slippage effects of the conservation reserve program
TRANSCRIPT
Amer. J. Agr. Econ. 82(4) (November 2000): 979–992Copyright 2000 American Agricultural Economics Association
Slippage Effects of theConservation Reserve Program
JunJie Wu
Each year, billions of dollars of public funds are expended to purchase conservation easements onfarmland. One unintended impact of these programs is that they may bring non-cropland into cropproduction. Such a slippage effect can be caused by increased output prices and by substitutioneffects. This article shows that for each one hundred acres of cropland retired under the Conser-vation Reserve Program (CRP) in the central United States, twenty acres of non-cropland wereconverted to cropland, offsetting 9% and 14% of CRP water and wind erosion reduction benefits,respectively. Implications of these results for the design of conservation programs are discussed.
Key words: conservation programs, environmental benefits, land use changes, slippage effects.
Few programs in recent U.S. history havehad such a large and sweeping effect onfarmland use as the Conservation ReserveProgram (CRP). Established under the Con-servation Title of the 1985 Food Security Actand re-authorized in subsequent Farm Bills,the CRP contained approximately 33 millionacres of cropland as of January 1997, with anannual cost of over $1.6 billion (Osborn andRibaudo, p. 287). Annual erosion reductionon CRP land totaled 626 million tons as ofJanuary 1997, or about 19 tons/acre/year, gen-erating substantial social benefits. For exam-ple, CRP wildlife recreation benefits alonewere estimated to be $428 million per year(Feather, Hellerstein and Hansen, p. iii).Although the CRP was not intended specif-ically as a land use policy, the magnitude ofits impact on farmland use and rural com-munities has been recognized (Young andOsborn).One unintended impact of the CRP, which
may compromise its primary objectives,1 isthat it may cause non-cropland to be con-verted into crop production. For example, the
JunJie Wu is assistant professor in the Department of Agri-cultural and Resource Economics at Oregon State University.The author thanks two anonymous referees for their thoroughreview of this paper. Useful comments from Bill Boggess, DavidZilberman, Bruce Babcock, and participants of the 1999 RuralLand Use Change Workshop at University of Maine are alsogratefully acknowledged. Technical paper number 11672 of theOregon Extension and Experiment Station.1 The original goals of the CRP were (P.L. 99–198): a) reducing
soil erosion, b) protecting soil productivity, c) reducing sedimen-tation, d) improving water quality, e) improving fish and wildlifehabitat, f) curbing production of surplus commodities, and g) pro-viding income support for farmers.Also, see Feather, Hellerstein,and Hansen.
Economic Research Service’s CRP databaseshows that 17.63 million acres of croplandwere retired under the CRP in the CornBelt, Lake States and Northern Plain by 1992,but total cropland acres were reduced byonly 13.69 million acres in the region from1982 to 1992 according to the 1982 and 1992National Resources Inventories (NRI). Thisimplies that at least 3.69 million acres of non-cropland were converted to cropland in theperiod of 1982–92.2 Many factors may havecontributed to the slippage effect, but crop-land retirements under the CRP may be oneof them.Slippage effects may exist for at least
two reasons. First, some non-cropland maybe brought into production as a resultof increased output prices associated withreduced production on CRP land. Priceeffects of conservation efforts have beendocumented. For example, Berck and Bent-ley estimated that governmental takingsof old-growth redwood for inclusion inthe Redwood National Park significantlyincreased redwood lumber prices. The 1978
2 Using the 1982 and 1987 NRI data, we estimated that10.25 million acres of noncropland were converted to croplandfrom 1982 to 1992, 1.16 million acres of noncropland were con-verted to cropland and then to CRP land. Thus total acreage ofnoncropland converted to cropland is 11.41 million acres. Dur-ing the same period, only 8.35 million acres of cropland wereconverted to noncropland (excluding CRP land). Thus, the netincrease in total cropland is 3.06 million acres, which is smallerthan the estimate of 3.94 million acres based on the ERS’s CRPdatabase.The discrepancy arises because the NRI underestimatesthe total CRP acreage by 0.88 million acres for the study region.In this analysis, the ERS’s CRP database is used to estimate CRPacreage.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
