An Emprcal Investgaton of the Lnkages between Government Payments and Farmland Leasng Arrangements Feng Qu Barry K. Goodwn Jean-Phlppe Gervas Department of Agrcultural and Resource Economcs North Carolna State Unversty Last updated, March 010 Abstract: Ths artcle nvestgates the mpacts of decoupled and coupled support payments on farmland rental contract choces usng a prncpal-agent model. We consder cash and share contracts as well as hybrd contracts, whch are ncreasngly promnent n US agrculture. The conceptual framework suggests that restrctons on sharng payments between contractng partes are neffectve and nduce an offsettng contractual rearrangement. Emprcal results from a multnomal logt model confrm that government support programs have large sgnfcant effects on contract choces and that these effects vary by types of subsdes. Keywords: contract choce, hybrd contract, farm bll, program payments J.E.L. Classfcaton: Q1, Q15, Q18
An Emprcal Investgaton of the Lnkages between Government Payments and Leasng Arrangements About 45% of Unted States (US) farmland was operated by a tenant n 1999 (USDA/NASS 001). Hstorcally, contractual arrangements between landlords and tenants mostly ncluded ether cash payments or sharecroppng. More recently, a thrd form of leasng arrangement nvolvng both forms of payments - an arrangement that we desgnate as a hybrd contract - has ganed popularty. USDA/NASS (001) defnes a hybrd contract (also called a cash/share contract) as one under whch the tenant pays part of the rent n cash and part as a share of crops or lvestock products. 1 The use of hybrd contracts s ncreasng n the US farmland leasng market. In 1999, about 11% of all US leased farmland was under hybrd contracts, compared to only 3% n 1988. The ncdence of use of hybrd contracts was hghest n the Corn Belt and the Northern Plans (USDA/NASS 001). These two regons are manly comprsed of crop farms, whch are also the prmary benefcares of often generous government support programs. In 1999, 6% of leased farmland n Indana was rented under hybrd contracts, as compared to less than % n 1988. Smlar stuatons can be observed n Illnos, Oho, Kansas, Nebraska, Mssour, and Iowa. The lterature on farmland contract choce s consderable. Marshall (1890) lad out the early foundatons of the analyss of sharecroppng and llustrated the source of neffcency assocated wth sharecroppng (n relaton to a cash or wage contract). Sharecroppng dscourages the tenant s own nput use because he/she receves only part of hs/her margnal product. A number of studes challenged Marshall s concluson. Cheung (1969) argued that sharecroppng could be as effcent as other types of contracts f montorng s costless. Stgltz (1974), and Newberry and Stgltz (1979) ntroduced land tenure choces nto a prncpal-agent 1
framework. The standard agency model suggests that contracts are desgned to acheve a balance between effcent rsk-sharng and approprate ncentves to dscourage moral hazard. Allen and Lueck (199, 00) argued that n developed countres, where nsurance markets are well developed, rsk-sharng should not be the prmary determnant of contract choces. They argued that the beneft of a sharecroppng contract s that t curbs the tenant s ncentve to overuse the nputs (e.g., sol mosture and nutrents) suppled by the landlord. However, sharecroppng requres the output to be dvded between the landlord and the tenant and thus generates addtonal transactons and montorng costs for the landlord. More recently, Huffman and Just (004) ntroduced a prncpal-agent model whch allows for heterogenety n the characterstcs of prncpals and agents and relaxes the rsk-neutralty assumpton for landlords. They argued that the parameters of sharecroppng vary across tenants and landlords because of the former s heterogenety (e.g., the agent-specfc effort productvty). Huffman and Fukunaga (008), and Fukunaga and Huffman (009) provded recent emprcal evdence on the determnants of contract arrangements usng a model n whch agents choose between a share and cash-rent contract. They found that both rsk sharng and transacton cost ncentves are mportant determnants of the contract type. They also emphaszed the role of the landlords attrbutes nto the optmal landlord-tenant contract choce. The lterature has neglected two man ssues related to farm leasng arrangements. Frst, prevous studes have for the most part gnore the exstence of hybrd contracts, focusng on a bnary decson rule that entals cash rentals versus sharecroppng (e.g., Allen and Lueck 00; Fukunaga and Huffman 009). As argued above, hybrd contracts capture a growng share of leasng arrangements. Second, most studes gnored the mpacts of government support on contract choces (Berlen et al. 000, s a notable excepton). Government support programs are especally mportant n US agrculture. For example, the 00 Farm Securty and Rural
Investment Act was scored to provde more than $190 bllon of fnancal support to agrculture. In the 008 Farm Bll, taxpayer outlays were estmated to be nearly $300 bllon. Prevous studes demonstrated that the optmal contract choce depends on producton rsk, the tenant s and the landlord s rsk preferences, and the expected returns from rented land. Income support programs and subsdes wll affect the landlord-tenant contract choce because they potentally have mpacts on expected returns and ncome varablty as well as on the ndvduals degree of rsk averson. The contrbuton of ths study s twofold. Frst, we examne the effects of government programs on farmland rental contract choces. In an emprcal model, we break down aggregate government support nto fve dfferent programs and nvestgate to what extent each mpacts the probabltes of selectng a gven contract type. Second, we ntroduce hybrd contracts as a thrd alternatve n the contract set of landlords and tenants n order to nvestgate the determnants of an ncreasngly popular form of rental arrangement n US agrculture. The ndvdual-contract level data collected n the 1999 Agrcultural and Economcs Landlord Owner Survey (AELOS) and the 1999 Agrcultural Resource Management Survey (ARMS) are combned to carry out the analyss. The remander of the paper s structured as follows. The next secton ntroduces a smple prncpal-agent model to explan the landlord-tenant leasng arrangements. Secton three presents the dataset and the emprcal modelng strategy. Secton four dscusses the results of the estmaton procedure. Concludng remarks are offered n the last secton. CONCEPTUAL FRAMEWORK The model below bulds upon earler efforts of Huffman and Just (004), and Huffman and Fukunaga (008). We expand ther work by ntroducng agrcultural program payments nto the 3
