Results

Tables 2 and 3 compare the survival, positive growth, and success rates for small-scale beginning farms based on selected farm and operator characteristics, respectively. Among the different categories of farms considered, success rates are higher for those that receive government payments, have relatively high productivity, have DTC sales, produce grains, oilseeds, dry beans, or dry peas, or produce milk or dairy products. In terms of producer characteristics, success rates are higher for farmers who are younger, or white, or male. These statistics could be useful to policy makers and others who seek to target groups based on their business performance.

Table 2. Survival, growth and success rates by farm characteristics: small-scale beginning farms

Notes: Small-scale farms are those with less than $250,000 (constant 2018 dollars) in farm real estate assets in the initial period.

Source: USDA-NASS Census of Agriculture, 2007, 2012, 2017.

Table 3. Survival, growth and success rates by operator characteristics: small-scale beginning farms

Notes: Small-scale farms are those with less than $250,000 (constant 2018 dollars) in farm real estate assets in the initial period.

Source: USDA-NASS Census of Agriculture, 2007, 2012, 2017.

The comparisons in Tables 2 and 3 quantify success rates for groups with similar characteristics, but do not necessarily reveal how the group’s shared attributes affect farm performance since confounding factors will also vary across groups. For example, as Table 2 shows, larger farms are more likely to survive than smaller farms and farmers who receive agricultural payments are more likely to survive than farmers who do not receive payments. However, since operators of larger farms are more likely to receive payments, it is not clear from the summary statistics whether, or by how much, each factor is affecting farm survival. Probit regressions are used to determine the marginal effects of individual variables on the likelihood of farm survival, positive growth, and success, while holding other exogenous factors constant. Tables 4 and 5 show the estimated marginal effects at the means for small- and large-scale beginning farms, respectively.Footnote ¹⁰ Classification tables corresponding to the probit regressions in Tables 4 and 5 are shown in Appendix Table A1.

Table 4. Probit regression marginal effects: small-scale beginning farms

* p < 0.01;

** p < 0.001.

Notes: Small-scale farms are those with less than $250,000 (constant 2018 dollars) in farm real estate assets in the initial period.

Source: USDA-NASS Census of Agriculture, 2007, 2012, 2017.

Table 5. Probit regression marginal effects: large-scale beginning farms

* p < 0.01;

** p < 0.001.

Notes: Small-scale farms are those with less than $250,000 (constant 2018 dollars) in farm real estate assets in the initial period.

Source: USDA-NASS Census of Agriculture, 2007, 2012, 2017.

The regression results (Tables 4 and 5) are qualitatively consistent with the effects implied by the mean values displayed in Tables 2 and 3. Since the estimated coefficients obtained for small and large farms are generally similar, we focus the discussion on small-scale beginning farms and note the differences with larger operations when they are substantial. For small-scale beginning farms (Table 4), farm size is strongly correlated with survival: doubling farm assets increases the probability of surviving by about 5 percentage points.Footnote ¹¹ Larger farms may have a higher survival rate because larger farms are more profitable and have higher household income and wealth and are hence more resilient to economic shocks (Whitt, Todd, and MacDonald, Reference Whitt, Todd and MacDonald2020). While the survival rate increases with scale, among those farms that survive, farm size is negatively correlated with positive growth: doubling farm size reduces the likelihood of positive growth by about 11.6 percentage points. Larger farms might be less likely to expand because they are already closer to an optimal scale, and therefore, there is less incentive to expand production compared to smaller farms. With scale being positively correlated with survival and negatively correlated with growth, the effect of scale on success (survival and positive growth) is limited: size has a very small negative association with success.

The results also show that land tenancy arrangements matter for farm business performance. Compared to similar farmers who rent in land (the missing category), those who do not rent in land or who rent out land have survival rates that are about 4 and 5 percentage points higher, respectively. Those who rent out land might have higher survival rates because they earn rental income which can provide a financial cushion during periods of low prices or yields. In contrast, farmers who rent in land likely face higher per-acre costs in the form of rent, which makes them more vulnerable to financial shocks.