980 November 2000 Amer. J. Agr. Econ.
taking alone increased redwood lumberprices by 26%. Ironically, these priceincreases led to increased profits and harvest-ing of old-growth redwood on other lands.Another possible reason for slippage is
substitution effects. When some croplandis taken out of production, farmers maysubstitute other land for crop productionbecause of scale economies and fixed inputeffects (see the detailed discussed in thenext section). However, no CRP provisionprohibits participating farmers from convert-ing non-cropland to cropland. The “sod-buster” and “swampbuster” provisions of the1985 Food Security Act prohibit farmers whoconvert highly erodible land or wetland intocrop production from receiving price andincome support payments under the com-modity provisions. But temporal cycles in theagricultural economy, as well as definitional,implementation, and enforcement problems,have hampered the effectiveness of theseprovisions (Heimlich 1995; Miranowski andCochran).The primary objective of this article is to
determine if a slippage effect was presentin the CRP and, if so, how large was theeffect. The objective is achieved by combin-ing the Economic Research Service’s county-level CRP file with data from the NationalResources Inventories (NRI) and the Censusof Agriculture. The 1982 and 1992 NRIs wereused to estimate the land use changes from1982 to 1992. The CRP database was used toestimate the acres retired during the period.The 1982 and 1992 Census of Agriculture wasused to estimate changes in land value, farmsize, and other farm characteristics affectingland use. Regression analysis was then usedto determine if land retirements under theCRP increased conversions of non-croplandto cropland. To make the analysis more man-ageable, we focused on twelve states in thecentral United States where more than halfof the CRP acres were concentrated.Previous analyses of CRP efficiency have
demonstrated that specific benefits fromthe CRP could have been achieved at alower cost than what was actually incurred.Babcock et al. (1996) estimated that 90%of the available water erosion benefits couldhave been achieved with 50% of the CRPbudget. Heimlich and Osborn compared twohypothetical enrollment patterns and foundthat selecting five million CRP acres to renewon the basis of maximizing soil erosion reduc-tion per dollar would cost $0.54 per ton
of soil saved compared to $1.01 under atargeting criterion of minimizing total cost.Reichelderfer and Boggess estimated that thecost per ton of soil erosion reduction couldhave been decreased by 15% if erosion ben-efits were maximized. Ribaudo showed howtotal benefits of the CRP could be increasedif the off-site benefits were accounted for.Osborn and Thurman analyzed recent CRPsignups and found that the CRP began toenroll land based on comparisons of environ-mental benefits and government costs withan objective of improving the environmentalperformance of the CRP. These studies, how-ever, have not examined the slippage effectsof the CRP.In the next section, I model the slippage
effects associated with output price increasesand substitution effects and show that slip-page reduces environmental gains of a con-servation program, and in some cases, maymake it counterproductive. Subsequent sec-tions discuss data, empirical specifications,and land use changes for twelve states in thecentral United States. Empirical results andpolicy implications are discussed in the lasttwo sections.
Modeling Slippage Effects
In this section, I first model the slippageeffects associated with output price increases.I show that when the output demand is suf-ficiently inelastic, slippage may make a pro-gram counterproductive (resulting in neg-ative net environmental benefits). Then Ipresent a model to show that substitutioneffects also give landowners incentives toconvert non-cropland to cropland when somecropland is retired under a conservationprogram.
Output Price Feedback Effects
I analyze the situation where some croplandis retired from crop production for environ-mental benefits (e.g., soil erosion reduction)in a nation or a large region. Land quality isdifferentiated by the output and environmen-tal benefits per acre. Let y be the output peracre, and let b be the environmental bene-fit per acre if the land is retired. Across thelandscape y and b vary and have a joint prob-ability distribution function of s(y� b), wherey varies from 0 to y and b varies from 0to b. Production cost per acre is assumed to
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
Wu Slippage Effects of the Conservation Reserve Program 981
Figure 1. Slippage effects from output priceincreases
be constant and is denoted by c.3 Withoutthe land retirement program, all land withπ = p0y − c ≥ 0 or y ≥ (c/p0) is assumed tobe used in crop production (figure 1), wherep0 is the initial output price.Suppose some cropland is targeted for con-
servation practices. Each targeting criterioncorresponds to selecting a subset of (y� b)from the base resource set R0 ≡ {(y� b);c/p0 ≤ y ≤ y ; 0 ≤ b ≤ b}. The first ninesignups of the CRP, conducted under author-ity of the 1985 Food Security Act, were sub-ject to mandatory minimum annual enroll-ment levels as established in the Act. In aneffort to meet these enrollment levels, bidswere accepted as long as land eligibility crite-ria were met and the rental rate requested bythe producer did not exceed a USDA maxi-mum acceptable rental rate (MARR) estab-lished for a multiple county area or state(Economic Research Service, pp. 294–95).Thus, the experience of CRP implementationbefore 1990 was consistent with acreage max-imization (Reichelderfer and Boggess). Landwith lower yields had greater incentives toenroll because the MARR offered a rentalpremium. Let y∗ be the highest yield of landsaccepted into the program. Then, land in N0will be retired from crop production (seefigure 1). Because of the output reduction,
3 Relaxing this assumption would complicate our model, butwould not change our basic result as long as the convertible landhas a lower yield than the currently cropped land. Empirical evi-dence indicates that crop yields are generally lower on convertedcropland than on other cropland. For example, the NRI datashow that the average land capability class of the converted crop-land is higher (i.e., has a lower quality) than the other cropland(see table 2). In addition, the USDA’s Area Study Data showthat crop yields and land capability classes are negatively corre-lated across fields. For example, the average corn yield in centralNebraska Basin is 143 bushel per acre on the first class land, butonly 120 bushels on classes II and III land and 113 bushels onland with a capability class IV or higher.
