a detailed farm financial database with participation from 13 states across the US, and aggregate data are publicly available. In Minnesota, data are collected through a collaboration of CFFM at the University of Minnesota and the Minnesota FBM program, facilitated through the Minnesota State Colleges and Universities system.

CFFM and Minnesota FBM are members of the National Association of Farm Business Analysis Specialists. Minnesota FBM offers student-led programs to help farmers meet business goals across eight community college programs (AgCentric, 2022). The FBM program provides participating farmers with farm management assistance, access to educational opportunities, improved record keeping, confidential peer farm benchmarking and individual consultation (Andersson and Olson, 1996; Gustafson et al., 1990). Farmers may enter or exit the program on an annual basis. Thus, the number and composition of farms in the data varies on an annual basis, resulting in an unbalanced panel which is further discussed in the concluding remarks. Furthermore, some farmers self-select into the FBM program, while others are required by their lender to participate in order to gain access to loans.

Minnesota FBM data collection are based on accrual accounting practices. Data collection across the Minnesota FBM program is standardized through a series of data reviews performed to verify that the data provided are consistent with accrual-based practices. As data are loaded into the database, an additional check for outliers is performed based on two standard deviations from the average for each variable (FINBIN, 2023). All data included in FINBIN are de-identified to protect confidentiality, with an anonymous identifier assigned to each farm so data can be matched on an annual basis to create a time series.

A unique feature of the FINBIN data is that it is comprised of whole farm and enterprise-specific datasets. The whole farm dataset consists of general non-enterprise-specific variables in which all the enterprise-level financial values are combined. These variables include the start date of the operation, an indicator variable for farms receiving the MDA's Beginning Farmer Scholarship and detailed financial characteristics at a whole farm level (i.e. depreciation expenses, total current assets, net farm income) (FINBIN, 2022b) [7]. Financial characteristics are also detailed at the enterprise level. The dairy enterprise data contain dairy-specific variables, including number of cows, herd somatic cell counts, veterinary expenses and milk yield (FINBIN, 2022a).

Farmer groups

We divided principal operators into one of three mutually exclusive groups: FGBF, SGBF, or EF and provide pairwise comparisons of each of these groups in the analysis [8]. Each farm observation in FINBIN reports the start date of the farming operation, which we used to calculate the years of experience of the principal operator. We classified principal operators with more than 10 years of experience as EF. BF are operators with 10 years of experience or less, which is consistent with the USDA Farm Service Agency definition. In 2014, the MDA initiated the Beginning Farmer Scholarship to provide BF with scholarships or cost-shares to participate in the MN FBM program, which is a requirement of the BF loan. Any BF farmer that received the scholarship also received a unique identifier, which can be used to identify BF from 2014 onwards. Prior to 2014, BF could only be identified by an operator's years of experience. We further separate BF into FGBF and SGBF. In FINBIN, the ID associated with a farm location does not change as principal operators change, but the year a farm started does adjust for the new principal operator. This allowed us to identify if a farm was transferred from one generation to another, which is the basis of the SGBF definition. Specifically, we identified SGBF according to two criteria: (1) the principal operator has 10 years or less of experience and the farm operation start date changed by 20 years or more or (2) the farmer had 20 or more years of experience but was receiving an MDA Beginning Farmer Scholarship (2014 and onwards). The second criteria for SGBF accounts for a scenario in which the years of experience recorded in the dataset are still associated with the initial principal operator on the farm. In other words, the principal operator information was linked to the initial principal operator and was not updated to reflect the operator transition, so the 20 years or more of experience is associated with the farm's initial principal operator. We know that these farms are second-generation farms because a farm could not receive a BF scholarship if its principal operator did not meet the USDA definition of BF. We identify FGBF as (1) those with 10 years or less of experience that did not experience a start date change or (2) those receiving an MDA Beginning Farmer Scholarship with 10 years or less of experience.

Profitability variables

Following previous literature (Detre et al., 2011; Gloy et al., 2002; Gloy and LaDue, 2003; Jablonski et al., 2022b; Katchova and Dinterman, 2018; Kropp and Katchova, 2011; Mishra et al., 2009), the two measures of financial performance we used as dependent variables are the OPM and RROA. The OPM measures the amount of revenue retained as profit and accounts for the opportunity costs of labor and management [9]. It is calculated as the operating profit (the sum of net farm income and farm interest less the value of operator labor and management) divided by total revenue. The RROA measures the return of all investments on a farm. It is calculated as the return on farm assets (equivalent to the operating profit) divided by the average value of the assets (reported in the beginning- and end-of-year balance sheets).

Operator characteristics

Farm operator characteristics in the model include the number of operators and an off-farm income indicator. We expected the number of operators to be positively related to profitability. As a farm increases its number of operators, each operator may specialize in different areas on the farm, which has been associated with improved financial performance (Detre et al., 2011; Mishra et al., 2009; Williamson, 2016). The off-farm income indicator is a dummy variable taking the value of 1 if the farm recorded off-farm income and a value of 0 otherwise. Working off-farm takes time and resources away from the farming operation and we expected participation in off-farm work to lower farm profitability and performance, as indicated in previous literature (Detre et al., 2011; Key and Lyons, 2019; Mishra et al., 2009).

Farm and herd characteristics

Farm characteristics include the size of the farm, captured by the total number of cows (and their squared value), total acres operated (owned and rented) and percentage of acreage operated that is owned. We measured herd size in thousands of cows for ease of interpretation of estimated coefficients and because a single cow would have a very small impact on farm profitability within the context of this analysis. We expected an increase in herd size to positively impact farm performance because large farms can take advantage of efficiencies, lowering per cow expenses. However, we expected this positive association to show a decreasing rate because herd size has decreasing marginal returns. Operating more acreage indicates that the farmer can produce more feed on the farm and diversify the operation to include grain commodity sales. Therefore, we expected total acreage and percentage of acreage owned to be positively associated with a farm's financial performance (Ahearn and Newton, 2009; Jablonski et al., 2022b; Williamson, 2016).

