WHO REALLY BENEFITS FROM AGRICULTURAL SUBSIDIES? EVIDENCE FROM FIELD-­‐LEVEL DATA Barrett Kirwan University of Maryland

Michael J. Roberts North Carolina State University February 2010 [Preliminary and Incomplete] ABSTRACT The idea that agricultural subsidies are fully capitalized into farmland values forms the foundation of the argument that subsidies are entitlements and removing them would drastically reduce farmland asset values. Surprisingly little evidence substantiates this claim. Using field-­‐level data and explicitly controlling for potentially confounding variables we find that landlords only capture between 14 – 24 cents of the marginal subsidy dollar. The duration of the rental arrangement has a substantial effect on the incidence. Initially, landlords extract 44 cents of the marginal subsidy dollar, but the incidence falls by 1.5 cents with each additional year of the rental arrangement. This duration effect reveals that rental market frictions play an important role in the farmland rental market.

Introduction In the twenty-­‐first century, subsidized American farmers have gleaned, on average, $8,824 annually in subsidy payments, making agricultural subsidies one of the largest per-­‐capita transfer programs in the U.S. Although originally motivated by equity concerns, farm subsidies today are generally considered entitlements.1 According to conventional wisdom, since subsidies are fully capitalized into farmland values, the first generation of subsidy recipients reaped a windfall gain in the form of higher asset values, while all subsequent generations have purchased the “right” to agricultural subsidies by purchasing subsidy-­‐inflated farmland.2 Recent evidence from farm-­‐level data, however, indicates that only 23 to 64 cents of the marginal subsidy dollar gets capitalized into farmland values (Goodwin et al 2005; Kirwan 2009). These estimates weaken, and at the low end of this range effectively eliminate, a primary rationale supporting transfer payments to American farmers. The fundamental unit in the theory of agricultural subsidy incidence is the plot of

land being transacted, i.e., the field. This paper augments recent farm-­‐level analysis focused on subsidy capitalization by using nationally representative field-­level data to test the theory at the ultimate level of analysis. These unique field-­‐level data pair a field’s subsidy rate with its rental rate. Farm-­‐level analysis, in contrast, pair the farm-­‐average rental rate to the farm-­‐average per-­‐acre subsidy. Although previous analysis (Roberts et al 2005, Kirwan 2009) used farm-­‐level longitudinal data to control for farm-­‐level unobserved

1 Agricultural subsidies began in 1933, when average farm income was 32% of average non-­‐farm income (Gardner, 1992), and aimed to give “agriculture a fair share in the national income” (Nourse et al 1937). By 1970 farm and non-­‐farm income had essentially converged. 2 Orden, Paarlberg, and Roe (1997) provide a historical perspective on the evolving justification for farm subsidies, emphasizing the ‘entitlement’ argument.

2

heterogeneity, the incidence question fundamentally calls for field-­‐level data to relate changes in the subsidy per acre to changes in the per-­‐acre rental rate. The farm-­‐level incidence estimate may be biased downward if farmers own most of their subsidized acres and rent unsubsidized acres. In the extreme case where no subsidized acres are rented, but the farmer owns subsidized acres, a farm-­‐level analysis would estimate no relationship between subsidy changes and rental rate changes, even if the implicit rental rate of subsidized acres rises one-­‐for-­‐one with the marginal subsidy dollar This paper overcomes the farm-­‐level aggregation issue by bringing field-­‐level data

to bear on the question. Using a nationally representative sample of soybean and rice fields, we find, similar to Kirwan (2009), that landlords extract less than one-­‐quarter of the marginal subsidy dollar through higher rental rates. These findings provide further evidence that subsidies are not fully capitalized into the land values and, consequently, are not entitlements. The findings also restore confidence in the usefulness of farm-­‐level data to address important policy questions. This paper also attempts to explain why theory and evidence diverge substantially by examining the institutions surrounding the farmland rental market. We find evidence that landlords are typically retired and often depend heavily on the income stream provided by the farmland, as if it were an annuity. Evidence of thin rental markets, i.e., few tenants, combines with landlords’ inelastic demand for tenants to provide the marginal tenant with substantial bargaining power.