model. We begn by addressng the role of decoupled payments n leasng arrangements. Decoupled payments were ntroduced n the 1996 Farm Bll, and were renewed n the 00 and 008 Farm Blls. They refer to lump-sum ncome transfers that are ndependent of current producton actvtes and market performance. For example, the Producton Flexblty Contract (PFC) payments under the 1996 Farm Bll are completely ndependent of current farm output levels. The only requrement for a producer (or a landlord) to receve PFC payments s that he or she has establshed base acreage. There are restrctons on the dstrbuton of decoupled payments. Legslaton requres that the payments be shared among producers and landlords subject to the contract on a far and equtable bass. Under a cash rental arrangement, 100 percent of the decoupled payments are allocated to the farm operator. Under a share contract, the government dstrbutes payments to both the landlord and the tenant operator accordng to the shares terms of lease. For smplcty, we assume that each landlord contracts wth only one tenant. The prncpal s the landlord and the agent s the tenant operator. Followng Huffman and Just (004), we allow heterogenety n rsk preferences of agents and prncpals. We also allow heterogenety n the productvty of effort, cost of effort, and reservaton utlty. A Prncpal-Agent Model wth Decoupled Payments The output of tenant operator on one unt of leased land (or net revenue wth approprate normalzaton) s defned as: (1) ( ) y e = ae + ε, where e s tenant s effort/labor nput and a s the tenant-specfc productvty of labor. Dfferences n productvty may be related to human captal n the form of farmng experence 4
(Huffman and Just 004). Output s also functon of a stochastc term ε whch s assumed to have zero mean and varance σ. Followng Huffman and Fukunaga (008), we assume the landlord offers a lnear ncentve contract to the tenant operator. The tenant operator s compensaton s: I e = α + β ae + ε + g 0.5ke, () ( ) ( ) d where k s the tenant-specfc effort cost parameter. A hgh (low) value of k ndcates a steep (flat) margnal cost curve. The varable g d represents decoupled government payments. The parameter α s the tenant-specfc cash payment of the contract. A postve α represents the cash wage pad by the landlord to the tenant; a negatve value for α means that cash rent payments are made to the landlord. The parameter β ( 0 β 1) s an ncentve rate representng a share of output. Hence, when β = 1 and α < 0, the leasng arrangement s a cash contract as opposed to 0 < β < 1 and α = 0 whch ndcate a share contract. More mportantly n the context of ths paper, 0 < β < 1 and α < 0 ndcate that the leasng arrangement s a hybrd type contract. Assume that the tenant has well-defned preferences over ncome summarzed by the utlty functon U ( I ). Expected ncome of the th tenant operator s ( ) 0.5 EI = α + β a e + g k e. The varance of the tenant s ncome s d VI = β σ. Let RP 0.5rVI denote the rsk premum where r U / U s the tenant s Arrow-Pratt coeffcent of absolute rsk averson. Under the expected utlty model, EU ( I ) U ( EI RP ) U ( I ) s an ncreasng functon of ncome, maxmzng ( ) =. Gven that EU I s equvalent to maxmzng 5
the expresson ( EI RP) (Chavas, 004). Therefore, the tenant operator s optmal effort s determned by maxmzng hs/her certanty equvalent CE = EI RP: (3) max CE = max [ EI 0.5rVI ] e e ( ae g ) ke r = max α + β + d 0.5 0.5 βσ. e The optmzaton problem defned n (3) solves the optmal effort e = β a k. * l Smlarly, the l th landlord s expected return from ownershp of the rented land equals = [(1 )( + ) ] and ts varance s ( ) Vπ β σ Eπ E β y g α d l =. As n the case of the tenant 1 operator, we wrte the landlord optmzaton problem n terms of the certanty equvalent return: (4) max CE = max [ Eπ 0.5rVπ ] β l β l l l = ( d ) max (1 β ) ae + g α 0.5 r (1 β ) σ β * l, subject to the partcpaton and ncentve compatblty constrants: (5) ( ) α + β ae + g 0.5ke 0.5rβ σ μ, * * d (6) e = arg max α + β ( ae + g ) 0.5ke 0.5rβ σ, * d e where r l s the Arrow-Pratt measure of the landlord s absolute rsk averson, and μ s tenant s reservaton utlty. The landlord s optmal choce of α wll be determned by the bndng partcpaton constrant. Substtutng e * and α nto (4) and optmzng over β yelds the optmal ncentve rate offered to the th tenant operator: (7) * c + rσ rσ β = = c r r c r r l 1 + ( + l) σ + ( + l) σ, 6
where c a k s an ndex of tenant-specfc effort productvty. Substtute the optmal share rate nto the partcpaton condton to obtan the optmal cash component of the contract: (8) 0.5 ( ) α = μ β c rσ β g. * * * d The optmal share rate β * n (7) emphaszes the role of the landlord s and the tenant operator s degree of rsk averson. If a tenant operator s rsk neutral ( r = 0), the optmal share rate equals one and a cash contract s the optmal outcome. Smlarly, the optmal share ncreases towards one as the landlord s coeffcent of rsk averson goes to nfnty (.e. r l ). Rsk, represented by the varance of the ncome, s negatvely correlated wth the optmal share rate. The hgher s the varance of ncome, the smaller s the optmal share rate. Therefore, an ncrease n ncome volatlty can have a negatve mpact on the choce of a cash contract, ceters parbus. A quck look at (7) suggests that decoupled payments do not have a drect mpact on the share rate. However, these payments may affect the contract choce ndrectly, through the mpact on the degree of rsk averson. If an ndvdual has constant absolute rsk averson (CARA) preferences, decoupled payments wll not have an mpact on the soluton n (7). However, f ndvduals rsk preferences ental decreasng absolute rsk averson (DARA), decoupled payments reduce the degree of rsk averson through the mpact on wealth. Decoupled payments however do have a drect effect on the optmal cash component of the contract. It reflects a pass-through of program benefts from the tenant operator to the landlord. From (8), the optmal cash component wth no decoupled payments (.e., g d = 0 ) s ( * ) ( c r ) μ 0.5 β σ, whch s greater than that wth postve decoupled payments. The * α dfference s β g, whch equals the share of decoupled payments gong to the tenant operator * d under the current legslatve envronment for US farm programs. Hence, the landlord captures 7