Renting out farmland is associated with a higher survival rate, but a lower probability of positive growth. Operators who rent out land may have less time to work on their farm because they have better off-farm employment opportunities. With less time available to work on-farm, farm expansion may be more difficult. Tenancy arrangements, therefore, have conflicting effects on the components of success: farmers who rent out land are more likely to survive but less likely to grow that similar farmers who rent in land. In net, the results indicate that farmers who rent out land have a business success rate that is 3 percentage points higher than similar farmers who rent in land.

The results also show that farmers who depend more on government agricultural payments tend to have better business performance. Increasing the “payment intensity” (agricultural payments as a share of payments plus the total value of production) increases survival, positive growth, and success rates. Increasing the payment share by 10 percentage points is associated with a 1.2 percentage point increase in the likelihood of success. These findings are consistent with earlier research that found that agricultural program payments are associated with an increase in the likelihood and duration of farm business survival of crop farms (Key and Roberts, Reference Key and Roberts2006, Reference Key and Roberts2007). The results of this study show that this relationship also holds for beginning farmers and that the combined effects of payments on survival and growth also increase the likelihood of farm business success.

There are several mechanisms through which agricultural payments could improve farm outcomes. Higher government payments raise net returns, which could reduce the likelihood of financial insolvency and allow farms to remain in business longer. Higher payments may also make agriculture more profitable relative to alternative occupations, which could reduce the incentive to quit farming, especially if off-farm occupations are in different locations or require a significant degree of specialization. The finding that higher payments allow farms to expand in size could mean that some farmers face binding liquidity constraints that inhibit their farms from achieving an optimal scale and that payments relieve these constraints. Commodity program payments provide cash, some degree of insurance (due to links with commodity prices), and perhaps also a means to leverage greater resources from lending institutions, all of which may lower farmers’ capital costs. Lower capital costs may allow farms to more easily adopt new technologies and expand in scale.

As expected, farm productivity (value of production relative to assets) is positively and significantly correlated with survival, growth, and success. More productive farms are generally more profitable and hence better able to cope with financial shocks and have the financial resources to expand production. The productivity index value for a small beginning farm has a mean value of 0.16 with a standard deviation of 0.50 (Table 1). The results imply that increasing the average farms’ productivity value by one standard deviation to 0.66 would raise the survival rate by 1 percentage point, the growth rate by 1.2 percentage points, and the success rate by 0.8 percentage points.

For small-scale beginning farms, using DTC markets is associated with an 8 percentage point increase in the survival rate. It is possible that DTC sales provide a relatively secure income stream, which increase survival rates. Income from DTC marketing depends on the margin between the wholesale and retail price. Even when the farmer’s input and output prices vary, the markup between the wholesale and retail prices tends to be relatively stable. Hence, the additional income that farmers can earn by selling directly to consumers should be less influenced by price fluctuations, making this income less risky.

While farmers who directly market to consumers are more likely to continue farming than those who do not, they were about 1.5 percentage points less likely to have positive asset growth. Farms with DTC sales might be less likely to expand because of market constraints. DTC markets may have a limited size within a local region, making it more challenging for farms to expand in scale without substantial additional travel and marketing costs.

Despite the small negative effect on farm size growth, the relatively large positive effect on survival causes DTC sales to have a positive association with farm business success: farms with DTC sales were about 4.6 percentage points more likely to succeed than similar farms without DTC sales. This effect on success is larger for small farms compared to large farms (Table 5). For larger operations, having DTCs resulted in smaller effect on growth and a larger negative effect on survival.