The paper is structured as follows. First, we search the historical literature for clues to answer the incidence question, and we highlight the importance and novelty of this approach by contrasting it with the previous research. We explain the advantages of farm-­‐

3

level data in the context of past research. Next we examine the available evidence on the institutional structure of the farmland rental market. We then develop our empirical model and perform the analysis, and finally we interpret the results in light of the institutions surrounding the farmland rental market. The Incidence Question A common weakness of the literature is the inability to identify the effect of

subsidies. This identification problem arises from estimating a fundamentally unidentified system. Consider the standard workhorse model of farmland value determination: the present value model (e.g., Melichar, 1979; Robison et al., 1985). Many unknowns characterize the present value model. The unknowns of greatest concern are the expected subsidy stream, the discount rate, and the proportion of the subsidy that becomes capitalized into the land value, i.e., the incidence. The present value model, however, is a single equation. As a system, the present value model of land price determination is underidentified. No parameter of this system can be identified without further information or restriction on the other unknown parameters. Typically, investigators assume (often implicitly) that the entire subsidy dollar gets capitalized into the land value, thereby restricting the incidence parameter to equal one. Ultimately, investigators have left unanswered the most policy-­‐relevant question: what proportion of the marginal subsidy dollar is capitalized into land values? The answer to the question confirms or repudiates the conventional wisdom that the landowner primarily benefits from agricultural subsidies; it clarifies who benefits from agricultural subsidies and how much they benefit; and it supports or discredits the contention that subsidies are entitlements. The incidence question lays the groundwork for welfare

4

analysis of agricultural subsidies and illustrates the distribution of subsidies. Knowing the incidence of the marginal subsidy dollar enlightens our understanding of the political process by illuminating the value transferred to key benefitting constituencies. To study the incidence question, one must appropriately deal with expected subsidies and the discount rate. Roberts et al. (2003) and Kirwan (2009) demonstrate how to do this by focusing on the farmland rental market. Because rental rates are per-­‐period prices reflecting the productive value of agricultural land, this approach avoids making assumptions about expectations or the discount rate over future periods. By focusing on farmland rental rates, these investigators can cleanly estimate the incidence without relying on strong assumptions about other key parameters. The Farmland Rental Market In the U.S., farmers rent 360 million acres of farmland, an area equal to 38 percent of all farmland and comparable to all the farmland in the 13 Corn Belt states.3 Table 1 reports statistics on the farmland rental market from the 1992, 1997, and 2002 Censuses of Agriculture. In 2002, farmers paid $8.7 billion in cash rent, a 47% increase from 1992. This expense accounted for 11.3 percent of total production expenditures by renters. At the same time the number of tenants dropped 25 percent while their farm sizes increased 11 percent.

The importance of the farmland rental market cannot be overstated. The farmland rental market is also an important market for subsidized land. The

1999 Agricultural Economics and Land Ownership Survey (AELOS) reveals that farmers rent a majority of cropland acres, which are the only type of land that is subsidized. Data

3According to the 2002 Census of Agriculture, there are 361.7 million acres of farmland in Iowa, Indiana,

Illinois, Ohio, South Dakota, Nebraska, Kansas, Minnesota, Wisconsin, Michigan, Missouri, Kentucky, and North Dakota.

5

from the 1996 Agricultural Resource Management Survey (ARMS) demonstrate that 45 percent of subsidized acres are rented.

According to the conventional wisdom, landowners are the primary beneficiaries of agricultural subsidies. In the U.S., non-­‐farmer landlords own 87 percent of rented farmland.4 Since the vast majority of rented, subsidized land is likely owned by non-­‐ farmers, the conventional wisdom implies a significant share of agricultural subsidies leaves the agricultural sector and accrues to non-­‐farmer landlords. A New Approach This paper overcomes the shortcoming of previous work by using novel, acreage-­‐

level data to estimate the incidence of agricultural subsidies on farmland rental rates. In 2006, the USDA’s Agricultural and Resource Management Survey (ARMS) phase II survey of soybean and rice fields elicited information on rent paid and subsidy payments received on all cash-­‐rented parcels of land. The survey also obtained information on expected yields and historical rotations, which serve as powerful controls for land quality. The data are unique in that they link subsidies to the specific cash-­‐rented parcels to which the subsidies are tied. Table 2 contains summary statistics from these data. Table 2 illustrates some differences between soybean fields and rice fields. Notably,

the average rice-­‐field rental rate is 30 percent higher than that of the average soybean field. This reflects both the higher yield, 41.6 bushels/acre of soybeans versus 72.7 bushels/acre of rice, and an historically higher price. Rice acres receive substantially higher subsidies, too. At nearly $60/acre, the average rice subsidy is 3½ times greater than the $17/acre average soybean subsidy. Further contrasts are that the rice field is almost twice as likely

4 The 1999 AELOS data reveal that non-­‐farmer landlords also own 55 million farmland acres that are not rented out and, presumably, are left unfarmed. 6

to be harvested for seed, and the crop rotation for rice fields is evenly spread between a rice-­‐soybean rotation, continuous rice, and a rice-­‐fallow rotation. In contrast most soybean fields are in a corn-­‐soybean rotation. Overall, the summary statistics suggest that rice and soybean fields are sufficiently distinct to warrant separate analysis. Empirical Strategy The model we use to estimate the incidence of agricultural subsidies on farmland rental rates is (1)