the benefts that go to the tenant by chargng an extra cash amount of sze β g, and restores the * d equlbrum that would have been attaned under no governmental restrcton, ceters parbus. 3 Under the optmal leasng arrangement, the landlord s able to capture all of the benefts dstrbuted to the tenant operator, gven the condtons that payments are decoupled and the wealth effect s neglgble. A governmental restrcton on payment dstrbuton does not nfluence the actual beneft dstrbuton between landlords and tenant operators n the end. It merely results n offsettng contractual rearrangements. 4 Ths s consstent wth Lence and Mshra (003), and Goodwn, Mshra, and Ortalo-Magne (009) who have found evdence that landlord capture 6%-86% of the entre benefts from decoupled payments by rasng cash rents. When referrng to the equlbrum leasng arrangement, t s nterestng to look at the comparson between the equlbrum contract choce wth decoupled payments and the choce wthout the payments under three general crcumstances. Assume both partes have CARA * * preferences and defne β and 0 α 0 as the optmal share rate and cash component of the contract under no decoupled payments. Frst, f the optmal contract s a share contract (.e., 0 < β < 1, and α = 0 ), the ntroducton of decoupled payments wll change the equlbrum to * * 0 0 a hybrd contract, ncreasng the cash payment to the landlord and keepng unchanged the share * * * rate ( 0 < β < 1 s constant, and α = 0 β g < 0 ). In a second case, f the equlbrum contract d wth no decoupled payments s a cash contract ( β = 1, and α < 0 ), the ntroducton of * * 0 0 decoupled payments would leave the share rate constant and the cash payment to the landlord * * * * would ncrease ( β = 1, and α = α β g < 0 ). Therefore, the equlbrum contract choce 0 d wll stll be a cash contract; however the cash rent would ncrease. Fnally, f the optmal leasng arrangement s a hybrd contract wthout decoupled payments, decoupled payments would not change the equlbrum contract type. The cash payments to the landlord would smply ncrease. 8
In summary, under the CARA assumpton, the ntroducton of decoupled payments ncreases the use of hybrd contracts and decreases the use of share contracts. Decoupled payments have no effect on the choce of cash contracts. On the other hand, f both partes have DARA preferences, decoupled payments wll lower the degree of rsk averson. From the solutons n (7) and (8), a cash contract wll emerge n the case where the landlord s degree of rsk averson goes to nfnty. Under more general condtons, the drect and ndrect effects of decoupled payments on the cash component of the contract ( α ) can go n dfferent drectons. The net effect depends on the rsk preferences of both contractng partes. Table 1 summarzes the effects of decoupled payments on optmal leasng arrangements by rsk preferences. The ambguous causal effect of decoupled payments on leasng arrangement can only be resolved emprcally. However, we examne the effects of coupled payments on contract choces before consderng the emprcal nvestgaton. A Prncpal-Agent Model wth Coupled Payments Coupled payments are based on current producton and/or market prce. Many forms of coupled programs exst n the US. These nclude prce and/or yeld support mechansms and dsaster relef programs. For smplcty, we nvestgate a per-unt producton subsdy n ths secton. As before, the landlord and the tenant share the subsdes n the same proporton as they share output. The per-unt producton subsdy rate s φ > 0 and coupled support equals gc = φ y. The landlord and the tenant operator s payments are β φ y and (1 β ) φ y, respectvely. Maxmzng the objectve functon defned n (3) accountng for coupled support yelds the optmal effort level e = (1 + φβ ) a k. The optmal share rate and the cash payment that maxmze the landlord s * objectve functon are: 9
(9) * (1 + φ) c + r(1 + φ) σ rσ β = = (1 + φ) + ( + )(1 + φ) σ + ( + ) σ l 1 c r rl c r rl, (10) * ( ) * α 0.5 1 ( μ φ β c rσ ) = +. The per-unt producton subsdy has a drect mpact on ncome varablty and the margnal productvty of effort. These two effects however cancel each other when determnng the optmal share rate, whch remans constant f no wealth effects are present. Turnng our attenton to potental wealth effects, coupled payments have an ambguous mpact on the share rate. The effect depends on both the landlord and tenant s rsk averson. As dscussed n the prevous secton, the optmal share ncreases towards one as the landlord s rsk averson ncreases to nfnty. However, coupled support may decrease the landlord s rsk averson coeffcent, whch could ental a swtch from a cash rental type to hybrd or sharecroppng. Producton subsdy payments have a drect mpact on optmal cash payments. If β = the cash payments decrease from μ 0.5( * ) ( β c rσ ) * * 0 β, to ( ) ( ) 0.5 1 * + c r μ φ β σ, whch suggest that cash payments ncrease as per-unt producton subsdes ncrease. Note that a decrease n cash payments s possble f the wealth effect decreases the landlord s coeffcent of absolute rsk averson. The mpacts of coupled payments on contract choce can dffer substantally accordng to the support types. Coupled payments nfluence the optmal share rate and cash payments through one or more of the followng factors: ncreases n expected returns, changes n ncome varablty, changes n the (value of) margnal productvty of effort, and mpacts on the contractng partes degree of rsk averson. Programs that decrease ncome varablty and/or ncrease a tenant s effort ncrease the optmal share rate, thus have a postve effect on the choce of a cash contract, unlke n the current case producton subsdes are postvely correlated wth rsk. Whle the latter 10
types of programs have not been explctly modeled, we use the nsghts of ths secton to state the hypotheses related to the causal relatonshp between US coupled farm payments and leasng arrangements. DATA AND EMPIRICAL MODEL The data used n ths study come from fve sources: the 1999 Agrcultural Economcs and Land Ownershp Survey (AELOS), the 1999 Agrcultural Resource Management Survey (ARMS), the Regonal Economc Informaton Systems (REIS) dataset for the 1990-1999 perod, unpublshed county level government program payments data from the Unted States Department of Agrculture (USDA) over the 1996-1999 perod, and county level farmland data from the 1997 Census of Agrculture. In contrast to other studes that only use one source of data (e.g., Huffman and Fukunaga 008, 009), we combne the above datasets to ncrease the explanatory power of the emprcal model. The AELOS s an ntegrated survey of farm fnance and land ownershp. It ncludes comprehensve nformaton collected from both tenants and landlords. Each observaton n ths dataset represents a unque contractual relatonshp between a landlord and a tenant operator. The ARMS s a natonal survey that provdes observatons of farm-level producton practces, economc attrbutes, and operator households characterstcs. We use ths dataset to obtan nformaton about rented farms, as well as addtonal operator characterstcs that may mpact the leasng arrangements. The REIS contans economc data and annual estmates of personal ncome for the resdents of the entre naton as well as states, metropoltan areas, and countes. We obtaned the county level farm ncome (cash recepts from marketng) data from REIS and county level farmland acres from the 1997 Agrcultural Census. 11