Herd characteristics include milk yield per cow and a herd health indicator. We used these variables to capture farm management styles (Lai et al., 2019; Neyhard et al., 2013; Gloy and LaDue, 2003). Holding all else equal, an increase in milk production will increase profitability (Gloy et al., 2002). Farmers generally manage the tradeoffs between culling cows and incurring additional veterinary expenses and we captured these decisions with the herd health indicator. We generated this variable, which is used to capture herd health problems, by assigning a value of one to farms allocating 5% or more of their operating expenses to veterinary expenses and zero otherwise [10]. We expected the herd health indicator to be negatively associated with farm financial performance.

Farm financial characteristics

Financial characteristics used in the model are interest expense per cow, depreciation expense per cow, a hired labor expense indicator, government payments as a percentage of total revenue, crop income as a percentage of total revenue and the current ratio. A high interest expense per cow indicates that the farm has debt, which may be used to expand its operations, potentially compromising short-run profits to achieve higher long-term profits. Key (2022) noted that credit-constrained farmers, as measured by interest expenses, have lower survival and growth. We expected the association between interest expense per cow and financial performance to be negative. We measured investment in the farm business by depreciation, which is a noncash expense capturing the annual loss in capital assets, such as the value of machinery and equipment, due to wear and tear. In these data all assets are depreciated using straight-line depreciation. High depreciation expenses indicate recent purchases of capital assets and are negatively associated with financial performance.

Farms have tradeoffs between current profits and future profits. Many dairy farms utilize hired labor, but some may use family labor, which we captured through the opportunity cost of labor and management in our OPM and RROA numerators. Because there is a large disparity across reported hired labor expenses, we generated an indicator variable. Following Jablonski et al. (2022b), we measured hired labor expenses relative to total variable expenses. The hired labor expense indicator takes the value of one if the farm reported hired labor expense above 40% of total expenses and zero otherwise [11]. We expected this variable to have a negative association with financial performance. Government payments have previously been measured in dollars and found to increase financial performance (Katchova and Dinterman, 2018; Mishra et al., 2009). This study measures government payments as the percentage of total revenue comprised of government payments and we expected the sign of the relationship to be negative because government payments tend to be distributed during periods of low commodity prices. Government payments comprising a larger share of revenue indicates that a farm relies on outside revenue sources rather than sales of milk and other commodities, and it may signal a weak farm economy, which may be related to lower profitability. The proportion of total revenue derived from crop income is a diversity measure, and we expected it to be positively related to profitability, as diversified farms face lower economic and financial risk. Lastly, the current ratio measures the farm's liquidity, and we expected a strong current ratio to positively impact profitability.

Summary statistics and difference across farm groups

We computed summary statistics using Stata for each of the subsamples considered in the analysis (FGBF, SGBF and EF; Table 1) (StataCorp, 2021). We performed T-tests to determine whether the characteristics of the three farm groups were statistically the same on average, with the p-values for the difference in means reported in the last three columns. We winsorized the OPM, RROA and current ratio at the 1% and 99% levels to eliminate extreme outliers (Hastings et al., 1947; Ludwig-Mayerhofer, 2020) [12]. We inflated interest expense per head and depreciation expense per head values to 2021 dollars (Federal Reserve Bank of Minneapolis, 2022).

SGBF have the lowest profitability on average (Table 1). Meanwhile, EF have the strongest short-run profitability (OPM) and FGBF have the highest long-run profitability (RROA). SGBF have more operators on average (1.7) and are less likely to participate in off-farm work (57.8% of farms) than FGBF (1.2 operators, 72.9% of farms). As expected, FGBF have the smallest herd size and acres operated and they own less of their land compared to the other two groups. FGBF also have a lower proportion of farms with high hired labor expenses compared to EF and SGBF. SGBF have a small incidence of high veterinary expenses, as measured by the herd health indicator (2.1%). SGBF also have the lowest interest expense per head ($268/head) and highest percentage of revenue from government payments (6.0%).

T-test results indicate that FGBF and SGBF differ from a statistical standpoint for each of the characteristics except the OPM. Though SGBF have a lower OPM on average (9.20% versus 12.20% for FGBF) this difference is not statistically different. FGBF and EF behave differently for each of the characteristics, but they have no statistically significant differences in their OPM and off-farm income participation. SGBF and EF are the most alike, sharing a number of characteristics that have no statistically significant differences, including acres operated, hired labor expenses, depreciation expenses and the farm's current ratio. However, while many explanatory variables did not differ in this pairwise comparison, both OPM and RROA, were statistically different. This was not the case for the other two comparisons, in which only the RROA was statistically different.

Empirical results

Estimated coefficients and their differences across groups

Estimated coefficients of equation (1) for the three farm groups and two profitability measures, OPM and RROA, are presented in Tables 2 and 3, respectively. In both tables, the first column includes the estimated coefficients for the FGBF group, the second includes those for SGBF and the third includes those for EF.

When OPM is used as a profitability measure (Table 2), the number of operators appears to have an inverse relationship to profitability across all farm groups, but the estimated coefficient is not statistically different from zero for SGBF. The estimated coefficients for off-farm income show a similar pattern. For the FGBF group, the coefficient is negative and more than twice as large as that of EF (−3.4 and −1.6, respectively, and statistically different from zero). For SGBF, we find no association between the OPM and off-farm income.

Herd size has a concave relationship with OPM, and the estimates of the coefficients of the linear and quadratic terms are statistically different across farm groups, suggesting that farm size is related to short-run profitability differently across the three subgroups. The marginal effects of adding one additional cow (at each one of the subsamples' means) are 0.046, 0.102 and 0.006, respectively, for FGBF, SGBF and EF. At the margin, expanding the number of dairy cows would be the most profitable for SGBF, but these farms would start experiencing diseconomies first.