,

where ri is the rental rate for field i. The per-­‐acre Direct Payment subsidy is gi,. Xi is a vector of observable covariates, including field-­‐level costs and returns. Regional differences in production practices are accounted for by fj, a fixed effect for region j. In the U.S., agricultural subsidies and farmland rental rates are mechanically

connected. Agricultural subsidies are a function of program yield, a parameter that reflects the underlying productivity of the subsidized field. Since more productive fields command a higher rental rate, the simple correlation between subsidies and rental rates not only reflects the causal effect of subsidies, but it also incorporates a spurious, positive mechanical relationship, resulting in upwardly biased incidence estimates. Typically, a field’s underlying productivity is unobserved, and the econometrician

has to resort to other methods to account for the spurious upward bias, e.g., fixed effects. The 2006 ARMS, however, explicitly asked farmers for their expected yield on the field. Presumably the farmer knows the field’s underlying productivity and bases his answer on this, thus providing an explicit measure of each field’s underlying productivity. Armed with this information, we account for the spurious mechanical relationship by explicitly

7

controlling for each field’s underlying productivity. We thereby isolate the causal effect of subsidies on farmland rental rates. Results—Rental Rates Incidence In light of the differences between rice and soybean fields illustrated in table 2, we

initially perform the analysis separately for each crop. Results from this analysis are reported in tables 3 and 4. Table 3 contains the results using the sample of soybean fields to estimate equation 1. The first column reports the results of a simple bivariate regression of rental rates on subsidies. At the field level, before controlling for any field characteristics we estimate an incidence of 0.45; in other words, rental rates increase by 45 cents with the marginal subsidy dollar. Even the simple correlation rejects the conventional wisdom that the landlord extracts a substantial portion of the marginal subsidy dollar. Controlling for observable characteristics of the soybean fields lowers the incidence estimate to 0.25. Accounting for region-­‐specific unobserved heterogeneity with region fixed effects lowers the estimate to 0.24. Interestingly, this estimate is nearly identical to that found by Kirwan (2009) using farm-­‐level data. Table 4 contains the results for the sample of rice fields. Here the story is very much

the same. The bivariate relationship in column one is very different than the conventional wisdom, which posits the coefficient should equal 1. Here, the coefficient is 0.16, but it is statistically indistinguishable from zero. Adding covariates to control for field level observable characteristics provides a statistically significant estimate of 0.19. Including region fixed effects results in a still-­‐significant 0.14. This estimate is substantially smaller than the soybean-­‐subsidy incidence, suggesting that differences in the farmland-­‐rental market structure could help explain the lower-­‐than-­‐expected incidence estimates.

8

Although the soybean and rice farmland rental markets appear to be somewhat

different, both reveal that landlords capture a surprisingly low portion of the marginal subsidy dollar. Table 5 combines the two crops into a single analysis. The bivariate relationship reveals a 0.27 estimate; the estimate falls to 0.14 after controlling for observable field characteristics, and ultimately the combined analysis yields an incidence estimate of 0.08. In any case, the conventional wisdom that landlords extract a substantial share of the marginal subsidy dollar appears to be wrong. Results—Net Returns The analysis above establishes that 75 – 85 cents of the marginal subsidy dollar are

not extracted by the landlord and are available for the tenant. Whether the remaining subsidy accrues to the tenant or another factor of production owner, however, is an empirical question. Following the identification strategy outlined above, we examine the effect of the marginal subsidy dollar on tenants’ net returns, controlling for field quality and farming practices. Column 4 of table 4 reports the results for soybean fields. Sixty-­‐ three of the remaining 76 cents are retained by the tenant farmer. Contrary to conventional wisdom and standard economic theory, the vast majority of the marginal subsidy dollar stays with the tenant farmer. Discussion One of the fundamental assumptions behind the standard model of economic incidence is perfect competition. Yet the farmland rental market might not be perfectly competitive. Kirwan (2009) found evidence that tenants have some market power. Market power may arise as farms grow larger and distance between parcels embodies transactions costs that could limit local competition. Additionally, Young and Burke (2001) hypothesize