We refne the combned dataset followng these steps. Frst, we focus on the landlords who have only one renter. Ths accounts for 9% of the entre dataset. Second, some outlers (less than % of the avalable sample) are excluded from the analyss because they represent atypcal stuatons (for example, landlords reportng land rent exceedng $,000 per acre). Thrd, because crop farm producers are the man recpents of farm subsdes, farms that reported lvestock product sales that exceeded of 50% of ther farm sales are excluded. Farms for whch more than 50% of total sales were nursery products, fruts, or vegetables are also dropped from the sample. After ths selecton procedure, a total of 16,117 observatons reman for the analyss. In the AELOS dataset, each landlord/operator observaton has a dfferent weght to represent ther weght n the underlyng populaton, as f a complete census had been carred out. 5 We present the weghted results n ths artcle. We address the choce of leasng arrangements usng a multnomal Logt (MNL) model, appealng to the concept of random utlty derved by ndvdual n from a set of j = 1, K, J dfferent alternatves (Tran, 003): (11) nj nj ε nj U = V + j, where Vnj represents nformaton that s known by researchers and ε nj s the unobservable component of utlty. Let x be a vector of ndvdual-specfc characterstcs and β a correspondng vector of estmated coeffcents. If ε nj s unknown but follows a logstc dstrbuton, the choce probablty s (Long and Freese 006): (1) P nj α jb xβ jb = J exp( + ) j= 1 exp( α + xβ ) j b j b, 1
where b refers to the base alternatve whch s defned here as a share contract here. We normalze αbb = 0 and βbb = 0 so that the log of the odds of an alternatve compared wth tself s always zero. The log lkelhood functon for the MNL model s: (13) N ln L( β ) d ln P J =. n= 1 j= 1 nj nj The ndcator vector d nj equals 1 f ndvdual n chooses alternatve j, and equals zero otherwse. Model Specfcaton In the followng emprcal nvestgaton, we use a generalzed MNL model wth an alternatvespecfed constant (Tran 003). Each observaton n the dataset consttutes a landlord and tenant operator par (landlord-tenant hereafter) who s nvolved n a specfc farmland contract. The landlord-tenant chooses a contract among three alternatves: a cash contract, a share contract, or a hybrd contract. The decson s made condtonal on a set of ndependent varables whch are specfc to the landlord-tenant par n and are ncluded n the vector x n. Ths vector can be decomposed nto four dfferent parts that nclude farm program payments, farmng rsk and rsk preferences, tenant operator's effort productvty, and other factors, each of dmenson I G, I R, I P, and I M, respectvely. Alternatves are assumed to be mutually exclusve. The utlty functon can be wrtten as: IG IR IP IM. G R P M (14) V, = α + β GovP + β Rsk + β EffP + β Other nb b gb ng, rb nr, pb n, p mb nm, g= 1 r= 1 p= 1 m= 1 The subscrpt n (14) refers to ether the cash or the hybrd contract. The parameter α b s the th alternatve specfc constant whch can be nterpreted as the average effects of 13
unobserved factors. The varables GovP cg, are subsdes (per acre) receved from government program g. The varables Rsk nr, are proxes to capture farmng rsk and both partes rsk preferences. One potental proxy canddate for rsk s the coeffcent of varaton (CV) for farm ncome at the ndvdual farm level. However, ths may rase endogenety concern f the ndvdual CV s correlated wth unobserved farm characterstcs, such as land attrbutes. Therefore, we use the per-acre CV of cash recepts n the county where the ndvdual farm s located over a ten-year perod. A tenant operator s rsk preference s represented by household s wealth measured usng all assets. On the landlord s sde, we do not have access to ths varable. We follow Goodwn and Mshra (005) and use whether the landlord purchased nsurance for the target farm as a proxy to the landlord s rsk preference. The varables EffP np, represent the tenant operator s productvty. We employ farmng experence to capture the tenant operator s effort productvty. The varables n Other nm, nclude dfferent factors such as landlord s resdence and the farm type. Government program payments nclude sx components. They are the Producton Flexblty Contract (PFC) payments, the Market Loss Assstance (MLA) payments, the Loan Defcency Payments (LDP), Agrcultural Dsaster Payments (whch nclude all market loss or dsaster assstance payments, but exclude Federal Crop Insurance ndemnty and other ndemnty payments), the Conservaton Reserve Program (CRP), and a fnal category ncludng all other mnor program payments. As dscussed n the conceptual framework secton, decson makers use expectatons of future payments to determne the contract type. PFC payments are known wth certanty. However, dsaster, MLA, and LDP payments are not predetermned. They are trggered by market and producton condtons. Measurement ssues arse f actual reported payments are used to represent expectatons as noted n Goodwn, Mshra, and Ortalo-Magne 14