The number of acres has no statistically significant relationship with OPM across all farmer groups, with the percentage of acres owned positively related with OPM for FGBF, SGBF and EF. As expected, there is a positive relationship between yield and short-run profitability. FGBF show the largest benefit from increasing milk yield when compared to the other two groups. An inverse relationship between hired labor cost and OPM exists, and the magnitude of the estimated coefficient is largest for SGBF.

Interest expense per head and depreciation expense per head show a negative, statistically significant relationship with OPM for FGBF and EF. This relationship is positive but not statistically different from zero for SGBF. This suggests that FGBF and EF may have large debt loads or recently purchased equipment, which decrease profitability.

Estimated coefficients for government payments as percentage of a farm revenue show the expected negative sign, which is in line with the existing literature. However, the association is not statistically significant for FGBF. Crop income is positively associated with OPM for all farm groups, indicating that diversifying farms increases profitability and the estimated coefficient is highest for BF (0.498 FGBF, 0.481 SBGF, 0.269 EF). Higher liquidity ratios are associated with higher short-run profitability for the three subsamples of farmers, even though the relationship is not statistically different from zero for SGBF. Similar to the results for short-run profitability, the operator characteristics (number of operators, participation in off-farm income) are negatively associated with long-term profitability for FGBF and EF (Table 3) and show no relationship with RROA for SGBF.

Among farm characteristics, herd size shows the same concave relationship with profitability found in the OPM results. However, the relationship is statistically significant for FGBF and EF in this case, while it is not for SGBF. The marginal effects of adding one additional cow (at each of the subsamples' means) are 0.019, 0.018 and 0.005, respectively, for FGBF, SGBF and EF. A herd expansion would improve RROA the most for FGBF, even though this group would experience diseconomies first, compared to other groups. While total acreage operated does not seem to be related to RROA for FGBF and EF, it is for SGBF. The relationship between the percentage of acres owned and RROA was statistically significant for SGBF, holding the expected positive relationship.

An increase in milk yield increases the RROA for each of the farmer groups, all else being equal. Similar to the OPM results, FGBF have the largest benefit from increasing milk yield, compared to SGBF and EF (0.06, 0.05 and 0.04, respectively). The coefficient for herd health was only marginally significant for EF, holding the expected negative association.

The inverse relationship between hired labor cost and OPM is also found with RROA. The magnitude is again largest for SGBF (2.6; 1.7 FGBF, 0.9 EF). Interest and depreciation expenses per head have negative associations with RROA for FGBF and EF, but the relationship is not statistically significant for SGBF for either of these two variables. Interestingly, the negative associations for interest and depreciation expenses per head are larger for OPM than RROA, indicating that short-run profitability is compromised more than long-run profitability by investment in assets. Government payments as a percent of total farm revenue show a relationship with RROA similar to its relationship with OPM. There are negative and statistically significant associations for SGBF and EF, but the relationship is not statistically significant for FGBF. The percentage of total revenue that is crop income is associated with higher long-term profitability for all farm groups, which again supports the conclusion that diversity increases profitability. The impact on FGBF has the largest magnitude (0.2), and SGBF and EF have a similar coefficient (approximately 0.1). An increase in the current ratio is associated with a statistically significant increase in RROA for all farmers.

Decomposition results

Results of the Blinder-Oaxaca decomposition are presented in Table 4. We compare FGBF and SGBF which is our primary comparison, but also provide comparisons for these groups with EF (FGBF compared to EF, SGBF compared to EF). Comparing FGBF and SGBF are our novel contribution, and we include the comparisons of each BF type with EF allows us to contextualize findings within previous literature and provide additional information for future policy discussions. In each of the pairwise comparisons for the decomposition, “Group 1” is the reference group, which is represented by FGBF for the first decomposition and EF for the decomposition of the profitability gap between the two types of beginning farmers. The “Overall” panel of Table 4 presents the aggregate decomposition results. The results in the “Explained” section illustrate how differences in the subsample averages of each covariate contribute to the profitability gap (explained portion). Assuming a positive profitability gap, a positive (negative) statistically significant coefficient indicates that the difference in the endowment of a covariate between the two groups is associated with an increase (decrease) in the profitability gap. The bottom panel of Table 4 presents detailed results for the portion of the gap that is unexplained by differences in characteristics and can be attributed to differences in estimated coefficients. Continuing to assume a positive profitability gap, a positive (negative) statistically significant result in the unexplained component indicates that the difference in the relationship between a given covariate and profitability increases (decreases) the profitability gap.

FGBF compared to SGBF

This decomposition uses FGBF as its reference group. As noted by the pairwise t-tests (Table 1), there is no statistically significant difference in short-run profitability (OPM) among these two groups, which is again verified in the decomposition (Table 4). As FGBF and SGBF are not statistically different for the OPM, the decomposition results are not discussed, despite a marginal, statistically significant coefficient for the unexplained portion of the profitability gap.

FGBF have a 3.1% higher RROA than SGBF (7.6% versus 4.5%). The difference is statistically different from zero and is driven by the size of the asset base which differs between beginning farmer groups. Specifically, SGBF have a larger asset base than FGBF resulting in a larger denominator in the RROA for SGBF. The RROA difference is equally explained by differences in endowments (“explained” component) and differences in estimated coefficients (“unexplained” component).

If FGBF and SGBF had the same levels of observed characteristics, half of the difference in RROA (1.6% of 3.1%) would be eliminated. Herd size, milk yield, depreciation expense per head and the number of operators contributed the most to the endowment gap. If FGBF and SGBF employed the same number of operators, then the profitability gap would increase by 0.74%, all else being equal. Characteristics found to increase the profitability gap holding positive coefficients include the number of operators, acreage, herd health concerns, hired labor expense, depreciation expense per head and percentage of total revenue that is crop income. Meanwhile, observed characteristics such as off-farm income, number of cows, milk yield and interest expense per head decreased the RROA inequality between the groups.