9

that social norms may play a role in share-­‐rent contracts, and they find evidence supporting this hypothesis. Examining the available data on the farmland rental market and the tenant-­‐landlord relationship could inform us of the degree of market competition and the reasonableness of Young and Burke’s social norms hypothesis. Farmland Rental Market Frictions The empirical results demonstrate the failure of the standard economic model to

explain the farmland rental market. As with many other markets, the perfect-­‐information, frictionless-­‐markets assumptions appear untenable. Explaining the incidence results requires a closer look at the how the farmland rental market functions. One exceptional characteristic of the farmland rental market is the longevity of tenant-­‐landlord relationships. Allen and Lueck (2002) report an 11.5-­‐year average tenant-­‐landlord relationship duration in Nebraska and South Dakota. In Illinois, Sotomayer, Ellinger, and Barry (2000) report the mean tenant-­‐landlord relationship duration to be 14.4 years. According to the 2006 ARMS data, the average rental duration among soybean producers is 13 years, and the median duration is 10 years. Long-­‐lived contracts may be indicative of several rental-­‐market frictions. For

instance, heterogeneous land and farmer quality might result in a matching problem. Once tenants and landlords find a suitable match, the likelihood of separation will be low and decreasing with longer matches. Alternatively, transactions costs could explain long-­‐lived contracts, but the separation likelihood would increase with contract length as fixed transactions costs are spread over a longer period. We explore the role of rental arrangement duration in the incidence findings

introducing the duration of the landlord-­‐tenant relationship into the model and interacting

10

it with the subsidy measure. Table 6 contains the results of this analysis. Column 1 repeats the results from column 3 of table 3. The rental duration has a substantial direct effect— reducing the rental rate by 32 cents for every year of rental arrangement. Adding only the rental duration, however, has no effect on the incidence estimate. Interestingly, interacting the rental duration with the subsidy results in a substantial change to the direct incidence estimate, which nearly doubles to 0.44. The direct effect of the rental duration becomes insignificant, while the interaction term is marginally significant with a p-­‐value of 0.105. According to these estimates, the rental rate incidence falls by about 1.5 cents for every year of rental duration. At the median duration (10 years) the rental rate is 15 cents (roughly 33 percent) lower than in the first year of a the rental arrangement. The Tenant-­Landlord Relationship The tenant-­‐landlord relationship is an important, yet relatively unstudied aspect of the farmland rental market. Allen and Lueck (2002) report simple contractual arrangements, often “sealed with a handshake.” Typically, rental contracts are renegotiated annually (Allen and Lueck, 2002), but, as noted above, tenant-­‐landlord relationships appear to be quite long-­‐lived. Little more, however, is known of the representative tenant-­‐landlord relationship. Nationally representative data do not exist to answer fundamental questions such as the following: “How do tenants and landlords match?”; “Why are tenant-­‐landlord relationships so long-­‐lived?”; and “Do rental rates adjust annually, or only when there is a new tenant-­‐landlord match?” Understanding tenant and landlord characteristics can facilitate a better understanding of tenant-­‐landlord relationships. As reported in Kirwan (2009), the average tenant appears to farm about 30 percent more land than the typical subsidized-­‐crop

11

producer. Tenants are more profitable on average, earning 6 percent more per acre than the average farm, and they are 8 percentage points more likely to receive subsidies.

Generally, farmland rental contracts are between a farmer and a non-­‐farmer landlord; the 1999 AELOS, which surveys every landlord associated with a random sample of farm operators, gives a glimpse at the characteristics of non-­‐farmer landlords. As reported in table 2, the median non-­‐farmer landlord is retired (52 percent are), and nearly half (42 percent) of the retired landlords are retired farmers. The median age among retired landlords is 74. Figure 1 illustrates the importance of rental income as a share of total income by landlord type. As illustrated in figure 1 and reported in table 2, rental income comprises more than half of total income for 11 percent of non-­‐farmer landlords. Retired farmer landlords are particularly dependent on rental income; 29 percent of them derive over half of their income from renting out farmland.

Another important characteristic of non-­‐farmer landlords is their proximity to the rented land. A majority (51 percent) of non-­‐farmer landlords live within 5 miles of the land they rent out. Nearly 70 percent live within 25 miles, and only 13 percent live more than 150 miles away. Although landlords generally live near their farmland, not all landlords are rural residents. Forty-­‐eight percent of all landlords live in a non-­‐rural setting. Thirty-­‐ eight percent of all landlords live on a farm. These data provide some insight into the workings of the farmland rental market and the landlord-­‐tenant relationship. They indicate that landlords are close enough to their land and, since they are retired, have enough time to monitor the tenants’ use of the land. Landlords also might be subject to social norms as Young and Burke (2001) suggest. Since most landlords are local, they likely interact with their tenants in other settings. Because