(003a). To control a potental errors-n-varables problem, we follow ther approach and use a four-year county average of payments to proxy expected program subsdes. 6 Future PFC payments are decoupled and known n advance of when a contract s sgned. Therefore, we use realzed farm-level payments to represent expected future payments. Table presents the defnton of key varables and summary statstcs. In the crop farm sample, 57% of farmland contracts were on a cash bass whle 18% were share contracts. The remanng 5% were hybrd contracts, makng the latter form of marketng arrangement more popular than pure share contracts for crop farms. From 1996 to 1999, farms receved $13.10 PFC payments per acre on average at the county level. The correspondng MLA, LDP, and Dsaster payments were $9.41, $9.64, and $1.77 per acre on average. Fnally, the county average CRP payments were $.44 per acre. All monetary values were adjusted by the consumer prce ndex to represent 004 dollars. Tenant operators had 5.6 years farmng experence on average. Fftyseven percent of landlords lved n a rural area and 39% of landlords are defned as absentee landowners and lved n an urban area. Prncple crop farms defned as olseed and gran farms - account for 68% of the crop farm sample. Expected Impacts of Key Factors on Contract Choces The PFC payments are decoupled subsdes whch are ndependent of current producton and market prce. The mpacts of decoupled payments on leasng arrangements are summarzed n Table 1. More specfcally, when the wealth effects are small or neglgble, the PFC payments wll entce agents to move from a share contract to a hybrd contract and thus wll redstrbute the benefts between contractng partes. The MLA, LDP, and dsaster assstance programs are coupled and are assocated wth current producton and/or market condtons. When wealth effects are neglgble, we can expect the followng mpacts of program payments 15
on the contract choce. The coupled programs (MLA, LDP, and dsaster) lower ncome varaton and have a postve effect on the optmal share lease rate. Thus, they rase the probablty of selectng a cash contract. However, f a wealth effect nfluences the degree of rsk averson of both partes, the effects of program payments can have opposte mpacts. In general, government programs can shft ncentves to use a partcular type of contractual arrangement and can redstrbute ncome and rsk between the landlord and tenant. The CRP s a specal type of program when consderng the mpacts of government payments on leasng arrangements. In most cases, payments are not related to the leased land. Tenant operators receve payments from ther own land. The CRP pays land owners annual rents to set ther land asde under a ten to ffteen year lease agreement. Land commtted to CRP must be removed from producton. Because the CPR payments usually do not nvolve rented land, they do not affect the landlord s ncentves. However, they may have an mpact on the contract choce by affectng the tenant s degree of rsk averson (through wealth effects). Accordng to the optmal share rate derved n (7), rsk s expected to have a negatve mpact on the optmal share rate. Thus, t reduces the probablty of choosng a cash contract. One concern at the emprcal stage s the possblty that a partcular type of prncpals contracts wth certan types of agents, a phenomenon dubbed endogenous matchng by Ackerberg (00). Ackerberg argues that f: 1) there exst ncentves for partcular partes to contract wth a specfc subset of the other partes (e.g., a rsk-averse tenant beng more lkely to contract wth a rsk neutral landlord); and ) some characterstcs (e.g., landlord s rsk preference) of contractng partes are not observable, explanng the outcome may nvolve a possble bas f the endogenety s not addressed. To nvestgate ths possblty, we carred out a two-stage regresson procedure that nvolves n the frst stage, regressng the tenant operator s rsk preference (represented by all assets owned by the farm household) on the landlord s rsk 16
preference (proxed by purchase of nsurance) and other exogenous factors may that may have an mpact on matchng (e.g., contractng partes age). We found no sgnfcant correlaton between the contractng partes rsk preferences. In a second stage, we use the predcted value of the tenant operator s rsk preference proxy and estmate the multnomal logt model. The results from the second stage are qute smlar to the uncorrected MNL estmaton results whch does not control for endogenous matchng. Intutvely, the smlarty between the results s consstent wth pror studes (e.g., Sherrck and Barry 003; Allen and Lueck 00) that emphasze how contracts emerge from long-run busness relatonshp due to close tes between the landlord and the tenant. Therefore, t s not unreasonable to treat the matchng of contractng partes as exogenous to the leasng arrangements n the US farmland market. In the US, a tenant operator normally contracts wth several dfferent landlords (on average, one tenant operator contracted wth four landlords n 1999). Some correlaton among observatons from the same tenant operator may exst. Therefore, clustered robust standard errors are used and based on the tenant operator s d number n ths artcle. The logt model mplctly mposes the Independence of Irrelevant Alternatves (IIA) assumpton whch states that the probablty of choosng among two alternatves s unaffected by the presence of addtonal alternatves. We test the IIA usng the Ch-Square test statstc proposed by Hausman and McFadden (1984). We are not able to reject the null hypothess that the IIA assumpton s vald at a hgh level of sgnfcance. Tests for combnng alternatves (Long and Freese 006) are also computed to examne f hybrd contracts are dstngushable from share and cash contracts. The Wald tests reject the hypothess that any two of the alternatve contracts are ndstngushable at a 0.01 level. 17
RESULTS Table 3 reports the estmates of the coeffcents n the three-alternatve MNL model whle Table 4 reports the margnal or dscrete changes n predcted probabltes for each alternatve derved from the estmates n Table 3. Government Program Payments Recall that program payments are measured n 004 dollars. Not surprsngly, the change n the predcted probablty followng a dollar ncrease s small. Therefore, we report the effects of a standard-devaton change n Table 4. We defne a standard devaton ncrease (centered on the mean) as one unt change when we refer to the margnal/dscrete effects. Table 4 shows evdence that the PFC payments have a postve mpact on the selecton of hybrd contracts and a negatve effect on share contracts. When a PFC payment ncreases by one unt, the probablty of choosng a hybrd contract ncreases by 1.1% and the probablty of choosng a share contract decreases by 3.7%. Ths s consstent wth the theoretcal explanaton that landlords are more lkely to capture the program benefts through a hybrd contract. The mpact of decoupled payments on choosng a cash contract s postve. A drect payment changes the wealth level and decreases rsk averson under DARA-type preferences. Both the dsaster payments and the loan defcency payments encourage the choce of a cash contract by reducng the ncome volatlty as s summarzed n table 1. If a tenant operator receves an addtonal unt of loan defcency payments, the probablty of choosng a cash contract ncreases by 0.6%. The predcted probablty of choosng a cash contract s 4.9% hgher followng a one unt ncrease n the tenant operator s dsaster payments. Meanwhle, both the LDP and the dsaster assstance payments decrease the probabltes of choosng a share contract. In contrast, the MLA payments have negatve mpacts on the cash contract choce. The margnal 18