The unexplained portion of the profitability gap captures the difference in coefficients or returns between the groups, as well as unobservable characteristics. The differential effect of the number of operators was the largest contribution, closely followed by the percentage of total revenue from government payments, depreciation expense per head and off-farm income. The negative and statistically significant coefficient (−3.9) for the number of operators indicates that the return FGBF achieve for each operator is lower than SGBF. Specifically, each additional operator is associated with 3.9% lower RROA for FGBF compared to SGBF (at SGBF average number of operators). This suggests that centralized decision-making, that is, a smaller number of operators is highly preferable for FGBF (compared to SGBF) and it may lead to relatively higher RROAs, which validates that FGBF have on average one operator while SGBF have multiple (Table 1). Similarly, off-farm income participation adversely impacts FGBF more than SGBF. If the difference in returns from off-farm income is eliminated, the RROA gap would decrease by two-thirds, or 2.2%. Differences in the returns of the percentage of acres owned is also related to a lower RROA gap of −1.4%, at the SGBF average, which is mainly driven by the positive association between ownership of land and RROA for SGBF. The relationship between RROA and interest expense per head for FGBF and SGBF is associated with lower RROA gaps because of the large negative and statistically significant relationship between RROA and this expense in FGBF (Table 3). This negative relationship and its impact is also found with depreciation expense per head. Given the strong negative relationship between the percentage of government payments over total sales for SGBF (Table 3), we find that the RROA will increase during years when farms receive a larger portion of their revenue from government payments or experience worse economic conditions. At the SGBF average, the different returns for this variable account for (ceteris paribus) 90.7% of the overall RROA gap (estimated coefficient of 2.8). The difference in the percent of crop income coefficients for the two groups of farms accounts for about one-fifth of the overall RROA gap, suggesting that a diversified portfolio can be a more effective strategy for FGBF than SGBF to increase RROA.

FGBF compared to EF

EF are the reference group for this sub-analysis. The average OPM is 12.2% for FGBF and 13.3% for EF, and this difference is not statistically different from zero, which is consistent with the t-test results (Table 1). Although the OPM difference between the two groups is not statistically different from zero, the explained portion is statistically significant, indicating that the groups differ in endowments, or observed characteristics.

FGBF had an RROA that was 1.2% higher than EF, and both groups achieved a fair return (CFFM, 2022a). The difference in RROA between FGBF and EF was statistically different from zero. Despite these two groups showing different operator, farm, herd and financial characteristics on average (Table 1), the difference in RROA does not appear to be driven by the difference in endowments. Rather, differences in farm behavior (returns), that is, the unexplained component drives the difference. The positive and statistically significant coefficient of the number of operators suggests that the return FGBF achieve for each operator are much higher than EF, and due to the RROA gap between FGBF and EF being negative (−1.2, Table 4), this results in a 1.7% lower RROA for FGBF compared to EF (at FGBF average number of operators). Off-farm income participation negatively impacts FGBF more than EF. EF operate farms that are larger than FGBF on average (143 vs 90 cows, Table 1), but FGBF experience higher returns compared to EF. Also, milk yield difference in returns alone (evaluated at the FGBF yield) would account for an RROA gap three times as large as the estimated average. Differences in returns for herd health also contribute to the RROA gap between FGBF and EF, however, its contribution is limited (−0.2, or about one-seventh of the estimated gap). The relationship between RROA and depreciation expense per head for these two groups of farmers is large enough to be associated with larger RROA. Like the analysis for FGBF and SGBF, the large negative relationship between this expense and RROA in FGBF drives this finding, and this result is consistent for interest expense per head as well. The difference in the percentage of crop income coefficients for the two groups of farms accounts for more than two-thirds of the RROA gap.

SGBF compared to EF

Here, EF are the reference group. SGBF and EF have statistically different performances for both short-run and long-run profitability that are driven by explained factors. The average OPM was 9.2% for SGBF and 13.3% for EF, which is a 4.1% gap. There are no statistically significant impacts from the unexplained components; the groups only diverged because of differences in endowments. If they had the same endowments, there would be no difference in their short-run profitability. The variables that increased the OPM gap include the number of operators, percent of acres owned, percent of total revenue derived from government payments and percent of total revenue derived from crop income. The percentage of total revenue derived from government payments and the percentage of total revenue derived from crop income accounted for 14% (0.577/4.197) and 33% (1.388/4.197) of the gap in the explained portion of the operating profit margin. Meanwhile, if SGBF and EF had the same off-farm income participation, herd size, milk yield and interest expense per head, the OPM gap would decrease by 0.2%, 0.4%, 1.6% and 0.2%, respectively.

Long-run profitability results for SGBF and EF are largely the same as they are for short-run profitability. SGBF had an RROA of 4.5% and EF had an RROA of 6.3%. The difference was statistically significant and driven by the explained portion. Milk yield and crop income had the greatest contribution to the difference but opposing impacts on the gap. If SGBF and EF had the same milk yield, the gap would decrease by 0.7%, all else being equal, which is a 34% reduction in the gap of the explained portion. Similarly, if SGBF had a more diversified revenue portfolio such that the percent of revenue derived from crop income matched that of EF, the gap would increase by 0.6%, all else being equal, which is a 32% gap increase. The number of operators, percent of total revenue derived from government payments and percent of total revenue derived from crop income increase the gap, while off-farm income participation, herd size, milk yield, herd health concerns and interest expense per head decrease the gap.

Discussion

The results presented above have some important policy implications. Specifically, this study supports a more nuanced understanding of how access to low-interest credit via BF loans affects the comparative performances of FGBF and SGBF, along with another group, EF.

