The Effect of the Conservation Reserve Program on Land Values€¦ · Konyar 1990; Feather, Hellerstein, and Hansen 1999; Wu 2000; and Kirwan, Lubowski, and Roberts 2005). For exam-ple,

The Effect of the Conservation Reserve Program onLand Values

JunJie Wu and Haixia Lin

ABSTRACT. This paper evaluates the effects of theConservation Reserve Program (CRP) on landvalues. A theoretical model is presented to analyzethe interaction between farmers' CRP participationdecisions and land values. Empirical models areestimated to evaluate the effects of the CRP on landvalues. Results suggest that CRP participation hadthe largest effects in the Mountain, Southern Plains,and Northern Plains regions, where it increasedaverage farmland values by 5% to 14%, 4% to 6%,and 2% to 5%, respectively. The CRP also had astatistically significant effect on developed landvalues, but the percentage increases were smaller.Implications of the results for the design ofconservation programs are discussed. (JEL Q24,Q28)

I. INTRODUCTION

The Conservation Reserve Program(CRP), one of the largest conservationprograms in U.S. history, was estabhshedby the Food Security Act of 1985 and wasreauthorized in all subsequent farm bills.Under this program, farmers convert highlyerodible cropland or other environmentallysensitive acreage to vegetative cover such asnative grasses, trees, or filter strips; inreturn, they receive an annual rental pay-ment for a contract period of 10 to 15 years.By 2004, over 34 million acres of croplandhad been enrolled in the CRP, with anannual rental payment of approximately $2billion (USDA 2004). In some countiesmore than 20% of cropland has beenconverted to vegetative cover under theCRP.

The economic and environmental bene-fits of the CRP have been well documented(e.g.. Young and Osbom 1990; Osbom and

Land Economics • February 2010 • 86 (1): 1-21ISSN 0023-7639; E-ISSN 1543-8325© 2010 by the Board of Regents of theUniversity of Wisconsin System

Konyar 1990; Feather, Hellerstein, andHansen 1999; Wu 2000; and Kirwan,Lubowski, and Roberts 2005). For exam-ple, based on 1997 enrollments, the CRP iscredited with reducing soil erosion by 224million tons annually, generating a total of$500 million of on-site and off-site econom-ic benefit per year (Sullivan et al. 2004, 22-23). With about 8% of the nation's croplandenrolled into the CRP, this program mayalso have an impact on land values.However, only a few studies have examinedthe effects of the CRP on farmland values,and their results seem inconsistent. Shoe-maker (1989) analyzed the first five CRPsign-ups, from 1986 to 1987, and found thatCRP participation provided a huge windfallto farmers but had little effect on farmlandvalues. Goodwin, Mishra, and Ortalo-Magné (2003) evaluated the effect of theCRP and other farm programs on farmlandvalues; their results indicate that CRPpayments correlated with lower farmlandvalues. Lence and Mishra (2003) usedcounty-level data from 1996 to 2000 toexamine effects of the CRP and other farmpayment programs on cash rental rates inIowa; their results indicate that the effect ofthe CRP was positive or zero, depending onthe models used.

Many studies, however, have examinedthe effects of government commodity pro-grams on farmland values (Rosine andHelmberger 1974; Castle and Hoch 1982;Alston 1986; Goodwin and Ortalo-Magné1992; Clark, Klein, and Thompson 1993;Barnard et al. 1997; Ryan et al. 2001;Schmitz and Just 2002). The conventional

The authors are, respectively, Emery N. CastleProfessor, Department of Agricultural and ResourceEconomics, Oregon State University; and economist.Dell, Inc. We thank two anonymous referees for theiruseful comments and suggestions.

Land Economics February 2010

wisdom is that because the supply ofagricultural land is highly inelastic, govem-ment payments are largely capitalized intofarmland value. Recently, Kirwan (2008)presented a direct test of this theory usingfarm-level data and found that only 20 to 25cents of the marginal subsidy dollar isreflected in increased rental rates, whereastenant net returns rise by 70 to 75 cents.Roberts, Kirwan, and Hopkins (2003)found that each dollar of governmentpayments increases land rents between 34and 41 cents. Several studies have examinedthe proportion of land value that can beattributed to govemment subsidies. Forexample, Barnard et al. (1997) examinedthe ef'fect of eliminating the Federal Agri-cultural Improvement and Reform Act of1996 (FAIR.) on cropland values in theUnited States and found that croplandvalues would be reduced by 12% to 69% inthe eight examined regions if governmentprograms were eliminated. Just and Mir-anowski (1993) found that governmentpayments accounted for 15% to 25% offarmland values.

Although many previous studies haveevaluated the effects of govemment com-modity programs on farmland values,relatively few have focused on the effect ofthe CRP. Furthermore, no study, to ourknowledge, has examined the effect of theCRP or any other govemment commodityprograms on developed land values. Thelack of analysis is surprising given that (1)farmland is a main asset of the agriculturalsector, (2) the opportunity cost of farmlandrepresents a major production expense(Lence and Mishra 2003), (3) the CRP isthe largest land retirement program in U.S.history, and (4) how farmland values areaffected is a critical issue in farm policydebates (Goodwin, Mishra, and Ortalo-Magné 2003).

The primary objective of this study is toevaluate the effects of the CRP on prices offarmland and developed land. To achievethis objective, we first present a theoreticalmodel to guide our empirical investigation.The model integrates Capozza and Li's(1994) land price model with an optimal

bidding behavior model (Latacz-Lohmannand van der Hamsvoort 1997). Under theCRP, any farmers with highly erodible landor other environmentally sensitive acreagecould submit an application for CRPparticipation by indicating the parcels theywish to enroll in the CRP and the annualrental payments they require. However, thehigher the rental rate a farmer requires, theless likely it is his application will beaccepted. Thus, farmers' CRP rental rates,the probability of CRP participation, andthe impact of the CRP on land values aretreated as endogenous variables that aresimultaneously determined in our model. Inprevious studies, government payments aretypically treated as exogenous variables infarmland value equations. Based on thetheoretical analysis, empirical models arethen estimated to evaluate farmers' CRPparticipation decisions and the resultingeffects on values of farmland and developedland.

Results suggest that the CRP increasedthe average farmland value by between $18and $25 per acre in the United States in1997. The effect was largest in the Moun-tain, Southern Plains, and Northern Plainsregions, where it increased the averagefarmland values by between 5% and 14%,4% and 6%, and 2% and 5%, respectively.The CRP also had a statistically significanteffect on developed land values; however,the percentage increases were smaller,although the absolute increases were larger.Agricultural returns accounted for about40% of total farmland values in the UnitedStates, and growth premium and optionvalue accounted for the remaining 60%.Implications of the results for the design ofconservation programs are explored below.

II. THE THEORETICAL MODEL

Under the current CRP rules, any farm-ers with highly erodible land or otherenvironmentally sensitive acreage can applyfor CRP participation during a sign-upperiod by indicating the parcels they wish toenroll and the annual rental payments theyrequire. Whether an application will be

86(1) Wu and Lin: Conservation Policy and Land Values

accepted depends on the rental rate thefarmer requires and the level of potentialenvironmental benefits the parcel offers ifconverted to vegetative cover. Potentialenvironmental benefits may include wildlifehabitat, water quality, reduced soil erosion,improved air quality, and conservationpriority area (USDA 1997). Based on thesebenefits, an environmental score (S) iscalculated for each offered parcel via aformula established by the U.S. Depart-ment of Agriculture (USDA), which is thencombined with the farmer's bid rent (B) toobtain an environmental benefit index"(EBT): EBI=g(S,B), where dEBI/dS>0,dEBI/dB<0. Bids with an EBI above acutoff'level Xare accepted, and bids with anEBI below X are rejected.

Because a farmer does not know thecutoff level, X, in preparing his application,he faces a trade-off in choosing his bid rent;if the bid rent is too high, it will not beaccepted; if the bid is too low, he will losethe opportunity to receive a higher rentalrate. It is plausible to assume that eachfarmer forms expectations about X, whichcan be characterized by the density func-tion f(X) and the distribution functionF{X). The probability that a bid is acceptedequals

Pr = prob(X < EBI) = F(EBI). [1]

If the bid is accepted, the farmer's annualnet return will be B; if it is rejected, thefarmer's annual net return will be theagricultural rent A. If the farmer is riskneutral, he will choose B to maximizeexpected net payoff: BE(EBI) + A[l-F(EBI)]. The first-order condition of thismaximization problem implicitly definesthe optimal bid rent B* for the farmer:

B'=A-F(EBI)

f(EBI)gB{S,B) • [2]

The optimal bid consists of two compo-nents: the foregone agricultural rent and theinformation premium, which depends onthe farmer's private information about thecutoff level oí EBI as described hyf{X) andE{X). For example, farmers may form their

expectation of X based on the acceptedrental rates in past CRP sign-ups.