Being organized as a family farm was associated with a higher rate of survival, but a lower likelihood of farm growth. Unlike family farms, partnerships (the most common non-family farm arrangement) might lower survival rates because they are vulnerable to dissolution from any single partner. Higher survival rates and slower growth for family farms might be explained by different attitudes toward farm work by the farm operators. Researchers have found evidence that non-pecuniary benefits from self-employment explain why small business owners remain in business despite earning less income (Hamilton, Reference Hamilton2000). It is possible that non-pecuniary benefits from farming (e.g., greater autonomy, independence, and lifestyle factors) provide operators of family farms a greater incentive to remain in business even with lower business expansion possibilities, compared to non-family farms that may be organized solely for pecuniary reasons (Howley, Reference Howley2015). Because of their higher survival rates, family farms are about 2 percentage points more likely to be successful than otherwise similar non-family farms.

Commodity specialization is an important determinant of farm performance. There are substantial differences in performance rates between farms having different specializations: with survival, positive growth, and success rates varying by about 10, 18, and 12 percentage points, respectively. Holding all other characteristics constant, farms that produce “grains, oilseeds, dry beans, or dry peas” (the missing category) have the highest probability of success. Compared to that missing category, being a producer of “vegetables, melons, potatoes, or sweet potatoes” or a producer of “hogs and pigs, poultry or eggs” or a producer of “horses, ponies, mules, burros or donkeys” reduces the success rate by 10 to 12 percentage points. While there could be numerous factors, including differences in pest, climate, and market conditions that cause variation in farm performance across commodity groups, it is plausible that insurance availability plays a role. The categories associated with the highest (lowest) success probability are comprised of commodities with historically higher (lower) Federal crop insurance programs participation rates. Crop insurance may reduce the risk of default and bankruptcy and result in higher survival rates. Greater crop insurance coverage might also enhance farmers’ access to credit with which to expand production (Ifft, Kuethe, and Morehart, Reference Ifft, Kuethe and Morehart2015).

The results indicate a U-shaped relationship between the operator’s age and the likelihood of farm survival. Holding all else constant, farms with operators younger than 30 (the missing category) or over 69 have a lower likelihood of surviving than farms with operators between the ages of 30 and 69. The youngest farmers may be less like to survive because they are the most financially vulnerable to financial shocks—they tend to be more highly leveraged and have accumulated less net worth. As was discussed in the literature review, the oldest farmers probably have higher exit rates because they are more likely to retire or leave farming for health reasons.

There is a monotonic negative relationship between age and the likelihood of positive growth. Compared to having an operator under 30 years old, having an operator age 40–49 reduces a farm’s likelihood of positive growth by about 7 percentage points, while having an operator over age 69 reduces the likelihood of positive growth by about 10 percentage points. Older beginning farmers start farming later in life so have had more time to accumulate net worth. Consequently, older farmers may be able to start a farm closer to their desired scale and so they would have less incentive to expand farm size. Older farmers also have a shorter lifetime investment horizon, so may find it less advantageous to incur the sunk costs associated with increasing their farm size.

Operator age does not have a large effect on the probability of success except when the operator is 70 or older. Farms with these older operators are about 7.5 percentage points less likely to be successful compared to farms with younger operators. Farms having the oldest operators have a lower likelihood of success because they are less likely to survive in business and less likely to have positive growth.

Whether or not the principal operator had farming as his or her main occupation had no significant effect on the likelihood of farm success. In contrast, if the farm household is more dependent on farm income, then the farm business is somewhat more likely to survive, to have positive growth, and to succeed. Increasing the farm’s share of household income by 10 percentage points increases the likelihood of success by about 0.8 percentage points. Being more dependent on farm income might incentivize the operator to provide additional effort and attention to assuring the success of the farm.

Several demographic characteristics of the principal operator have a significant influence on farm business performance. Having a female operator is associated with an 8.5 percentage points reduction in the survival probability, a 4-point reduction in the likelihood of positive growth, and an 8-point reduction in the likelihood of success, holding other observable factors constant. The race of the primary operator had a relatively small effect on the farm outcomes. Non-white or Hispanic operators had success rates that were about 2–4 percentage points lower than similar farms with white operators.