We find no statistically significant differences in average OPM between FGBF and SGBF, but the groups have a statistically different RROA. FGBF have an average herd size of 90 cows and operate approximately 256 acres, while SGBF have an average herd size of 198 cows and 473 acres. This difference is captured in the denominator of the RROA as average assets, and lower asset levels for FGBF would push the RROA downwards. Even though the difference in farm size contributes to the difference in RROA, it is important to emphasize that composition (explained) and coefficient (unexplained) effects each explain about half of the RROA gap.

Explaining the contribution of the differences in endowments is straightforward. Given that SGBF in our sample operate dairy farms with a herd size about 2.2 times larger than FGBF, the fact that reducing the size of their operations to the level of FGBF would reduce the endowment effect to almost zero, ceteris paribus, is not surprising. One of the most interesting results is that, the number of operators, off-farm income, percentage of acres owned and the likely different structures of the loans SGBF and FGBF acquire can account for most of the unexplained portion of the RROA gap, ceteris paribus. FGBF have negative “returns” for both the number of operators and share of off-farm income, but these factors do not seem to affect RROA for SGBF. Therefore, the difference in coefficients contributes to lowering the RROA gap. Similarly, interest expense per head and depreciation expense per head are inversely related with RROA for FGBF, but they show no statistically significant association with it for SGBF. The percentage of acres owned is positively related to RROA for SGBF and has no association for FGBF. SGBF dairy farmers in our sample operate more acres on average than FGBF (473 versus 256) and have a higher share of land ownership (45.45% versus 37.46%), resulting in an average of 215 acres owned by SGBF versus 96 owned by FGBF. It should be noted that the amount and nature of the loans these two groups of farmers have access to also differs. SGBF have larger total and per-head depreciation expenses than SGBF, which is consistent with SGBF having more assets to depreciate. Intuitively, SGBF would need to access larger loans, which is consistent with their ability to turn land ownership to their advantage relative to FGBF. Given the difference in factors related to RROA for larger SGBF compared to smaller FGBF, it appears that creating differential policies that cater to the different needs of these two groups of BF may be worth considering. This work has demonstrated that the initial capital endowments for FGBF and SGBF differs. Specifically, it is highly likely that FGBF have limited to no preexisting physical infrastructure, while SGBF, by definition, are taking over an already established farm which has a collateral base. Thus, it is possible to envision policies tailored to the needs of FGBF which may focus on capital asset management and procurement, given that interest and depreciation expense have a large negative impact on FGBF's profitability. Jointly, SGBF operations are supporting multiple operators and tend to be larger in size, therefore, policies focused on supporting labor management and transition between generations may be more of a benefit to SGBF compared to a capital asset management and procurement focus.

Another noteworthy finding is that SGBF and EF differ in both short- and long-run profitability, with EF being more profitable. The gap in profitability being driven by the endowment effects lead to the question of whether the established farms that SGBF took over were less profitable than the average EF. As shown in Table 1, SGBF have higher yields and larger herd sizes than EF. Given the important role that these two characteristics have in determining the gap in profitability, it is likely that as SGBF took over management of the farm new practices focused on increasing efficiency, changing the herd size, or producing value-added were adopted to remain competitive in the industry. Interestingly, SGBF show the same depreciation expense per head as EF, but lower interest expense, which may be due to their ability to access low-interest financing because of their BF status. Based on our results, it is unclear whether changing the number of operators and accessing credit would result in better profitability for this group of farmers. However, these results and the comparison of FGBF and SGBF indicate that it is possible that changing operators may be a strategic decision for EF wishing to expand their farm while transitioning it to the next generation since SGBF will have access to lower-interest financing to facilitate operation expansion. If this is the case, having specific policies to facilitate transition of farming operations to new operators would be an effective approach that would alleviate the need to rebuild infrastructure. These policies could operate in lieu of, or in addition to, current BF programs, focusing more on larger, long-term loans to facilitate capital asset replacement and modernization of the operation, rather than operating and microloans, which may provide short-term financial relief to new entrants in agriculture.

Concluding remarks, limitations and future research

As the number of US dairy farms continues to decline, it is important to understand what drives the profitability of new farm operations, and whether it is more profitable to enter agriculture by means of a new operation or to operate an established farm. In this analysis, we used an unbalanced panel of Minnesota dairy farms to estimate the profitability gap between BF operating ex novo (First Generation Beginning Farmers – FGBF) and those operating on a preexisting farm (Second Generation Beginning Farmers – SGBF).

Overall, our findings indicate that the path to economic viability for BF differs depending upon the initial status of their operations. Results indicate that FGBF have a higher RROA than SGBF and EF. The differences between FGBF and SGBF are driven by both their different endowments (characteristics) and differences in returns, but the differences between FGBF and EF are due to differences in returns. EF outperform SGBF in short- and long-run profitability, and the profitability gap is due to differences in the characteristics of the operations. When considering the sources of differences in RROA between FGBF and SGBF in detail, it appears that their financial long-term commitment to dairy farming (which may be seen in the differences in share of off-farm income and the positive association between RROA and acreage owned for SGBF) may be one of the fundamental differences between these two types of farms. Previous research indicates that being located in areas with more off-farm income opportunities (Hartarska et al., 2022) and renting land instead of owning it (Stevens and Wu, 2022) may be positively related to the performance and overall well-being of BF, even though these results may only apply to the “average” BF (SGBF are only 13% of all BF in our sample). Our results indicate that for BF taking over a preexisting farm, such finding may not apply.