effects of the MLA payments on both the cash and share contracts are the largest among all government programs. Gettng an addtonal unt of MLA payments decreases the probablty of choosng a cash contract by 6.5%. The extent of LDP and dsaster payments was determned by the 1996 Farm Bll. However, the MLA was determned outsde of the Farm Bll. In 1998, the prces of many crops declned sgnfcantly. Congress authorzed about $13.8 bllon as emergency MLA payments (trgged by low market prce, but based on hstorc base acreages) to help farmers deal wth ncome losses. Therefore, MLA actually targeted hgher rsk farms/crops. Ths would n turn make MLA correlated wth hgher rsk (hgh cv) and thus have negatve mpacts on the optmal share rate (.e., a decrease n the probablty of choosng a cash contract). The mpacts of CRP payments on the landlord-tenant contract choces are found to be nsgnfcant. The other payment category reveals a postve mpact on the choce of a cash contract and negatve mpact on a hybrd or a share contract. In concluson, the results ndcate that benefts from all fve man US farm support programs (wth the excepton of MLA) have postve effects on the choce of a cash contract. Conversely, the mpacts on the probablty of selectng a hybrd or a share contract dffer dependng on the specfcs of the program. Most programs have negatve mpacts on the probablty to observe a share contract. Decoupled payments encourage the use of hybrd contracts relatve to share contracts. The mpacts of program payments on contract choces show that rsk-sharng and beneft dstrbuton are mportant determnants of farmland leasng arrangements. Rsk and Rsk Preference The results n Table 4 provde evdence that rsk sharng s an mportant determnant of leasng arrangements. Rsk (as proxed by the CV varable) has a negatve effect on the choce of a cash 19
* contract ( β decreases as σ ncreases ). A standard devaton ncrease n the coeffcent of varaton wll reduce the probablty of choosng a cash contract by 3.6%, and ncrease the probablty of choosng a hybrd contract and a share contract by 1.6% and.0, respectvely. The tenant operator s wealth s found to be a sgnfcant varable n the MNL model. It ndcates that the wealther (and thus less rsk averse under DARA) the tenant s, the more lkely the resultng leasng arrangement s a cash contract ( β * ncreases as r decreases ). The probablty of choosng a cash contract ncreases by 8.0% when the tenant s wealth ncreases by one unt whle the probablty of choosng a share contract decreases by 10.%. The landlord s purchase of nsurance s found to be a sgnfcant determnant of contract choces. The results show that f a landlord purchases crop nsurance for the rented farm (denotng possble rsk averson), he/she s less lkely to choose a cash contract. Ths s not consstent wth the ntuton summarzed n table 1. We expected that a rsk-averse landlord would be more lkely to choose a cash contract. One possble explanaton s that the purchase of nsurance ndcates a more rsky farmng actvty ( σ s large) whch deters the use of cash contracts. Productvty of Effort The tenant operator s farmng experence s not found to be a sgnfcant determnant of the landlord-tenant s contract choce. Other Attrbutes Table 4 also reports that a landlord lvng n a rural area s more lkely to choose a cash contract than those who lve n an urban area. The evdence supports the transacton cost hypothess proposed by Allen and Lueck (00) whch states that an absentee landlord s more lkely to choose a share contract, under whch the tenants ncentve to overuse the land s smaller than 0
under a cash contract. It does not lend support to the transacton cost hypothess that an absentee landlord s less lkely to choose a share contract snce the cost of montorng s relatvely hgh (e.g., Cheung 1969). The results show that the farm type sgnfcantly affects contract choces as well. If the target crop farm belongs to a prncple crop farm type (.e., olseed and gran farms), the probablty of choosng a hybrd contract ncreases 9.8%. CONCLUSIONS Ths paper provdes a smple conceptual model to evaluate the mpacts of government programs on contract choces n agrculture. The result shows that exogenous legal restrctons on the dstrbuton of program benefts between contractng partes, such as the restrcton on the drect payments dstrbuton between landlords and sharecroppers under the 1996, 00, and 008 Farm Blls, can cause an offsettng contractual rearrangement n order to restore the beneft dstrbuton to the unrestrcted level. The ncreasngly common use of hybrd contracts (and decreasng use of share contracts) on crop farms s a form of ths contractual rearrangement. We use data from a varety of sources to emprcally analyze the determnants of contract choces usng a multnomal logt (MNL) model wth alternatve-specfed constants. The results confrm that dfferent polcy mechansms have dfferent effects on the farmland contract choces. More specfcally, we fnd that a one standard devaton unt ncrease n the PFC (decoupled) payments ncreases the probablty of usng a hybrd contract 1.1% and decreases the probablty of selectng a share contract by 3.7%. Other farm programs are also found to be sgnfcant determnants of leasng arrangements. Ther effects vary by the types of programs. Rsk-sharng ncentves are mportant determnant of contract choces. Ths study generates two mportant polcy mplcatons. Frst, t llustrates the potental bases that may arse when restrctng the set of potental leasng arrangements to only cash 1
contracts and sharecroppng. Introducng hybrd contracts nto the analyss s especally mportant to understandng the mpact of program payments on leasng arrangements. Second, the analyss suggests that governmental and legal restrctons on beneft sharng between contractng partes are neffectve and nduce offsettng contractual rearrangements. The ncreasng use of hybrd contracts lkely reflects a redstrbuton of program benefts between contractng partes. Most exstng emprcal research that analyzes the dstrbuton of program beneft between landlords and tenants effects focuses on the cash rental contracts (e.g., Lence and Mshra 003). Only a few studes examne the beneft dstrbuton under share contracts (e.g., Goodwn, Mshra, and Ortalo-Magne 009). Future studes may fnd t helpful to consder dfferent types of contracts, especally hybrd contracts.