As an alternative to CRP participation, afarmer could convert his cropland todevelopment. The value of one unit offarmland at location z and time / equals thepresent value of the expected net returns tofarmland up to the date of conversion plusthe present value of the expected returns todeveloped land, minus the conversion cost:

where R'^'^'' is the expected net return to theparcel before it is converted to developmentai t+s\ R'^'^'' equals PrB*->r{\ -Pr)A if theparcel is eligible for the CRP, and A if not.Because both B* and Pr are explainedvariables, R'^'^'' is endogenous. R{T,Z) isthe expected net return to the parcel after itis converted to development, C is the one-time conversion cost, r is the discount rate,E{} is the expectation operator. FollowingPlantinga, Lubowski, and Stavins (2002),R{x,z) is specified as R{t,z) = R{i) + R{z),where the temporal component R{t) followsthe Brownian motion process with upwarddrift g and variances a^: R(t)=gt + (TB{t),and the spatial component R{z) is specifiedas a function of amenities at location z, a{z),and other locational characteristics such asthe distance to the nearest city center, d(z):R(z) = R[a(z),d(z)].

Assuming that the landowner chooses theconversion time (t + s) to maximize theexpected value of land, the optimal decisionrule can be derived following Cappozza andHelsley (1990): convert the parcel to devel-opment if

R(t,<--ag)

[4]

where a = [(g- + 2<7 /-)'/ —g]/cr . The valueof farmland can be derived as

Land Economics February 2010

r

-r

4r'-

[5]

where z* is the "boundary" of the developedarea and is defined by R{t,z) = R*. The valueof farmland has three components: thevalue of net returns from agriculture orthe CRP, growth premium, and optionvalue. Similarly, following Capozza andHelsley (1990), the value of developed landcan be derived as

j^CRPg , r-

R(z)-R{z')[6]

The value of developed land consists offive components, which have been referredto as the net return from agriculture or theCRP, conversion cost, growth premium,irreversibility premium, and amenity andaccessibility premium. The irreversibilitypremium represents the loss of option valueonce a parcel is developed. The amenity andaccessibility premium represents the valueof amenities and other locational advantag-es (e.g., proximity to work).

Although the CRP always increases thefirst component of equations [5] and [6], itreduces the growth premium and option valueof farmland because the CRP slows the paceof land development. With the CRP, less landis developed, and the growth premium andoption value for an undeveloped parcel arereduced. Similarly, the CRP reduces theaccessibility premium of developed landbecause it reduces the comparative advantageof a developed parcel over an undevelopedparcel located at the boundary, which is closerto the city center with the CR.P.

Values of farmland and developed landare also affected by the spatial componentof urban land rents R{z), which is a functionof amenities and other locational charac-teristics. Equations [5] and [6] suggest thatlocations with better amenities, easy access,and lower transportation costs have higherland values regardless of whether they are

agricultural or developed land. Amenitiesaffect farmland values because they affectgrowth premium and option value.

III. EMPIRICAL MODELS

Specification

Equations [1], [2], [5], and [6], whichdescribe the relationship between farmers'CRP participation decisions and land val-ues, provide a theoretical foundation forour empirical analysis. The empirical coun-terparts of these equations are specified asfollows:

r = F{EBI)=-1+ei

[7]

[8]

[9]

[10]

where Xi, X2, X3, and X4 are vectors ofvariables affecting the probability of bidacceptance, the optimal CRP bid rent,farmland value, and developed land value,respectively; ß's are vectors of parameters;and e's are error terms. Variables includedin X|, X2, X3, and X4 and the functionalforms of/ (/ = 2, 3, 4) are specified basedon equations [1], [2], [5], and [6].

As shown by equation [1], the variablesaffecting the probability of CRP acceptanceinclude the environmental score (5), the bidrent {B*), and a farmer's expectation aboutthe cutoff level of EBI (as described byF{)). The variables affecting the farmer'sexpectation about the cutoff level of EBImay include the average CRP rental rate inprevious CRP sign-ups (5_i) and thepercentage of cropland already enrolled inthe CRP in the county (CRP^i). Thepercentage of cropland already enrolledmatters because the USDA often facespolitical pressure to spread CRP dollarsacross geographical regions. For example.

86(1) Wu and Lin: Conservation Policy and Land Values

in CRP sign-up 26, the USDA kept theacceptance rates in Montana, North Dako-ta, South Dakota, and Texas artificially low(by setting a higher cutoff level of EBI) toensure that more CRP acres were allocatedto other states (USDA 2003). Thus, weassume Xi =(5,5*,5_i,C7?P_i). Becausenot all variables affecting the probabilityof CRP acceptance are known to theresearchers, an error term ei is added inequation [7]. In a discrete choice model, F{)is typically assumed to be a logistic ornormal distribution function, and thechoice between the two typically make littleempirical difference. For easy estimation,discussed below, the probability of bidacceptance [1] is specified as a logit model.

Equation [2] reveals the variables affect-ing the optimal level of the CRP bid rent.These variables include the net return tofarmland. A, the environmental score, S,and the variables affecting a farmer'sexpectation about the cutoff level of EBI,including 5_ i and CRP.- \. Thus, the CRPbid rent is specified as a function of

Equations [5] and [6] reveal the variablesaffecting values of farmland and developedland. These variables include the net returnfrom agriculture or CRP participation^j^CRP^^ the growth rate of developed landrents (g), the variance of developed landrents (a^), amenities (Ö(Z)), and otherlocational characteristics ld(z)). Thus, weinclude 7?"- , g, c^, a(z), and d{z) in X3 andX4. Previous studies of farmland values thatinclude proxy variables for future develop-ment rents, amenities, or locational charac-teristics include those by Hushak and Sadr(1979), Chicoine (1981), Shonkwiler andReynolds (1986), Palmquist and Danielson(1989), Vitaliano and Hill (1994), Shi,Phipps, and Colyer (1997), Hardie, Nara-yan, and Gardner (2001), and Cho, Wu,and Boggess (2003).

' Many previous studies have also examined thedeterminants of developed land prices or values, includingthose by Coulson and Engle (1987), Rosenthal andHelsley (1994), Colwell and Munneke (1997), Kowalskiand Paraskevopoulos (1990), and McDonald and McMil-len (1998).

Two alternative functional forms are usedin the estimation of the land value equations[5] and [6]. One simply specifies the values offarmland and developed land as a quadraticfunction of R^^, g, a^, a{z), and d{z).Although this specification allows us toestimate the impact of CRP participation onthe values of farmland and developed land,it cannot be used to estimate the impact ofCRP participation on individual compo-nents of land values. To do that, we mustimpose more structure on the functionalforms of/3 and/4. Specifically, we rewritethe farmland value equation [5] aspa^j^cRPi^j^^^^l^^y^[R(z)-R(z*)]^ where the

second term is the sum of the growthpremium and the option value. Note that(1/ra) is a function of g and aa, and^a{R(z)-R(z*)\ ig ^ function of T?^'^, g, G\a{z), and d(z) because z* depends on allthese variables. Approximating {\/ra) and^a[R(z)-R(z*)] by their first-order Taylorexpansion, the farmland value equationcan be specified as

z)

[11]

where the first term {\represents the values of net return fromagriculture or CRP participation, and therest of the terms represent the sum ofgrowth premium and option value. Notethat equation [11] does not include anintercept term because when R^'^^, g, and(7 are all zero, p" is also zero.^ With thisspecification, the impact of CRP participa-tion on farmland values can be estimated by

[12]

where the first term measures the effect ofthe CRP on agricultural returns, and thesecond term measures the effect of the CRP

0,By definition, a = [(g

= (T/(í-v/2r)-0 as (T-.0.'^-gya'^. When g =

Land Economics February 2010

on the growth premium and the optionvalue.

Similarly, the developed land value equa-tion [6] can be written asp'^ = ( '^^-l-(^3, where ^^=e-^^'^^'_g~a[R{.)-R(z*)]y^^ and

R{z*)]/r. Approximating (j)^, ^2, and ^-^ bythe first-order Taylor expansion, the devel-oped land value equation can be specified as

'R^'"' + ßlR'

[13]

The data and methods used in theestimation of the empirical models arediscussed below.