This analysis has two limitations. First, farmer participation in the FBM program is not random, and self-selection in the data collection may lead to sample-selection bias. However, BF in Minnesota receiving the Beginning Farmer Participant scholarship have been required to participate in the FBM program and data collection since 2014, so selection bias is more likely to occur for the EF group. If that is the case, it is likely that EF contributing to FINBIN are more likely to outperform other EF farms and may show better performances than average. As a result, the estimated performance gap between EF and BF is biased upwards. Given differences in performance between EF and FGBF that are either not statistically significant (OPM) or indicate that FGBF outperforms EF (RROA) and the expected direction of EF sample selection bias, our results should not be affected. However, it is possible that the results of the SGBF versus EF decomposition indicate an upper bound of the performance gap. Future research could repeat our analysis using data collected with a more appropriate sampling strategy. Second, because of data limitations, some of the operator characteristics often found in farm profitability analyses, such as the education level of the principal operator, are excluded from the model. It is possible that the estimated parameters of the group-specific OLS regressions may be affected by omitted variable bias, but there is no reason to believe that the coefficients obtained for one group of farmers will be affected more (or in an opposite direction) than the others. Thus, while some of the “explained” or “endowment effect” portion of the decomposed profitability gap may be affected for each pairwise comparison, the “unexplained” or “composition” component of the gap is likely unaffected. Future research could determine the magnitude of omitted variable bias in farm profitability studies, making it possible to reconcile estimates obtained from different datasets that may not share the same richness of variables.

Despite these limitations, this study suggests that the three groups of farmers (FGBF, SGBF and EF) have differences in their profitability and, within each group, different characteristics drive the differences. Understanding the different needs of FGBF and SGBF is important for the longevity of the dairy industry by allowing farmers with and without an agriculture background to have long-term success.

Figures

Figure 1

Number of licensed dairy farm operations, 2010–2022

Table 1

Summary statistics and pairwise t-tests results for variables used in the estimation

	Sample averages			T-test p-values
	FGBF	SGBF	EF	FGBF vs SGBF	FGBF vs EF	SGBF vs EF
Dependent variables
OPM (%)	12.199	9.203	13.279	0.104	0.137	0.018
RROA (%)	7.563	4.473	6.346	0.000	0.000	0.002
Explanatory variables
Operator characteristics
Operators	1.226	1.661	1.401	0.000	0.000	0.000
Off-farm income	0.729	0.578	0.718	0.000	0.385	0.000
Farm characteristics
Cows (1,000s)	0.090	0.198	0.143	0.000	0.000	0.000
Cows2	0.016	0.079	0.058	0.000	0.000	0.102
Acres (100s of acres)	2.562	4.732	4.471	0.000	0.000	0.287
Acres owned (%)	37.461	45.450	50.829	0.007	0.000	0.053
Herd characteristics
Milk yield (cwt)	187.270	215.220	198.377	0.000	0.000	0.000
Herd health	0.087	0.021	0.070	0.000	0.044	0.000
Farm financial characteristics
Interest expense per head ($1,000)	0.318	0.268	0.369	0.004	0.000	0.000
Depreciation expense per Head ($1,000)	0.296	0.387	0.396	0.000	0.000	0.601
Hired labor	0.250	0.427	0.417	0.000	0.000	0.783
Govt payments (%)	4.234	6.034	4.896	0.000	0.000	0.004
Crop income (%)	7.542	4.899	10.066	0.027	0.000	0.000
Current ratio	2.952	5.108	4.059	0.013	0.000	0.217
Obs	1,289	192	8,583

Note(s): Data are from Minnesota dairy farms from 1997 to 2021. p values of pairwise t-tests for the differences in means are reported. T-tests performed correct for the unequal variance of the variables in the different farm subgroups

Table 2

Equation (1) Regression results – dependent variable: OPM

	Estimated coefficients
	FGBF	SGBF	EF
Operator characteristics
Operators	−3.455***	−3.443	−0.897***
	(1.115)	(2.413)	(0.324)
Off-farm income	−3.400**	1.804	−1.554***
	(1.460)	(3.646)	(0.469)
Farm characteristics
Cows (1000s)	53.572***	160.596***	6.818***
	(17.760)	(44.346)	(2.505)
Cows2	−40.746***	−148.371***	−1.862**
	(15.640)	(40.465)	(0.873)
Acres (100s of acres)	−0.402	−0.832	0.067
	(0.405)	(0.917)	(0.075)
Acres owned (%)	0.090***	0.133***	0.061***
	(0.017)	(0.041)	(0.006)
Herd characteristics
Milk yield (cwt)	0.160***	0.073*	0.096***
	(0.018)	(0.041)	(0.005)
Herd health	2.605	1.499	−1.154
	(2.211)	(11.886)	(0.762)
Farm financial characteristics
Interest expense per	−11.340***	4.064	−1.969**
Head ($1,000)	(2.884)	(7.754)	(0.772)
Depreciation expense	−18.069***	8.476	−8.164***
per head ($1,000)	(3.014)	(7.247)	(0.694)
Hired labor	−5.306***	−7.210**	−2.517***
	(1.711)	(3.459)	(0.452)
Govt payments (%)	−0.114	−1.979***	−0.508***
	(0.197)	(0.616)	(0.063)
Crop income (%)	0.498***	0.481***	0.269***
	(0.060)	(0.123)	(0.017)
Current ratio	0.265**	0.164	0.158***
	(0.108)	(0.129)	(0.025)
Constant	−9.169	−9.942	−3.428*
	(6.390)	(14.001)	(1.932)
N	1,289	192	8,583
R2	0.276	0.540	0.252

Note(s): Data are from Minnesota dairy farms from 1997 to 2021. Coefficients for year dummy variables are not presented but are included in the model. “*”, “**”, “***” indicate significance at the 10%, 5% and 1% levels, respectively. Standard errors are in parentheses