12

rental income can be a large share of landlord income, it is important for landlords to find a tenant rather than leave their land idle. These characteristics suggest that tenants will have some bargaining power. That power derives from the landlord’s social and search costs associated with breaking a relationship with a current tenant and establishing a relationship with a new one. While such costs are likely bi-­‐lateral, it is not difficult to imagine how these departures from perfect competition could allow tenants to extract a share of subsidy benefits. Conclusion Economists have long suggested that agricultural subsidies become fully capitalized

into farmland values, and that subsidies only benefit farmers inasmuch as they are landowners. This rationale has lead to the argument that current landowners bought the “right” to the stream of subsidy payments when they paid for their land at subsidy-­‐inflated prices. In other words, by this line of thought, subsidies are entitlements. This paper refutes that notion by demonstrating that the landowner captures only 14 – 24 cents of the marginal subsidy dollar. Since subsidies have such a minor effect on farmland prices, the entitlement argument for continued agricultural subsidies falls. This paper improves on previous analysis by using data at the appropriate level of

aggregation and explicitly controlling for each field’s fundamental productivity, thereby overcoming omitted variable bias. Subsidies are a positive function of the subsidized land’s underlying productivity, hence, failure to account for the land’s fundamental productivity results in an upward biased incidence estimate. We explicitly control for the farmland’s underlying productivity by using farmers’ self-­‐reported expected productivity of the field. Using field-­‐level data, which is commensurate with the unit of analysis in standard

13

incidence theory, we find that farmland rental rates for subsidized soybean fields increase by 24 cents with the marginal subsidy dollar, and subsidized rice field rental rates increase by only 14 cents. To explain the low incidence estimate we look to the limited data on the farmland

rental market. Available evidence on tenant-­‐landlord relationships indicates that landlords have a relatively inelastic demand for tenants. Coupled with the increasing size of tenant farms, tenants appear to have substantial bargaining power, enabling them to extract most of the subsidy rents. In spite of the volume of research on the relationship between farmland values and agricultural subsidies, investigators, hampered by conventional wisdom and poor data, have not adequately addressed one of the most fundamental questions of agricultural policy analysis: “how much of the marginal subsidy dollar accrues to the landowner?” This paper answers the question using unique data that overcome several endogeneity concerns. We also highlight the need for more data on the farmland rental market to adequately explain the severe departure from economic theory.

14

References Allen, Douglas W. and Dean Lueck. 2002. The nature of the farm: contracts, risk and organization in Agriculture, Cambridge, Massachusetts: Massachusetts Institute of Technology Press. Goodwin, Barry K., Ashok K. Mishra, and François Ortalo-­‐Magné. 2005. “Landowners' Riches: The Distribution of Agricultural Subsidies.” Working paper, February 2005. Kirwan, Barrett E. 2009. “The Incidence of U.S. Agricultural Subsidies on Farmland Rental Rates." Journal of Political Economy. 117(1). Just, Richard E., and John A. Miranowski. 1993. "Understanding farmland price changes." American Journal of Agricultural Economics 75, (1) (Feb.): 156-­‐68. Melichar, Emanuel. 1979. "Capital gains versus current income in the farming sector." American Journal of Agricultural Economics 61, (5, Proceedings Issue) (Dec.): 1085-­‐92. Orden, D., R. Paarlberg, and T. Roe. 1999. Policy Reform in American Agriculture. Chicago, IL: The University of Chicago Press. Nourse, Edwin, Joseph Davis, and John Black. 1937. Three Years of the Agricultural Adjustment Administration. Washington, DC: Brookings Institution Press.

Roberts, Michael J., Barrett Kirwan, and Jeffrey Hopkins. 2003. The incidence of government program payments on agricultural land rents: The challenges of identification. American Journal of Agricultural Economics 85, (3): 762-­‐9. Robison, L. J., D. A. Lins, and R. Venkataraman. 1985. Cash rents and land values in united-­‐ states agriculture. American Journal of Agricultural Economics 67, (4): 794-­‐805. Sotomayer, N.L., P.N. Ellinger, and P.J. Barry. 2000. “Choice among Leasing Contracts in Farm Real Estate.” Agricultural Finance Review 60: 71-­‐84. U.S. Department of Agriculture. 1996. Agricultural Resource and Management Survey (ARMS) ERS and NASS. Washington DC. U.S. Department of Agriculture. 1999. Agricultural Economics and Land Ownership Survey (AELOS) NASS. Washington DC.

15

Young, H.P. and M.A. Burke. 2001. “Competition and Custom in Economic Contracts: A Case Study of Illinois Agriculture.” The American Economic Review 91 (June): 559-­‐ 573.

16

17

18

19

20

21

22

23

24