Data

The empirical models are estimated usingcross-section data from 2,851 counties in thecontiguous United States in 1997. Onehundred and ninety counties are omitteddue to missing data or absence of agriculturalland. Although individual CRP bid andcontract information is available, correspond-ing parcel-level data are unavailable forvalues of farmland and developed land. Wehave, however, average values of farmlandand developed land for the 2,851 counties.

CRP data are provided by the EconomicResearch Service (ERS) of the USDA.^ Thedata contain individual bid and contractinformation for sign-up 15, which was heldin March 1997. We could estimate the bidrent equation using the individual bid andcontract data and then estimate the CRPacceptance equation as a discrete choicemodel using predicted values of B. Howev-er, we would not be able to address theissues of spatial autocorrelations becausethe relative locations of the offered parcelsare unknown (locations and owners of CRP

' We thank Shawn Bucholtz of the EconomicResearch Service for providing the data.

bids are confidential). In addition, becauseboth the land value equations are estimatedusing county-level data, the parcel-levelpredictions of B and Pr would have to beaggregated to the county level to be used inthe estimation of land value equations. Forthese reasons, we chose the followingprocedure to estimate the acceptance andbid rent equations. First, using individualbid data, we estimate the probability ofacceptance in each county by calculatingthe ratio of the total accepted bids to totalbids submitted in sign-up Í5. With the data,we are able to convert the discrete choicemodel of CRP acceptance [7] to a contin-uous dependence variable model:

In Pn\-Pn

[14]

where Pr¡ is the percentage of total submit-ted bids accepted in county / estimatedusing the individual bid data, B* is theaverage bid rent per acre in county /, which

n n

is calculated hy (J2 bk * acre/c)/ Y, acre^,k=\ k=\

where bk is the per acre bid rent and acrek isacres offered by fanner k, and n is the totalbids submitted in a county. Using theindividual CRP bid data, the averageenvironmental score (5,) is computed foreach county. The average past CRP rentalrate (5_i,) and percentage of land alreadyenrolled in the CRP (CRP_u) are estimatedusing data from the ERS. The average pastCRP rental rates are calculated using rentalrates from all previous sign-ups (i.e., sign-ups 1-14). The percentage of land enrolledin the CRP is computed as the ratio of totalland enrolled in the CRP by December 1996to total cropland in a county. Percentage ofcropland eligible for the CRP in eachcounty is estimated using data from the1997 National Resource Inventory. Simi-larly, we estimated the following bid rentequation using the county-level data:

[15]

86(1) Wu and Lin: Conservation Policy and Land Values

Data on farmland values, developed landvalues, and annual net returns to farmlandin 1997 were obtained from Plantinga,Lubowski, and Stavins (2002). * They cal-culated the average return to farmland. A,using Census of Agriculture data by (77? +GP - TQ/TA, where TR is the totalrevenues from the agricultural productssold, GP is the total government pay-ments except CRP payments, TC is thetotal farm production expenses, and TA isthe total farmland acres. The farmlandvalue (p") is the county-level average of self-reported estimates by landowners. Thedeveloped land value (p'^) is a county-levelestimate of the average per acre value ofrecently developed land for single-familyhouses (Plantinga, Lubowski, and Stavins2002).

As in previous studies, we also divideland use into discrete classes: farmland anddeveloped land. However, land use canspan a spectrum of uses ranging fromexclusively intensive agriculture, to small-er-scale and hobby types of farm opera-tions, to something more like backyardgardening, to exclusively residential.^Farmland and developed land values de-pend not on these two discrete land uses,but rather on the capability of land tosupport a range of uses, the demands forland in those different ranges of uses, andthe densities of development that arefeasible and allowable under given circum-stances. For these reasons, p'' does notnecessarily represent true "developed landvalues" but rather serves only as a proxy forthe value of land in more intensive devel-oped uses. The percentage change in thevalue of land in less intensive uses (e.g.,mixed uses with farmland and developedland) under the CRP is likely between thoseestimated for farmland and developed landin this study.

Two alternative approaches are used tomeasure amenities in a county. One uses theamenity data generated by the NationalOutdoor Recreation Supply Information

"* We thank Andrew Plantinga for providing the data.' We thank an anonymous referee for pointing this

out.

System (NORSIS),'' developed and main-tained by the USDA's Forest ServiceWilderness Assessment Unit, Southern Re-search Station. The amenity data includemore than 250 variables describing climate,natural amenities, man-made amenities,and geographic information across countiesin the United States. To synthesize theinformation contained in the large numberof variables, we use principal componentanalysis to calculate amenity scores for eachcounty, following Délier et al. (2001).Principal component analysis is an ap-proach to compress higher-dimension var-iables into a single scalar, which is, inessence, a linear combination ofthe originalvariables with weights being the eigenvec-tors of the correlation matrix for the factorvariables. Because the principal componentis sensitive to scale, all variables used inprincipal component analysis are standard-ized to zero mean and unit variance, and theamenity score is calculated by Score =

J2 ^1^1, where A/ is the eigenvector com-puted from the variance-covariance matrixof the original data, x, is the standardizedamenity variable, and L is the number ofvariables in a category. The main advantageof this approach is that variables are notremoved from the empirical analysis due tomulticollinearity problems or limited degreeof freedom (Wagner and Délier 1998).

We constructed three amenity scores foreach county to measure amenities derivedfrom temperate climate (e.g., the number ofsunny days in January, low humidity inJuly), man-made recreation facilities (e.g.,the numbers of golf courses, swimmingpools, campgrounds), and natural recrea-tional resources (e.g., total outstandingriver miles, Whitewater miles). The amenityscores for climate and natural recreationalresources are constructed using 4 variables,and the amenity score for man-maderecreation facilities is constructed using 14variables.'

'' We thank Steve Délier of the University ofWisconsin for providing the NORSIS data.

' Variables in each category and their correspondingeigenvector are available upon request.

Land Economics February 2010

An alternative measure of natural ame-nities used in this analysis is the naturalamenity scale created by the ERS. The ERSnatural amenity scale was constructedbased on six factors: warm winter (averageJanuary temperature), winter sun (averageJanuary days of sun), temperate summer(low winter-summer temperature gap), sum-mer humidity (low average July humidity),topographic variation (topography scale),and water area (water area proportion oftotal county area).

In addition to amenities, other locationalcharacteristics that affect land values in-clude accessibility, transportation costs,and development pressure. Two alternativevariables are used to refiect accessibility anddevelopment pressure in a county. One isthe total mileage of interstate and otherprincipal arterial roads (for example, statehighways) in a county. The other is theUrban Influence Code (UIC) developed bythe ERS to capture an area's geographiccontext and economic opportunities basedon population and commuting data. The1993 UIC, the most recent available before1997, was obtained from the ERS. The 1997road mileage data were obtained from theU.S. Bureau of Transportation Statistics.

Based on work by Capozza and Helsley(1990), we used the growth and variance ofreal income to approximate g and G'^ in eachcounty because time series data on values offarmland and developed land are unavail-able; g and a^ were calculated using the1993-1997 data on county median house-hold income from the U.S. Census Bureau.

Dummy variables for the farm produc-tion regions deñned by the ERS areincluded in the equations to reflect regionaldifferences not captured by the explanatoryvariables. The 10 farm production regionsdefined by the ERS are the Pacific, Moun-tain, Northem Plains, Southern Plains,Lake States, Corn Belt, Delta States,Northeast, Appalachian, and Southeastregions.^ The Southeast was chosen as areference region. CRP acres are concentrat-

An alternative way to define the regional dummies isto use the farm resource regions defined by the ERS.

ed in the Great Plains (Northern Pains andSouthern Plains) and the western Corn Belt,with some increases in the Mountain regionsince the fifteenth sign-up. The descriptivestatistics of all variables used in theempirical analysis are listed in Table 1.

Estimation Methods

Three econometric issues arise in theestimation of equations [11], [13], [14], and[15]. First, these equations are not indepen-dent. The dependent variable of [15] ap-pears on the right-hand side of [14] as anexplanatory variable because the level of thebid rent B* affects the probability of bidacceptance Fr. In addition, both B* and Fraffect the values of farmland and deve-loped land because the expected net returnto farmland, R'^^''= (l~m)A + m[FrB* +(1 -Fr)A], is affected by B* and Fr, wherem = 1 if the parcel is eligible for the CRPand zero otherwise. These endogeneityissues must be addressed. Second, the errorterms (e) may be correlated. For example,because farmers' expectation about thecutoff level of EBI affects both the proba-bility of bid acceptance and the bid rent, if avariable affecting the expectation is omit-ted, the error terms ei and £2 would becorrelated. Likewise, £3 and 64 may becorrelated because there may be an omittedvariable that affects both the values offarmland and the values of developed land.These contemporaneous correlations mustbe taken into account in the estimation.Finally, spatial autocorrelation may existbecause counties located near each othermay be affected by the same omittedvariables (e.g., Bockstael 1996). Spatialautocorrelations have been identified inprevious studies of land values (e.g.. Belland Bockstael 2000; Irwin 2002; Irwin andBockstael 2001).