The year indicators are included in the model to adjust for aggregate changes in the national economy and to adjust for changes in the Census methodology that could have affected the way farmers were tracked over time. The coefficients indicate that the likelihood of farm survival, positive grow, and success was 3–4 percentage points higher in 2012–17 compared to 2007–12. This is somewhat counterintuitive since net cash farm income decreased substantially over the later period. However, the Great Recession occurred during the 2007–12 period, and unemployment rates were unusually high. There is substantial evidence that the nonfarm economy is an important determinant of farm bankruptcies (Dinterman, Katchova, and Harris, Reference Dinterman, Katchova and Harris2018; Shepard and Collins, Reference Shepard and Collins1982). This is reflected in farm Chapter 12 bankruptcy rates, which increased substantially from 2008 to 2011 (Dinterman, Katchova, and Harris, Reference Dinterman, Katchova and Harris2018). Beginning farmers are more reliant on off-farm employment for household income than are more experienced operations (Key and Lyons, Reference Key and Lyons2019). Contraction in the macroeconomy may have contributed to higher farm exits and slower farm growth in the wake of the Great Recession.

The importance of the nonfarm economy is also reflected in the positive coefficients associated with the county employment change. An increase in the county employment rate indicates improving off-farm employment opportunities. While the results do not indicate that changes in the local employment rate affect farm survival, an improving local job market is associated with a greater likelihood of positive farm growth rates and farm business success. This suggests that farmers are more inclined, or better able, to expand their operations when the local economy is improving. This finding seems to contradict the hypothesis that more rapid local economic growth “pushes out” farming and discourages farm growth. Instead, local job growth may allow for greater off-farm employment and higher household income, which can lead to more on-farm investment (Key, Reference Key2020). Local economic growth might also be associated with expanding local and DTC agricultural market opportunities.

In terms of location, having a farm in the Heartland and Northern Crescent regions is associated with significantly higher survival rates; while being in the Northern Great Plains, Prairie Gateway, Fruitful Rim, or Basin and Range regions is associated with higher probabilities of growth. Interestingly, there was no regional overlap between high survival and high positive growth rates. In fact, the regions associated with the highest rates of positive growth had among the lowest survival rates. It seems plausible that at the regional level, farm survival and growth are in tension. Farms need additional land in order to expand and the main source for this land is farms that are exiting production. Hence, in regions where there are relatively few farms exiting, fewer farms will be able to grow.

In terms of success, the results show that the regional effects on survival rates outweighed the effects on growth. Being located in the Eastern Upland, Southern Seaboard, and Fruitful Rim regions reduced a farm’s success rate by 2–3 percentage compared to being located in the Heartland, Northern Crescent, Northern Great Plains, Prairie Gateway, or Basin and Range regions. Being located the Mississippi Portal reduced a farm’s success rate by about 6 points compared to the same set of regions. Further research is needed to understand what factors (not controlled for in the model) are responsible for these differential regional effects.

As a robustness check, we explore how the growth and success probit regression results change using total value of production instead of real estate asset value as a measure of farm size (Appendix Table A2). As can be seen by comparing the results to Tables 4 and 5, the coefficients are very similar in magnitude and in statistical significance. In fact, out of the 76 coefficients shown in Table 2, only 1 statistically significant coefficient changed sign when using the alternative farm size measure.

We conclude this section by noting that the regression estimates could potentially suffer from selection bias if unobserved factors correlated with the outcomes were not adequately controlled for by the observable farm and operator characteristics. For example, the estimated relationship between payments and farm success could be biased if missing variables affect both the likelihood that farmers received agricultural program payments and the likelihood of business success, and if these missing variables were not sufficiently correlated with the observable variables. Future work could test the robustness of the estimates using experimental, or more feasibly, quasi-experimental econometric techniques.