Table 1. Effects of Decoupled Program Payments on Contract Choces Program payments No program payments Rsk preference Effect Optmal share rate * β rσ 1 * 1 μ ( ) c + r + β c rσ r ( l) σ rσ 1 * * 1 μ β ( c rσ ) β g c + r + r ( l) σ * Optmal cash-part payments α Effects on contract choce d Cash β =1 * * α <0 Hybrd β <1 0< * * α <0 Share β <1 0< * * α =0 Both CARA drect NO NO + ndrect NO NO NO NO NO Decoupled payments Both DARA TO CARA LL DARA TO DARA LL CARA drect NO NO + ndrect +/ +/ +/ +/ +/ drect NO NO + ndrect + +/ +/ drect NO NO + ndrect + +/ + Note. TO refers to the tenant operator and LL represents the landlord. 3
Table. Summary Statstcs (N=16,117) Varable Contract choce Frequency Percentage Cash contract 909 57.06 Share contract 395 18.13 Hybrd contract 889 4.80 Varable Defnton Mean Std. Dev. 1996-1999 County average program payments ($/acre) PFC Producton flexblty contract payments 13.10 8.70 MLA Market loss assstance payments 9.41 6.16 LDP Loan defcency payments 9.64 7.14 Dsaster Dsaster payments 1.77.31 CRP Conservaton reserve program payments.44.63 Other Other payments 0.15 0.90 Rsks and rsk preferences CV 10-year County average coeffcent of varaton of cash recepts (per acre) from market 0.13 0.7 Allassets_t Value of the tenant operator s all assets 1.56E6 4.14E5 Insurance_l 1 f landlord s purchase nsurance for the target farm 0.3 0.47 Tenant operator's effort productvty FarmngExp Tenant operator s farmng experence 5.59 1.89 Other factors Rural_l 1 f landlord lves n a rural area 0.57 0.49 Urban_l 1 f landlord lves n an urban area 0.39 0.48 Ft_man 1 f the farm type s a prncple crop farm (olseed and gran farmng) 0.68 0.47 4
Table 3. Maxmum Lkelhood Estmaton of MNL Models of Contract Choce Explanatory Varable Choce Coeff 1996-1999 County average program payments ($/acre) Robust Std Error PFC Cash 0.01* 0.003 PFC Hybrd 0.01* 0.003 MLA Cash -0.067* 0.014 MLA Hybrd -0.044* 0.015 LDP Cash 0.06* 0.013 LDP Hybrd 0.041* 0.013 Dsaster Cash 0.075* 0.033 Dsaster Hybrd 0.008 0.038 CRP Cash -0.008 0.09 CRP Hybrd -0.00 0.06 Other Cash 0.196* 0.073 Other Hybrd 0.107 0.008 Rsks and rsk preferences CV Cash -.839* 0.96 CV Hybrd 0.694* 0.94 Allassets_t Cash 1.88E-07* 5.74E-08 Allassets_t Hybrd 1.79E-07* 5.6E-08 Insurance_l Cash -0.913* 0.097 Insurance_l Hybrd -0.593* 0.091 Tenant operator's effort productvty FarmngExp Cash -7.11E-05.38E-04 FarmngExp Hybrd 4.80E-04 0.088 Other Factors May Affect the Contract Choce Rual_l Cash 0.35* 0.087 Rual_l Hybrd 0.037* 0.087 Ft_man Cash -0.9 0.0 Ft_man Hybrd 0.49* 0.00 Constant Cash 1.59* 0.197 Constant Hybrd -0.396 0.38 N=957,660 (weght used) Log pseudo-lkelhood =-90,959.6 Note: The astersks (*) ndcate that a coeffcent s sgnfcantly dfferent from zero at 0.05 or smaller level. 5
Table 4. Margnal and Dscrete Changes on the Predcted Probabltes Varable Margnal and Dscrete Changes n Predcted Probabltes (100%) Recept of program payments n 1999 Cash Hybrd Share PFC -+sd/.61 1.1-3.7 MLA -+sd/ -6.45 0.84 5.61 LDP -+sd/ 0.63.34 -.98 Dsaster -+sd/ 4.94 -.47 -.47 CRP -+sd/ 0.17-0.57 0.40 Other -+sd/ 3.35-0.78 -.57 Rsks and rsk preferences CV -+sd/ -3.61 1.58.03 Allassets_t -+sd/ 8.0.13-10.15 Insurance_l 0 1-14.63 1.39 13.4 Tenant operator's effort productvty FarmngExp -+sd/ -1.16 1.8-0.66 Other factors Rual_l 0 1 7.6-3.61-3.65 Ft_man 0 1-10.7 9.75 0.97 Note: -+sd/: change n predcted probablty as x changes from ½ standard devaton below base to ½ standard devaton above. 6
REFERENCES Ackerberg, D. A., 00. Endogenous Matchng and Emprcal Determnants of Contract Form Journal of Poltcal Economy 110 (3): 564-591. Allen, D. W., and D. Lueck. 199. Contract Choce n Modern Agrculture: Cropshare vs. Cash Rent. Journal of Law and Economcs 35 ():397-46. Allen, D. W., and D. Lueck. 1993. Transacton Costs and the Desgn of Cropshare Contracts. Rand Journal of Economcs 4 (1):78-100. Allen, D.W., and D. Lueck. 00. The Nature of the Farm: Contract Rsk, and Organzaton n Agrculture. Cambrdge, Massachusetts: The MIT Press. Berlen R., L.D. Parsch, B.L. Dxon, and B.L. Ahrendsen. 000. The 1996 FAIR Act: Measurng the Impacts on Land Leasng. Revew of Agrcultural Economcs ():336-354. Braverman, A., and J.E. Stgltz. 