These econometric issues (endogeneity,contemporaneous correlations, and spatialautocorrelation) are addressed using gener-alized spatial three-stage least squares(GS3SLS) developed by Kelejian and Pru-cha (2004). In the first stage, the modelparameters are estimated using two-stage

86(1) Wu and Lin: Conservation Policy and Land Values

TABLE 1VARIABLES AND DESCRIPTIVE STATISTICS

Variable Description Mean St. Dev.

P"p"hPS

y

Natural amenityClimateMan-made amenityRecreation resourceCRP-¡mRoadsUICrlr2r3r4r5r6r7r8r9rlO

Average CRP rental rate for sign-ups 1-14 ($)Net returns to farmland ($)Max{A,A{\ -m) + m[pb + (\ -p)A]}Farmland values ($)Developed land values ($)Average bid rent per acre at sign-up 15 (S)Probability of acceptance at sign-up 15Sum of environmental scoresMedian household income in 1997 ($)Mean of annual real income growth, 1993-1997Variance of income growth, 1993-1997Natural amenity scale created by ERSAmenity score for temperate climateAmenity score for man-made recreational facilitiesAmenity score for natural recreational resourcesPercentage of fannland enrolled in CRP in sign-ups 1-14Percentage of land eligible for CRP participationInterstate and principal arterial roads (1,000 miles)1993 Urban Influence CodeI if counties in Pacific, 0 otherwise1 if counties in Mountain, 0 otherwise1 if counties in Northern Plains, 0 otherwise1 if counties in Southern Plains, 0 otherwiseI if counties in Lake States, 0 otherwise1 if counties in Com Belt, 0 otherwise1 if counties in Delta States, 0 otherwise1 if counties in Northeast, 0 otherwise1 if counties in Appalachian, 0 otherwise1 if counties in Southeast, 0 otherwise

547781

1,36248,837

500.65

14032,377640

3,420-0.600004.20

45.30585.600.040.080.110.110.080.170.070.070.160.10

16.1678.4574.10

961.9345,052.50

22.460.3134.90

7,514.832,161.272,402.17

1.831.001.001.004.70

29.4086.432.640.200.270.310.310.270.380.260.260.360.31

Noie: CRP, Conservation Reserve Program; ERS, U.S. Department of Agriculture Economic Research Service.

least squares (2SLS) and instrumentalvariable techniques. All exogenous vari-ables are chosen as instrumental variables.The residuals from the 2SLS estimates areused to test for spatial autocorrelation ineach equation using Moran's /-statistic,I=N{è"Wè)M{è'è), where N is the numberof observations, ê is the vector of estimatedresiduals, W is the spatial weight matrixindicating'spatial structure of the data, andM is the standardization factor equal to thesum of the elements of W. We assume theerror structure takes the form s = pWe + y,where p is a scalar and v is a vector ofspherical disturbance with zero mean. W isconstructed in ArcView 3.2 using rookcontiguity criteria, which uses commonboundaries to define neighbors.

If spatial autocorrelation is identified,then in the second stage the residuals fromthe 2SLS are used to estimate the spatial

autoregressive parameter p for each equa-tion utilizing the generalized moment esti-mator (Kelejian and Prucha 2004). Afterthe spatial autoregressive parameter p isestimate^d, data are transformed using thematrix P=M—pW, where M is an A by A'identity matrix. If spatial autocorrelation isnot identified, no data transformation isperformed. The final stage addresses theissue of contemporaneous correlations us-ing seemingly unrelated regression estima-tors. Two simultaneous equation systemsare estimated in this study, one includes theCRP acceptance equation [14] and the bidrent equations [15], and the other includesthe land value equations [11] and [13]. Thesetwo sets of equations are estimated sepa-rately because we have different numbers ofobservations for land values and CRP data.We use the predicted values of B* and Prfrom the first equation system to estimate

10 Land Economics February 2010

, which is then used in the estimationof the simultaneous equation system forland values.

IV. RESULTS

Econometric Estimates

The CRP acceptance and bid rent equa-tions [14] and [15] are estimated using thedata and methods described in the last twosections. The estimated coefficients arepresented in the Appendix (Table Al).Spatial autocorrelations are detected andare adjusted for each equation. Specifically,Moran's /-statistic, with the standard devi-ation listed in parentheses, is 0.13 (0.0135)for the CRP acceptance equation and 0.45(0.0135) for the bid rent equations. The nullhypothesis of no spatial autocorrelationis rejected at the 1% level in each case.The spatial autocorrelation parameter p isestimated to be 0.30 and 0.68 for thetwo equations, indicating positive spatialautocorrelations. The system-weighted R-squared is 0.57. All coefficients except someregional dummies are statistically signifi-cant at the 1% level.

As expected, bids with higher environ-mental scores and lower annual rental ratesare more likely to be accepted into the CRP.Specifically, a 1% increase in the bid rentreduces the probability of acceptance by9%. A large amount of the existing CRPland in a county has a negative effect on theprobability of acceptance because theUSDA is more likely, to target land for theCRP in areas with low participation rates.A CRP bid is more likely to be accepted in acounty with a higher average CRP rentalrate in previous sign-ups. One possiblereason for this result is that counties withhigher CRP rental rates in previous sign-upsmay have fewer CRP applications becauseof higher opportunity costs of participation.Most regional dummies are statisticallyinsignificant in the CRP acceptance equa-tion, indicating that the probability ofacceptance does not vary systematicallyacross regions except for the variationsexplained by the explanatory variables.

All variables expected to affect CRPrental rates are statistically significant atthe 1% level. Counties with higher averageenvironmental scores tend to require lowerrental rates. This may refiect that parcelswith higher environmental scores tend tohave lower land quality and lower oppor-tunity costs of CRP participation. Incontrast, counties with higher net returnsto agriculture and higher average CRPrental rates in previous CRP sign-ups tendto require higher rental payments becausethe opportunity costs of CRP participationin those counties may be higher. Specifical-ly, a $1 difference in the average CRP rentalrate in the previous CRP sign-ups leads to a$0.75 difference in the current bid rents,whereas a $1 difference in the average netreturn to agriculture leads to only a $0.03difference in the current bid rent. Theseresults suggest that farmers rely heavily onprevious CRP rental rates to determinetheir bid rents. Counties with a largeamount of CRP land tend to have lowerCRP bid rents.

Five versions of the land value equationsare estimated. The results, labeled as ModelI to Model V, are reported in the Appendix(Tables A2-A5). Model I is our basic modeland is estimated using the functional formsspecified in equations [11] and [13]. ModelsII to V are estimated using a quadraticfunctional form and alternative measures ofamenities and other locational characteris-tics. Specifically, Models I, II, and III useour own constructed scores of amenities asexplanatory variables, while Models IV andV use ERS's amenity index. Models I, II,and IV use the length of interstate andprincipal arterial roads to refiect accessibil-ity and transportation costs, while ModelsIII and V use ERS's UIC. Only theinteraction terms that are shown to bepossible by the theoretical mode are includ-ed in the developed land value equations.

Overall, all five models fit the data well,with a system-weighted Ä-squared beingabout 0.87 for all models. Most of thecoefficients of interest are statistically sig-nificant at the 5% level or better. Spatialautocorrelations are detected in all models

86(1) Wu and Lin: Conservation Policy and Land Values 11

TABLE 2MARGINAL EFFECT OF SELECTED VARIABLES ON VALUES OF FARMLAND AND DEVELOPED LAND

Variable Model I Model II Model III Model IV Model V

Marginal Effect on Farmland Values (Slacre)

Natural amenityClimate 45* ]29***Man-made amenity I99*** 406***Recreational resource 8 -44UIC

Marginal Effect on Developed Land Values (Slacre)

Natural amenityClimate 293 2,191Man-made amenity Recreation 16,229*** 24,707***Recreational resource -2,090*** -\2\***UIC

42*93***

206***7

-60***

2,312*4,671***1,460*

-5,543***

20

60***

-67***

1,249**

-5,425***

' Significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.

and are adjusted accordingly. Because ofinteraction terms and nonlinear relation-ships, the sign and magnitude of individualcoefficient do not have clear interpreta-tions. For this reason, we calculate themarginal effect of amenity variables andUIC on values of farmland and developedland and report the results in Table 2, F-statistics for the null hypotheses that themarginal effects are zero were calculated toindicate the statistical significance (Judge etal. 1988).