Table 3

Equation (1) Regression results – dependent variable: RROA

	Estimated coefficients
	FGBF	SGBF	EF
Operator characteristics
Operators	−1.703***	0.652	−0.324**
	(0.484)	(0.826)	(0.126)
Off-farm income	−2.012***	1.779	−0.582***
	(0.634)	(1.248)	(0.183)
Farm characteristics
Cows (1000s)	22.184***	24.993	4.846***
	(7.716)	(15.183)	(0.976)
Cows2	−17.231**	−16.837	−1.086***
	(6.795)	(13.855)	(0.340)
Acres (100s of acres)	−0.284	−0.525*	−0.024
	(0.176)	(0.314)	(0.029)
Acres owned (%)	0.002	0.032**	−0.001
	(0.008)	(0.014)	(0.002)
Herd characteristics
Milk yield (cwt)	0.060***	0.054***	0.039***
	(0.008)	(0.014)	(0.002)
Herd health	1.440	−2.780	−0.558*
	(0.961)	(4.070)	(0.297)
Farm financial characteristics
Interest expense per	−6.918***	−2.342	−1.668***
Head ($1,000)	(1.253)	(2.655)	(0.301)
Depreciation expense	−9.556***	−2.346	−4.132***
per head ($1,000)	(1.309)	(2.481)	(0.270)
Hired labor	−1.687**	−2.602**	−0.863***
	(0.744)	(1.184)	(0.176)
Govt payments (%)	0.031	−0.437**	−0.152***
	(0.085)	(0.211)	(0.025)
Crop income (%)	0.237***	0.102**	0.121***
	(0.026)	(0.042)	(0.007)
Current ratio	0.110**	0.097**	0.057***
	(0.047)	(0.044)	(0.010)
Constant	0.107	−6.514	−0.867
	(2.776)	(4.794)	(0.753)
N	1,289	192	8,583
R2	0.294	0.532	0.302

Note(s): Data are from Minnesota dairy farms from 1997 to 2021. Coefficients for year dummy variables are not presented but are included in the model. “*”, “**”, “***” indicate significance at the 10%, 5% and 1% levels, respectively. Standard errors are in parentheses

Table 4

Blinder-Oaxaca decomposition results

	SGBF (2) vs FGBF (1)		FGBF (2) vs EF (1)		SGBF (2) vs EF (1)
	OPM	RROA	OPM	RROA	OPM	RROA
Overall
Group 1	12.199***	7.563***	13.279***	6.346***	13.279***	6.346***
	(0.699)	(0.307)	(0.219)	(0.088)	(0.219)	(0.088)
Group 2	9.203***	4.473***	12.199***	7.563***	9.203***	4.473***
	(1.793)	(0.609)	(0.699)	(0.307)	(1.793)	(0.609)
Difference	2.996	3.090***	1.080	−1.217***	4.076**	1.873***
	(1.925)	(0.682)	(0.732)	(0.320)	(1.807)	(0.616)
Explained	−0.510	1.577**	1.222***	0.070	4.197***	1.951***
	(1.774)	(0.765)	(0.342)	(0.147)	(0.874)	(0.374)
Unexplained	3.506*	1.513*	−0.143	−1.288***	−0.121	−0.079
	(2.117)	(0.811)	(0.679)	(0.294)	(1.558)	(0.542)
Explained
Operators	1.505***	0.742***	−0.157***	−0.057***	0.233***	0.084**
	(0.497)	(0.224)	(0.050)	(0.022)	(0.087)	(0.036)
Off-farm income	−0.514**	−0.304**	0.018	0.007	−0.217***	−0.081**
	(0.242)	(0.118)	(0.021)	(0.008)	(0.084)	(0.033)
Cows (1000s)	−5.796***	−2.400***	0.360***	0.256***	−0.378**	−0.269***
	(2.134)	(0.896)	(0.130)	(0.057)	(0.167)	(0.091)
Cows2	2.535**	1.072**	−0.077***	−0.045***	0.039	0.023
	(1.087)	(0.482)	(0.026)	(0.013)	(0.027)	(0.015)
Acres (100s of acres)	0.873	0.617	0.128	−0.046	−0.018	0.006
	(0.903)	(0.390)	(0.160)	(0.064)	(0.027)	(0.011)
Acres owned (%)	−0.721**	−0.015	0.815***	−0.013	0.328*	−0.005
	(0.301)	(0.058)	(0.108)	(0.033)	(0.172)	(0.014)
Milk yield (cwt)	−4.461***	−1.663***	1.069***	0.433***	−1.621***	−0.657***
	(0.860)	(0.315)	(0.139)	(0.054)	(0.370)	(0.148)
Herd health	0.172	0.095	0.019	0.009	−0.057	−0.028*
	(0.150)	(0.074)	(0.016)	(0.007)	(0.039)	(0.016)
Interest expense per head ($1,000)	−0.563**	−0.343**	−0.101**	−0.086***	−0.199**	−0.169***
	(0.257)	(0.135)	(0.047)	(0.021)	(0.093)	(0.042)
Depreciation expense per head ($1,000)	1.637***	0.866***	−0.820***	−0.415***	−0.080	−0.040
	(0.442)	(0.221)	(0.133)	(0.044)	(0.153)	(0.077)
Hired labor	0.941***	0.299**	−0.421***	−0.144***	0.025	0.009
	(0.363)	(0.150)	(0.079)	(0.031)	(0.091)	(0.031)
Govt payments (%)	0.205	−0.056	−0.336***	−0.101***	0.577**	0.173**
	(0.519)	(0.199)	(0.096)	(0.033)	(0.230)	(0.074)
Crop income (%)	1.317**	0.627**	0.678***	0.305***	1.388***	0.624***
	(0.634)	(0.292)	(0.125)	(0.051)	(0.343)	(0.146)
Current ratio	−0.571	−0.237	0.175***	0.063***	−0.165	−0.060
	(0.359)	(0.160)	(0.041)	(0.015)	(0.136)	(0.049)
Unexplained
Operators	−0.019	−3.913**	3.135**	1.690***	4.230	−1.623
	(6.058)	(1.627)	(1.316)	(0.581)	(5.826)	(1.446)
Off-farm income	−3.008	−2.191***	1.346	1.043**	−1.941	−1.365*
	(2.137)	(0.843)	(1.043)	(0.456)	(1.999)	(0.768)
Cows (1000s)	−21.205	−0.557	−4.205**	−1.559**	−30.469**	−3.992
	(13.202)	(4.367)	(1.669)	(0.702)	(12.800)	(4.105)
Cows2	8.467*	−0.031	0.640**	0.266**	11.527**	1.239
	(4.896)	(1.524)	(0.271)	(0.121)	(4.889)	(1.435)
Acres (100s of acres)	2.033	1.137	1.202	0.667	4.254	2.369
	(4.832)	(1.922)	(1.081)	(0.461)	(4.440)	(1.741)
Acres owned (%)	−1.919	−1.381**	−1.099	−0.104	−3.253	−1.507**
	(2.176)	(0.700)	(0.716)	(0.287)	(2.042)	(0.629)
Milk yield (cwt)	18.534	1.270	−11.860***	−3.838**	4.903	−3.141
	(12.606)	(3.715)	(4.263)	(1.529)	(11.788)	(3.348)
Herd health	0.023	0.088	−0.327	−0.174*	−0.055	0.046
	(0.118)	(0.094)	(0.206)	(0.098)	(0.112)	(0.083)
Interest expense per head ($1,000)	−4.133*	−1.228*	2.980***	1.669***	−1.619	0.181
	(2.384)	(0.638)	(1.117)	(0.433)	(2.204)	(0.533)
Depreciation expense per head ($1,000)	−10.260***	−2.787***	2.931***	1.605***	−6.431**	−0.691
	(3.273)	(0.934)	(0.947)	(0.404)	(3.076)	(0.782)
Hired labor	0.813	0.391	0.697	0.206	2.004	0.742
	(1.424)	(0.599)	(0.442)	(0.196)	(1.246)	(0.510)
Govt payments (%)	11.250**	2.824**	−1.666	−0.776	8.875*	1.718
	(4.884)	(1.315)	(1.293)	(0.496)	(4.588)	(1.150)
Crop income (%)	0.083	0.663**	−1.731**	−0.879***	−1.041	0.092
	(0.810)	(0.309)	(0.695)	(0.240)	(0.744)	(0.229)
Current ratio	0.515	0.067	−0.316	−0.155	−0.032	−0.201
	(0.820)	(0.349)	(0.388)	(0.179)	(0.498)	(0.178)
Constant	0.773	6.621	5.741	−0.973	6.514	5.648
	(16.164)	(5.357)	(6.521)	(2.481)	(15.128)	(4.881)
N	1,481	1,481	9,872	9,872	8,775	8,775