198. Sharecroppng and the Interlnkng of Agraran Markets. Amercan Economc Revew 7:695-715. Chavas, J-P. 004. Rsk Analyss n Theory and Practce. London. Elsever Academc Press. Cheung, S.N. 1969. The Theory of Share Tenancy. Chcago: Unversty of Chcago Press. Dasgupta, S., T.O. Knght, and H. A. Love. 1999. Evoluton of Agrcultural Land Leasng Models: A Survey of the Lterature. Revew of Agrcultural Economcs 1(1):148-176. Fukunaga, K., and W.E. Huffman. 009. The Role of Rsk and Transacton Costs n Contract Desgn: Evdence from Farmland Lease Contracts n U.S. Agrculture. Amercan Journal of Agrcultural Economcs 91:31-49. Gardner, B.L. 003. U.S. Commodty Poltcs of Farmland Values. In C. Moss, and A. Schmtz, eds. Government Polcy and Farmland Markets The Mantenance of Farmer Wealth. Ames. IA: Iowa State Press, pp. 81-96. 7
Goodwn, B.K., A.K. Mshra, and F. Ortalo-Magne. 003. What's Wrong wth Our Models of Agrcultural Land Values? Amercan Journal of Agrcultural Economcs 85:744-75. Goodwn, B.K, A.K. Mshra, and F. Ortalo-Magne. 003. Explanng Regonal Dfference n the Captalzaton of Polcy Benefts nto Agrcultural Land Values. In C. Moss, and A. Schmtz, eds. Government Polcy and Farmland Markets The Mantenance of Farmer Wealth. Ames. IA: Iowa State Press, pp. 3-14. Goodwn B.K. and A.K. Mshra. 005. Another Look at Decouplng: Addtonal Evdence on the Producton Effects of Drect Payments. Amercan Journal of Agrcultural Economcs 87:100-110. Goodwn, B.K., A.K. Mshra, and F. Ortalo-Magne. 009. The Buck Stops Where? The Dstrbuton of Agrcultural Subsdes. Workng paper, Department of Agrcultural and Resource Economcs, North Carolna State Unversty. Hausman J., and D. McFadden. 1984. Specfcaton Tests for the Multnomal Logt Model. Econometrca 5:119 140. Huffman, W.E., and R.E. Just. 004. Implcatons of Agency Theory for Optmal Land Tenure Contracts. Economc Development and Cultural Change Workng paper 54(3):617-64. Huffman, W.E., and K. Fukunaga. 008. Sustanable Land Use: Landlord tenant Contractng n the Unted States of Amerca. NJAS Wagenngen journal of lfe scences 55(4):379-396. Lence, S.H., and A.K. Mshra. 003. The Impacts of Dfferent Farm Programs on Cash Rents. Amercan Journal of Agrcultural Economcs 85:753-761. Long, J.S., and J. Freese. 006. Regresson Models for Categorcal Dependent Varables Usng Stata. College Staton, TX: Stata Press. Marshall, A. 1890. Prncples of Economcs, 8th edn. Macmllan, London. 8
Newbery, D., and Stgltz, J.E. 1979. Sharecroppng, Rsk Sharng and the Importance of Imperfect Informaton In J. Roumasset, ed. Rsk, Uncertanty and Agrcultural Development. New York: Agrcultural Development Councl. Sherrck B.J. and P.J. Barry. 003. Farmland Markets: Hstorcal Perspectves and Contemporary Issues. In C. Moss, and A. Schmtz, eds. Government Polcy and Farmland Markets The Mantenance of Farmer Wealth. Ames. IA: Iowa State Press, pp. 7-49. Stgltz, J.E. 1974. Incentves and Rsk Sharng n Sharecroppng. The Revew of Economc Studes 41():19-55. Tran, K. E. 003. Dscrete Choce Methods wth Smulaton. Cambrdge: Cambrdge Unversty Press. USDA/NASS (U.S. Department of Agrculture, Natonal Agrculture Statstcs Servce). 001. Agrcultural Economcs and Land Ownershp Survey (AELOS 1999). Internet Webste: http://www.nass.usda.gov/census/census97/aelos/aelos.htm USDA/NRCS (U.S. Department of Agrculture, Natural Resources Conservaton Servce). 008. Envronmental Qualty Incentves Program. Internet Webste: http://www.nrcs.usda.gov/programs/eqip/ 9
Endnotes 1 In what follows, we use a relatvely narrower defnton of the hybrd contract as one that conssts of a predetermned share percentage plus a fxed cash payment. Tenant operators nclude pure-tenant operators, who rent all of the operatng farmland from others; and partowner operators, who own part of the operatng farmland and rent part of the land from others. 3 We assume that transacton costs for renegotatng contracts are zero. 4 Cheung (1969, chapter 5) reaches a smlar concluson. 5 For more nformaton about the calculaton of these weghts, see General Explanaton for Agrcultural Economcs and Land Ownershp Survey (1999), whch s avalable at: www.nass.usda.gov/census/census97/aelos/appendxa.pdf 6 Market Loss Assstance payments were ntroduced n 1998 and we use the 1998-1999 average annual payments. For other programs, we use 1996-1999 average annual payments. 30