Overall, amenities seem to have a positiveand significant effect on the values offarmland and developed land. The resultsderived from the models that use the ERSnatural amenity scale and those using ourown amenity scores are generally consis-tent. Climate appears to have a positiveeffect on land values, although it is insig-nificant in the developed land value equa-tions in Models I and II. The positive signsuggests households prefer locations withbetter climate. Man-made recreation facil-ities have positive and significant effects onvalues of both farmland and developedland. The recreation facility index is deter-mined by the number of parks, tenniscourts, and golf courses, among otherthings. Counties with more man-maderecreation facilities are more attractive tohouseholds. The coefficient on the index ofnatural recreational resources is sensitive tospecification in both land value equations.Table 2 also reports the marginal effects of

UIC on values of farmland and developedland. The effects of UIC on land values arenegative and statistically significant, indi-cating that land located in counties withlower "urban influence" has lower values.

Effects of the CRP on Land Value

The effects of the CRP on the values offarmland and developed land are evaluatedusing each of the five models, and theresults are reported in Tables 3 and 4. TheCRP has a positive and statistically signif-icant effect on farmland values in allregions. This result is robust in terms ofthe sign and relative magnitude of theeffects. Nationwide, the CRP increased theaverage farmland values by between $18and $25 per acre (1.3'yc^l.8%), with thelargest effect in the Mountain, SouthernPlains, and Northern Plains regions, whereit increased average farmland values bybetween 5% and 14%, 4% and 6%, and 2%and 5%, respectively. These results are notsurprising, given that more than 60% ofCRP lands are located in these three regionsand that CRP rental rates are considerablyhigher than net returns to agriculture in thethree regions, which are generally below $30per acre in the Mountain and the SouthernPlains regions, and below $50 per acre in theNorthern Plains region.

The CRP also increased farmland valuesin the Corn Belt, Appalachian, and Pacificregions. However, the percentage increases

12 Land Economics February 2010

TABLE 3T H E EFFECTS OF THE CONSERVATION RESERVE PROGRAM ON FARMLAND VALUES, BY REGION

Region

PacificMountainNorthern PlainsSouthern PlainsLake StatesCorn BeltDelta StatesNortheastAppalachianSoutheastUnited States

Model I

36*** (2.25)60*** (9.79)28*** (4.52)40*** (6.41)

3*** (0.22)24*** (1.35)18*** (1.62)8*** (0.33)

28*** (1.51)18*** (1.19)25*** (1.84)

Changes

Model II

44*85*33*29*

6*26*12*9*

24*11*25*

** (2.74)"* (13.87)"* (5.32)** (4.65)** (0.43)•*(1.46)''*(1.08)'* (0.37)'•*(1.30)** (0.73)''*(1.84)

in Farmland Values

Model III

35***52***24***23*

6*23*

9*8*

21*9*

^*I t«

^*

i î *

| E *

^ *

22***

(2.18)(8.48)(3.87)(3.69)(0.43)(1.29)(0.81)(0.33)(1.13)(0.59)(1.61)

($/acre)

Model IV

35*55*17*"35**2*

17*"12*"7*

21*"11*"18*"

'*(2.18)** (8.97)•* (2.74)** (5.60)**(0.15)** (0.96)**(1.08)** (0.29)•*(1.13)' (0.73)** (1.32)

Model V

183211302

18107

211118

***(1.12)*** (5.22)***(1.87)*** (4.81)***(0.15)*** (1.91)*** (0.90)*** (0.29)***(1.13)*** (0.73)***(1.32)

Note: Percentages are in parentheses.** Significant at the 5% level; *** significant at the 1% level.

were relatively small. The small effects werea result of lower CRP participation ratesand smaller difference between CRP rentalrates and net returns to crop production inthese regions. The CRP effects account foronly a small percentage of farmland values,because farmland is more productive andvaluable in these regions, with averagefarmland values higher than $1,600 per acrein most counties. The effect of the CRP onfarmland values was smallest in the LakeStates and the Northeast regions. In theLake States, there was little differencebetween CRP rental rates and net returnsto agriculture. The Northeast had thesmallest CRP enrollment among the 10

regions. Only about 0.5% of CRP land islocated in the Northeast region.

The CRP also had a positive andstatistically significant effect on developedland values. However, the percentage in-creases were small in every region. Nation-wide, the CRP increased the average valueof developed land by between $6 and $274per acre, which accounts for less than 0.6%of developed land values. The CRP hadrelatively large effects in the Mountain,Southern Plains, Appalachian, and CornBelt regions. It is not surprising that effectsof the CRP on developed land values arerelatively large in the Mountain and South-ern Plains regions, where the positive and

TABLE 4

EFFECTS OF THE CONSERVATION RESERVE PROGRAM ON DEVELOPED LAND VALUES, BY REGION

Region

PacificMountainNorthern PlainsSouthern PlainsLake StatesCom BeltDelta StatesNortheastAppalachianSoutheastUnited States

Model I

540*** (0.31)843*** (0.78)275*** (0.60)244** (0.61)

63(0.15)277*** (0.67)118*** (0.45)186*** (0.26)347*** (0.98)150(0.44)274*** (0.56)

Changes in Developed Land Values ($/acre)

Model II

549*809*277*202*

75*271*119*184*297*114*273*"

•*(0.12)** (0.74)** (0.60)• (0.50)'•*(O.I8)** (0.65)** (0.45)** (0.26)** (0.84)'* (0.33)** (0.56)

Model III

203*** (0.12)341** (0.31)

- 2 0 (-0.04)19 (0.05)31*** (0.07)

155*** (0.37)15(0.06)

137*** (0.19)205*** (O.S%)

57* (0.17)155*** (0.32)

Model IV

715*** (0.41)901*** (0.83)

-139** (-0.30)368*** (0.92)

-57*** (-0.14)10** (0.02)45 (0.17)42 (0.06)64(0.18)79** (0.23)6 (0.01)

Model V

158*** (0.09)233** (0.21)

-191*** (-0.41)249*** (0.59)- 3 3 (-0.08)

78 (0.30)33 (0.05)53*** (0.07)

125*** (0.35)94*** (0.27)73(0.15)

Note: Percentages are in parentheses.* Significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level.

86(1) Wu and Lin: Conservation Policy and Land Values 13

TABLE 5THE EFFECTS OF THE CONSERVATION RESERVE PROGRAM (CRP) ON DIFFERENT COMPONENTS OF FARMLAND

VALUES, BY REGION

Region

PacificMountainNorthern PlainsSouthern PlainsLake StatesCorn BeltDelta StatesNortheastAppalachianSoutheastUnited States

Value

$/acre

813261364229588655535762688608542

of AgriculturalReturns

% of LandValue

50.742.558.636.843.036.848.231.637.240.139.8

Growth 1Premiumand Option Value

$/acre

792353256397782

1,125575

1,6481,165

903820

% of LandValue

49.357.541.463.257.063.251.868.462.859.960.2

CRP Effect on Valueof Agricultural

Returns

$/acre % Increase

51834259

7392515382437

6.32.2

11.76.21.26.04.72.05.64.06.9

CRP Effect on GrowthPremium and Option

Value

$/acre %

- 1 5- 2 3- 1 4- 1 9

- 4- 1 3

- 7- 7

- 1 0- 6

-12

Increase

-1.9-6.5-5.4-4.8-0.5- I . I-1 .2-0.4-0.9-0.7-1.5

larger effects of the CRP on farmland valuesdirectly contribute to the large increases indeveloped land values. However, the rela-tively large effect of the CRP on developedland values in the Appalachian and theCorn Belt regions is unexpected, given theeffects of the CRP on farmland values aremoderate there. One possible explanation isthat Appalachia and much of the Corn Belt(Ohio, Indiana, and Illinois) are highlydeveloped and are also concentrated withhighly productive farmland. A small reduc-tion in developable land caused by the CRPtranslated into a relatively large increase indeveloped land values in those regions.