Note(s): Coefficients for year dummy variables are not reported. “*”, “**”, “***” indicate significance at the 10%, 5% and 1% levels, respectively. Standard errors in parentheses

Notes

1. The USDA Farm Service Agency defines a beginning farmer (BF) as any individual or entity who has not operated a farm for more than 10 years (USDA-FSA, 2022).

2. Even though most research uses the USDA-FSA definition of a BF (Detre et al., 2011; Katchova and Dinterman, 2018; Kropp and Katchova, 2011; Mishra et al., 2009), the varying BF definition creates difficulty.

3. Several previous studies have used the Blinder-Oaxaca decomposition to analyze the impact of different factors on farm financial performance. Recent examples include Sackey (2018), Le Thi Kim et al. (2021) and Fisher et al. (2023).

4. An alternative model specification included county-level fixed produced results that were substantially identical to those presented in the main text. Parameter estimates are available upon request. Additionally, given the highly unbalanced nature of the panel dataset used, as discussed in more detail below, the authors did not control for farm-specific fixed effects.

5. This interpretation of the coefficient portion of the decomposition leads to the well-known “omitted group” problem, as illustrated in Oaxaca and Ransom (1999), where it is possible to envision that the term X¯−j′(β^j−β^−j) refers to a hypothetical omitted group (o) for which X¯o=0. This issue in the interpretation of the coefficient component is particularly obvious in the case of indicator variables, where the decision to assign the value of “0” to a specific subgroup of the data is arbitrary and not dictated by economic rationale (Jones, 1983). Although there are methods that have been proposed in the literature to obviate this problem, their application is beyond the scope of the analysis.

6. FINBIN codes farms by type (crop, dairy, crop and dairy, hog, crop and hog, beef, crop and beef, sheep, crop and sheep) based on which enterprise(s) generates 70% or more of the farm's income. If no crop, livestock, or crop and livestock enterprise generates over 70% of revenue, the farm is categorized as other.

7. Beginning in 2014, FINBIN reported information about beginning farmers that received a scholarship through the FBM Beginning Farmer Program, which is facilitated by the Minnesota Department of Agriculture and Rural Finance Authority (MDA, 2022a).

8. Farmers in the SGBF group may be third- or fourth-generation farmers, but for the purpose of this research, any farmer taking over an already-existing farm has collectively been termed a second-generation beginning farmer.

9. The current calculation for the opportunity cost of the value of labor and management is $30,000 + 5% * value of farm production (CFFM, 2022b).

10. The 5% cutoff value, indicating herd health concerns, represents the value for farms performing at the lowest 10% from 1997 to 2021, in accordance with the FINBIN benchmark report. From 2016 to 2021, the average fraction of operating expenses allocated to veterinary services ranged from 4.6% to 6.2% nationally and 6.4%to 8.5% in Minnesota (USDA-ERS, 2023).

11. The 40% cutoff value, indicating high labor expenses, represents the value of farms performing at the 50% level from 1997 to 2021, in accordance with the FINBIN benchmark report.

12. Values of OPM, RROA and current ratio lower (higher) than the first (ninety-ninth) percentile of their respective distributions was replaced by the value of the first (ninety-ninth) percentile.