Table 5 reports the decomposition offarmland value and the effect of the CRPon agricultural and development compo-nents values based on Model I. Agriculturalreturns (the first term of equation [11])account for 40% of U.S. farmland value,and growth premium and option values (thesum of all terms in equation [11] except thefirst and the error term) account for theother 60%. Our estimate of the share ofgrowth premium and option value is higherthan that by Plantinga, Lubowski, andStavins (2002), who estimated that thefuture rents from development account foronly 10% of the U.S. agricultural landvalue. The difference between the estimatesis caused mainly by their decision to include

the intercept term in the agricultural com-ponents (see their footnote 27). When theintercept shifters for New Jersey, Connecti-cut, and Massachusetts are included in thedevelopment component, their estimatedshares of development components in theagricultural land value for those three statesare 82%, 81%, and 65%, respectively. Asthey pointed out in their paper, whether toinclude the intercept term in the agriculturalor development component is somewhatarbitrary. In this study, we do not face thisarbitrary decision because Model I, whichwas specified based on the theoretical modeland was used to estimate the shares ofagricultural and development componentsin farmland value, does not include anintercept term.

Of the 10 regions, the share of agriculturalretums in farmland value is the largest in theNorthem Plains, where most farmland faceslow development pressure. Growth premi-um and option value account for about 68%of farmland values in the Northeast, highestamong all regions. This estimate is compa-rable with Plantinga, Lubowski, and Sta-vins's estimate for Massachusetts (65%).Consistent with the theory, the CRP had apositive impact on agricultural retums, buta negative impact on growth premiums andoption values. Specifically, the CRP increas-es the value of agricultural retums by about

14 Land Economics February 2010

$37 per acre in the United States, butreduces growth premiums and option valuesby $12 per acre.

V. CONCLUSIONS

As the largest conservation program inU.S. history, the CRP has been evaluated ina number of studies for its economic andenvironmental benefits. However, the ef-fects of the CRP on land values havereceived relatively little attention. Thispaper develops theoretical and empiricalmodels to evaluate the effects of the CRP onvalues of farmland and developed land. Thetheoretical results suggest that the CRPincreases agriculture returns but decreasesgrowth premium and option value. Basedon theoretical analysis, empirical modelsare then estimated to quantify the effect ofthe CRP on values of farmland anddeveloped land. Results suggest that theCRP increased the average fannland valueby between 1.3% and 1.8% in the UnitedSt;ates in 1997. The effects were largest inthe Mountain, Southern Plains, and North-ern Plains regions, where the CRP increasedfarmland values by 5.2% to 14.0%, 3.7% to6.4%, and 2.7% to 5.3%, respectively. TheCRP also had a positive effect on developedland values; however, the percentage in-creases were relatively smaller, although theabsolute increases were much larger. Agri-cultural returns were estimated to accountfor about 40% of the total fannland valuesin the United States, and growth premiumand option value together account for theremaining 60%. Climate and recreationamenities have positive effects on farmlandvalues because they increase both growthpremium and option value.

These results provide useful informationfor the design of land conservation pro-grams. By retiring highly erodible croplandand other environmentally sensitive acreagefor 10 to 15 years, the CRP providessignificant environmental benefits. Howev-er, a permanent easement program has anobvious advantage. In recent years, several

states including Minnesota and Marylandhave used the Conservation Reserve En-hancement Program (CREP) and otherUSDA programs to convert short-termeasements to permanent conservation. Ithas been suggested that since the presentdiscount value of rental payments during a15-year contract equals about 76% of thevalue of a perpetual program (assuming a10% discount rate), states need to pay onlyabout 25% more to secure permanenteasements. Even if a 5% discount rate isassumed, states need to pay only 48% more.Our results suggest that such calculationsare flawed, and 25% additional funding isgenerally not enough to convert a 15-yearcontract to a permanent easement.

CRP payment is calculated based on therelative productivity of soils within thecounty and the local dry land cash rent.Thus, the CRP payments reflect only thestream of agricultural returns, not growthpremium and option value. Our resultsshow that agricultural returns account foronly 40% of the total farmland value, andgrowth premium and option value accountfor the remaining 60%. This suggests thatCRP rental payments during the contractperiod account for only about 30% to 21%(0.40 X 76% to 0.40 X 52%) of land value,where 76% and 52% represent the percent-age of agricultural returns covered by CRPpayments during a 15-year contract for a10% and 5% discount rate, respectively. Theremaining 70% to 79% of land value mustbe compensated to convert a 15-year CRPcontract to a permanent easement. Thatwould be 2.6 to 3.8 times of the total CRPpayment (70%/30% = 2.6). Thus, in areaswhere land has large growth premium andoption value, governments would need topay much more than 25% to convert a 15-year contract to a permanent easement.However, in remote areas where land haslittle growth premium and option value,25% additional funding may be sufficientto secure a permanent easement, particu-larly from landowners with a high discountrate.

86(1) Wu and Lin: Conservation Policy and Land Values 15

APPENDIX

TABLE AlPARAMETER ESTIMATES OF THE PROBABILITY OF ACCEPTANCE EQUATION AND THE OPTIMAL BID RENT EQUATION

Variable

InterceptSB*ACRP-,B-\rlr2r3r4r5r6rlr8r9

Acceptance

Coefficient

-5.19***0.08***

-0.10***—

-4.00***0.06**0.12

-0.270.15

-0.310.06

-1.15***0.160.29

-0.13

Equation

St. Dev.

0.3490.0020.019—1.4060.0240.4860.3360.2940.3070.3220.3310.3380.3800.290

Bid Rent

Coefficient

2.85***-0.02***

0.03***-12.79***

0.75***-5.56***-1.88-0.31-0.17

1.3911.93***

-3.16**0.932.51**

Equation

St. Dev.

0.4430.005

0.0024.6040.0231.8671.2741.3641.1921.3841.3631.4061.5661.154

Note: Number of observations = 2,206. System-weighted R^ = 0.57.** Significant at the 5% level; *** significant at the 1% level.

16 Land Economics February 2010

TABLE A2PARAMETER ESTIMATES OF THE FARMLAND VALUE

EQUATION, MODEL I

TABLE A3PARAMETER ESTIMATES OF THE DEVELOPED LAND

VALUE EQUATION, MODEL I

Variable

j^CRP

g X Roadsg X yg X gQ ^ j^CRP

g X Climateg X Man-made amenity

g X Recreation resource

gXrlgXr2gXr3gXr4gXr5gXr6gXr7gXr8gXr9a X Roadsax yaxa— v R

a X Climatea X Man-made amenitya X Recreation resourceaxrlaxr2axr3axr4axrSaxr6axr7axr8axr9axg

Coefficient

6.6769***-7.16e-6***

2.4e-6***5.0e-6*

-0.0007***0.0386***0.0263***

-0.00470.2081***0.0513*0.0377

-0.02750.05840.01580.00470.0209

-0.0249-0.0064

0.0010***-0.2271***-0.0304***

0.28203.3102***0.1636

-2.4834-6.4826***-7.9109***-3.4203**-5.8818***-1.8051

0.25208.2746***3.6615***0.0004*

Note: R''"'' is calculated using the predicted bid

St. Dev.

0.26670.00017.76e-72.65e-64.6e-50.01000.01100.00660.03670.02830.02870.02340.03430.02480.02730.03130.02140.00510.00010.01670.00410.50230.42620.26872.01211.56311.47451.32691.81681.33351.51601.16761.17300.0003

rents and thepredicted probability of acceptance. The farmland value equationreported in this table and therpnorti>H m TÜKIP A 7 íirp pctii

developed land value equationmiitpH iiciniT (»pnprtiliTpH cníitiül

Variable

Intercept

p" X Roadsfxyp^xgp^xa„Ü w hCRP

p" X Climatep" X Man-made

amenityp" X Recreation

resource

R'^'^'' X Roads

j^CRP ^ p.

j \ X ffhCRP V, Ä/f^Y

R''^^ X ClimatejlCRP y^ Man-made

amenityJlCRP y Recreation

resourceRoadsyClimateMan-made

amenityRecreation

resourcerlr2r3r4r5r6r7r8r9

Coefficient

-2,757.180.0317***0.0004***

-0.0005***-0.0235-0.0114-2.9862**

-6.3869***

1.9253**-0.3720***

0.00070.0063**0.03770.04305.1441

61.6576***

3.7066-12.6217

0.9562***3,324.58*

21,679.98***

-4,966.70***39,878.5***71,433.5***14,793.9***9,657.0**

-12,048.5**-4,365.1

6,096.9-4,884.6-3,621.5

St. Dev.

2,547.000.01070.00010.00020.02160.00611.0399

0.9050

0.93610.12580.00070.00330.28940.02439.4251

10.5946

8.281619.64370.1755

1,745.20

2,096.20

,434.405,926.104,598.504,398.304,060.605,280.204,110.704,341.905,106.703,688.10

p g g pthree-stage least squares (GS3SLS). Number of observations =2,851. System-weighted R^ = 0.87.

• Significant at the 10% level; ** significant at the 5% level;•*• significant at the 1% level.

Note: The farmland value equation reported in Table A2 andthe developed land value equation reported in this table areestimated using generalized spatial three-stage least squares(GS3SLS). Number of observations = 2,851. Systetn-weightedR^ = 0.87.

* Significant at the 10% level; •* significant at the 5% level;*•* significant at the 1% level.

86(1) Wu and Lin: Conservation Policy and Land Values 17

TABLE A4PARAMETER ESTIMATES FOR THE FARMLAND VALUE EOUATION, MODELS I I - V

Variable

Interceptj^CRP

gaNatural amenityClimateMan-made amenityRecreation resourceRoadsUIChCRPl

?Natural amenity^Climate^Man-made amenity^Recreation resource^Roads^UIC?j^CRP ^

bCRP „ „

¡^CRP y^ Natural amenityj^CRP y^ Climate

J^CRP X Man-made amenityj^CRP y Recreation resourcej^CRP y^ Hoadsj^CRP y y¡(^

gxag X Natural amenityg X Climateg X Man-made amenityg X Recreation resourceg X RoadsgX UICa X Natural amenitya X Climatea X Man-made amenitya X Recreation resourcea X Roadsax UICRoads X Natural amenityRoads X ClimateRoads X Man-made amenityRoads X Recreation ResourceUIC X Natural amenityUIC X ClimateUIC X Man-made amenityUIC X Recreation Resourcerl

Model II

247.94***441***6.05e-30.02**

227.90***252.44***49.150.85

-7.00e-5***-6.35e-6**

1.29e-6**

66.18***-76.21***-7.64-0.01***

- 2.10e-4***1.47e-4***

-1.27***0.70***

-0.68***-4.59e-3**

-1.69e-6*

-8.40e-30.02**

-1.82e-3l.Ole^

2.92e-35.00e-3

-5 .40e^147e^***

-1.101.17***0.17

935.61***

Estimate

Model III

224.04***6.14***0.010.06***

215.83***358.06***-1.84

27.32-9.00e-5***-7.87e-6***

1.24e-6**

60.76***-10.18***-8.40*

-2.32***1.50e-4***I.26e-4***

-1.28***0.08

-0.55***

-0.46***5.02e-7

-2.37e-30.02**

-4.66e-3

5.88e-4

4.92e^-4.16e-3

7.27e^

-7.29e-3***

-3.53-28.18***

9.55*793.48***

Model IV

194.83***2.66***

-5.01e-37.60e-3

5L23***

2.04***

2.00e-6-9.67e-6***

2.36e-6***5.51**

-7.70e-4**

-2.50e-4***1.82e-4***0.43***

4.58e-3***

-6.47e-7

2.67e-4***

-7.16e-3***

8.2e-5

-0.26***

642.28***

Model V

248.81***6.11***0.03*0.04***

126.35***

-24.46-7.00e-5***-7.67e-6***

1.65e-6***1.52

1.67-1.8e-4***

1.39e-4***0.15***

-0.57***6.91e-7

-3.32e-3-6.64e-3***

-5.74e-3***

-10.55***

685.39***

table continued on following page

18 Land Economics February 2010

TABLE A4PARAMETER ESTIMATES FOR THE FARMLAND VALUE EQUATION, MODELS I I - V

Continued

Estimate

Variable

r2r3r4r5r6rlr8r_9

Nole: The farmland value equation reported in this table and the developed land value equation reported in Table A5 are estimatedusing generalized spatial three-stage least squares (GS3SLS) for each model. Number of observations = 2,851. System-weighted 7? = 0.87for each model.

* Significant at the 10% level; ** significant at the 5% level; *•* significant at the 1% level.

Model II

160.56-104.41-324.87***

278.46**601.30***-5.31

1,640.10***708.56***

Model III

118.21-87.58

-309.14***250.33*572.09***36.63

1,616.72***707.32***

Model IV

-29.11-92.06-90.17319.19**688.39***177.76

2,095.62***770.56***

Model V

90.69***-16.94-68.70398.91***716.95***195.03

2,027.49***760.79***

86(1) Wu and Lin: Conservation Policy and Land Values 19

TABLE A5PARAMETER ESTIMATES FOR THE DEVELOPED LAND VALUE EQUATION, MODELS I I - V

Estimate

Variable Model II Model III Model IV Model V

Interceptj^CRP

ga

i

Natural amenityClimateMan-made amenityRecreation resourceRoadsUICj^CRP

Natural amenity^Climate^Man-made amenity^Recreation resource^Road/UIC?

y^ fifatural amenityf X Climate¡{CRP y^ Man-made amenityjfCRP ^ Recreation resource

' X RoadsX UIC

Roads X Natural amenityRoads X ClimateRoads X Man-made arnenityRoads X Recreation

resourceUIC X Natural amenityUIC X ClimateUIC X Man-made amenityUIC X Recreation resourcerlr2r3r4r5rorlr8r9

11,886.58***61.58***

-2.18***2.72***2.31e-4

-8.00e--1.60e-4**

4,963.67***17,786.51***-1749.59

82.40***

-2.70e-3***

-1,125.00-4,769.40***

-277.41-0.42***

-14.10*26.44***10.80

-0.21**

-28.10**82.83***13.03

160,295.60***85,559.43***21,190.61***

8,077.42*6,969.769,380.34**1,553.03

17,926.73***2,707.72

32,700.81***111.10***- 0.96**

1.57***3.39e-4**

-8.00e-5**

-1,952.4820,396.41***

-1,872.00

-2,1086.50***-3.40e-3***

1,864.22**-359.80***-372.33*

1,495.91***

-11.52*2.49

14.49**

-14.81***

927,-2,844,

385,161,634100,585,30,427,15,395,14,190,15,595,11,147,23,784,12,918,

94***45***

4540***

80***77***77***87***50***64**98***97***

6,849.01***12.34

-2.74***2.87***

7.00e-5**

1,258.77**

194.99***

732.89***

-0.09***

25.30***

0.07

1.84

123,496.40***64,044.16***13,566.22***8,102.67*2,699.35

10,059.34**4,159.05

33,193.82***5,117.96

39,437.21***108.13***- 1.34***

1.44***0.42e-4***

-l.OOe-4***-1.30e-4***

4,031.25***

-25,281.40***-3.31e-3***818.07***

1,862.91***12.06***

-15.59***

-496.14***

133,580.80***87,345.16***27,505.60***19,246.03***14,891.74***16,544.13***13,820.66***37,402.49***12,326.99***

Note: The farmland value equation reported in Table A4 and the developed land value equation reported in this table are estimatedusing generalized spatial three-stage least squares (GS3SLS) for each model. Number of observations = 2,851. System weighted Ä^ = 0.87for each model.

* Significant at the 10% level; ** significant at the 5% level; •** significant at the 1% level.

20 Land Economics February 2010

References

Alston, Julian M. 1986. "An Analysis of Growth inU.S. Farmland Prices: 1963-82." AmericanJournal of Agricultural Economics 68 (1): 1-9.

Barnard, Charles H., Gerald Whittaker, DavidWastenbarger, and Mary Ahearn. 1997. "Evi-dence of Capitalization of Direct GovernmentPayments into U.S. Cropland Values." Ameri-can Journal of Agricultural Economics 79 (5):1642-50.

Bell, Kathleen P., and Nancy E. Bockstael. 2000."Applying the Generalized-Moments Estima-tion Approach to Spatial Problems InvolvingMicrolevel Data." Review of Economics andStatistics 82 (1): 72-82.

Bockstael, Nancy E. 1996. "Modeling Economicsand Ecology: The Importance of a SpatialPerspective." American Journal of AgriculturalEconomics 78 (5): 1168-80.

Capozza, Dennis R., and Robert W. Helsley. 1990."The Stochastic City." Journal of Urban Eco-nomics 2^ {I): 187-203.

Capozza, Dennis R., and Yuming Li. 1994. "TheIntensity and Timing of Investment: The Caseof Land." American Economic Review 84 (4):889-904.

Castle, Emery N., and Irving Hoch. 1982. "FarmReal Estate Price Components: 1920-1978."American Journal of Agricultural Economics 68(1):8-18.

Chicoine, David L. 1981. "Farmland Values at theUrban Fringe: An Analysis of Sales Prices."Land Economics 57 (3): 353-62.

Cho, Seong-Hoon, Junjie Wu, and William Bog-gess. 2003. "Measuring Interactions amongUrbanization, Land Development, and PublicFinance." American Journal of AgriculturalEconomics 85 (4): 988-99.

Clark, J. Stephen, Kurt K. Klein, and Shelley J.Thompson. 1993. "Are Subsidies Capitalizedinto Land Values? Some Time Series Evidencefrom Saskatchewan." Canadian Journal ofAgricultural Economics A\ (2): 155-68.

Colwell, Peter F., and Henry J. Munneke. 1997."The Structure of Urban Land Prices." Journalof Urban Economics 41 (3): 321-36.

Coulson, N. Edward, and Robert F. Engle. 1987."Transportation Costs and the Rent Gra-dient." Journal of Urban Economics 2\ (3): 287-97.

Délier, Steven C , Tsung-Hsiu Tsai, David W.Marcouiller, and D. B. K. English. 2001. "TheRole of Amenities and Quality of Life in RuralEconomic Growth." American Journal of Agri-cultural Economics 83 (2): 352-65.

Feather, Peter, Daniel Hellerstein, and LeRoyHansen. 1999. Economic Valuation of Environ-mental Benefits and the Targeting of Conserva-tion Programs: The Case of the CRP. Agricul-tural Economics Report 778. Washington, DC:U.S. Department of Agriculture, EconomicResearch Service.

Goodwin, Barry K., Ashok K. Mishra, andFrançois. N. Ortalo-Magné. 2003. "What'sWrong with Our Models of Agricultural LandValues?" American Journal of Agricultural Eco-nomics 85 (3): 744-52.

Goodwin, Barry K., and François N. Ortalo-Magné. 1992. "The Capitalization of WheatSubsidies into Agricultural Land Values."Canadian Journal of Agricultural Economics 40(1): 37-54.

Hardie, Ian W., Tulika A. Narayan, and Bruce L.Gardner. 2001. "The Joint Influence of Agri-cultural and Nonfarm Factors on Real EstateValues: An Application to the Mid-AtlanticRegion." American Journal of AgriculturalEconomics 83 (1): 120-32.

Hushak, Leroy J., and Kazem Sadr. 1979. "ASpatial Model of Land Market Behavior."American Journal of Agricultural Economics 61(4): 697-701.

Irwin, E. G. 2002. "The Effects of Open Space onResidential Property Values." Land Economics78 (4): 465-80.

Irwin, Elena G., and Nancy E. Bockstael. 2001."The Problem of Identifying Land Use Spill-overs: Measuring the Effects of Open Space onResidential Property Values." American Journalof Agricultural Economics 83 (3): 698-704.

Judge, George G., R. Carter Hill, William E.Griffiths, Helmut Lütkepohl, and Tsoung-ChaoLee. 1988. Introduction to the Theory andPractice of Econometrics. 2nd ed. New York:Wiley.

Just, Richard E., and John A. Miranowski. 1993."Understanding Farmland Price Changes."American Journal of Agricultural Economics 75(1): 156-68.

Kelejian, Harry H., and Ingmar R. Prucha. 2004."Estimation of Simultaneous Systems of Spa-tially Interrelated Cross Sectional Equations."Journal of Econometrics 118 (1-2): 27-50.

Kirwan, Barrett. 2008. "The Incidence of U.S.Agricultural Subsidies on Farmland RentalRates." Working paper. College Park, MD:Department of Agricultural and ResourceEconomics, University of Maryland.

Kirwan, Barrett, Ruben N. Lubowski, and Mi-chael J. Roberts. 2005. "How Cost EffectiveAre Land Retirement Auctions? Estimating theDifference between Payments and Willingness

86(1) Wu and Lin: Conservation Policy and Land Values 21

to Accept in the Conservation Reserve Pro-gram" American Journal of Agricultural Eco-nomics %1 {5): 1239-47.

Kowalski, Joseph G., and Christos C. Paraskevo-poulos. 1990. "The Impact of Location onUrban Industrial Land Prices." Journal ofUrban Economics 27 (1): 16-24.

Latacz-Lohmann, Uwe, and Carel V. van derHamsvoort. 1997. "Auctioning ConservationContracts: A Theoretical Analysis and anApplication." American Journal of AgriculturalEconomics 79 (2): 407-18.

Lence, Sergio H., and Ashok K. Mishra. 2003."The Impacts of Different Farm Programs onCash Rents." American Journal of AgriculturalEconomics %5 O): 753-61.

McDonald, John F., and Daniel P. McMillen.1998. "Land Values, Land Use, and the FirstChicago Zoning Ordinance." Journal of RealEstate Finance and Economics 16 (2): 135-50.

Osborn, C. Tim, and Kazim Konyar. 1990. "AFresh Look at CRP." In Agricultural Outlook,33-37. Washington, DC: U.S. Department ofAgriculture, Economic Research Service.

Palmquist, Raymond, and Leon E. Danielson.1989. "A Hedonic Study of the Effects ofErosion Control and Drainage on FarmlandValues." American Journal of AgriculturalEconomics 7\ (1): 55-62.

Plantinga, Andrew J., Ruben N. Lubowski, andRobert Stavins. 2002. "The Effects of PotentialLand Development on Agricultural Land Pric-es." Journal of Urban Economics 52 (3): 561-81.

Plantinga, Andrew J., and Douglas J. Miller. 2001."Agricultural Land Values and the Value ofRights to Future Land Development." LandEconomics 77 (1): 56-67.

Roberts, Michael J., Barrett Kirwan, and JeffreyHopkins. 2003. "The Incidence of GovernmentPayments on Agricultural Land Rents: TheChallenges of Identification." American Journalof Agricultural Economics 85 (3): 762-69.

Rosenthal, Stuart S., and Robert W. Helsley. 1994."Redevelopment and the Urban Land PriceGradient." Journal of Urban Economics 35 (2):182-200.

Rosine, John, and Peter Helberger. 1974. "ANeoclassical Analysis of the U.S. Farm Sector,1948-1970." American Journal of AgriculturalEconomics 56 (4): 717-29.

Ryan, James, Charles Barnard, Robert Collender,and Ken Erickson. 2001. "Government Pay-ments to Farmers Contribute to Rising LandValues." In Agricultural Outlook, 22-26. Wash-ington, DC: U.S. Department of Agriculture,Economic Research Service.

Schmitz, Andrew, and Richard E. Just. 2002. "TheEconomics and Politics of Farmland Values."In Governtnent Policy and Farmland Markets:Maintenance of Landowner Wealth, ed. C.Moss and A. Schmitz. Ames, IA: Iowa StatePress.

Shi, Yue Jin, Timothy T. Phipps, and Dale Colyer.1997. "Agricultural Land Values under Urban-izing Influences." Land Economics 73 (1):90-100.

Shoemaker, Robbin. 1989. "Agricultural LandValues and Rents under the ConservationReserve Program." Land Economics 65 (2):131-37.

Shonkwiler, J. S., and J. E. Reynolds. 1986. "ANote on the Use of Hedonic Price Models in theAnalysis of Land Prices at the Urban Fringe."Land Economics 62 (1): 58-61.

Sullivan, Patrick, Daniel Hellerstein, Leroy Han-sen, Robert Johansson, Steven Koenig, RubenLubowski, William McBride, David McGrana-han, Michael Roberts, Stephen Vogel, andShawn Bucholtz. 2004. The Conservation Re-serve Program: Economic Implications for RuralAmerica. Agricultural Economic Report 834.Washington, DC: U.S. Department of Agricul-ture, Economic Research Service.

U.S. Department of Agriculture (USDA). 1997.FSA Handbook: Agricultural Resource Conser-vation Program 2-CRP (Revision 3). Washing-ton, DC: Farm Service Agency.

. 2003. Conservation Reserve Program(CRP) Sign-up 26: Questions and Answers.News Release No. Q&A0294.03. Washington,DC: Office of Communications, News Room460-A.

. 2004. CRP Contract Summary and Statis-tics: Monthly Summary. December 2004. Wash-ington, DC: Farm Service Agency.

Vitaliano, Donald F., and Constance Hill. 1994."Agricultural Districts and Farmland Prices."Journal of Real Estate Finance and Economics 8(3): 213-23.

Wagner, John E., and Steven C. Délier. 1998."Measuring the Effects of Economic Diversityon Growth and Stability." Land Economics 74(4): 541-56.

Wu, JunJie. 2000. "Slippage Effects of the Con-servation Reserve Programs." American Journalof Agricultural Economics 82 (4): 979-92.

Young, C. Edwin, and C. Tim Osborn. 1990. TheConservation Reserve Program: An EconomicAssessment. Agricultural Economic Report 626.Washington, DC: U.S. Department of Agricul-ture, Economic Research Service.

Copyright of Land Economics is the property of University of Wisconsin Press and its content may not be

copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written

permission. However, users may print, download, or email articles for individual use.