the price of the output will increase.4 As aresult, the marginal land in N1 will be putinto production. The net reduction in crop-land acreage equals
A = A(N0)−A(N1)(1)
=∫ y∗
c/p0
∫ b
0s(y� b)dydb
−∫ c/p0
c/p1
∫ b
0s(y� b)dydb�
The net environmental gain equals
B = B(N0)− B(N1)(2)
=∫ y∗
c/p0
∫ b
0bs(y� b)dydb
−∫ c/p0
c/p1
∫ b
0bs(y� b)dydb�
The output price p1 after the land retirementis defined by the market equilibrium condi-tion:
∫ y
y∗
∫ b
0ys(y� b)dydb(3)
+∫ c/p0
c/p1
∫ b
0ys(y� b)dydb = D(p1)�
where D(p) is the market demand curve forthe output.To see that expressions (1) and (2) can both
be negative, note that when the market isin equilibrium, the change in demand mustequal the change in supply under the conser-vation program:
A(N0)E(y|N0)−A(N1)E(y|N1)(4)
= D(p0)−D(p1)�
where E(y|Ni) is the average yield for landin Ni. As the demand for the output becomesmore inelastic, both sides of equation (4)become smaller. Because E(y|N1) < E(y |N0),A(N1) will eventually become greater thanA(N0). Specifically, if the demand is perfectlyinelastic, A(N0)E(y|N0)−A(N1)E(y|N1) = 0,which implies that A(N0) < A(N1) and A <0. In this case, more land will be activated forcrop production than retired. Furthermore, if
4 While reduced output would tend to raise the commodityprice in a static setting, it may only serve to mitigate further pricedeclines that may occur due to other market factors in a dynamicsetting. We thank a referee for pointing this out.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
982 November 2000 Amer. J. Agr. Econ.
environmental benefits and crop yields arenegatively correlated, the net environmentalgain B is also negative.5To provide further insights, we assume that
(y� b) follows the bivariate normal distribu-tion, N(µy� µb� σy� σb� ρ), where µy and µb
are the mean of crop yields and environ-mental benefits, σy and σb are their standardderivation, and ρ is the correlation coefficient.Under this normal distribution assumption, A, Y , and B have the following relation-ship (see the Appendix):
σy B = (µbσy − ρµyσb)(5)
× A− ρσb Y�
where Y = D(p1) − D(p0) ≤ 0.Equation (5), together with equation (4),indicates that if the demand is perfectlyinelastic ( Y = 0), both A and B arenegative if ρ < (µbσy)/µyσb. In this case,more land will be put into production thanretired, and the conservation program willbe counterproductive. On the other hand, ifρ > (µbσy)/µyσb, A < 0 and B > 0, whichimplies that the conservation program willresult in a positive environmental gain evenif more land will be activated than retired.Natural resource agencies tend to target
resources for conservation based on biolog-ical, agronomic or hydrologic criteria. Forexample, the U.S. Fish and Wildlife Service,in its Wetland Reserve and similar programs,targets wetlands and other water resourcesbased primarily on biological criteria. Inrecent signups, the CRP began to enroll landbased on comparisons of environmental ben-efits and contract costs. By using a similarmodel, it can be shown that these criteria mayalso cause slippage.
Substitution Effects
Another possible cause of slippage is sub-stitution effects, which can be understood
5 Empirical evidence indicates that the correlation between anenvironmental benefit and crop yields can be positive, negative,or zero, depending on the specific environmental benefit consid-ered. For example, Babcock et al. (1997) found that CRP rentalrates are negatively correlated with wind erosion (with a correla-tion coefficient of −0�26), but are positively correlated with watererosion and wildlife habitat benefits (with a correlation coeffi-cient of 0.19 and 0.14, respectively). The correlation between theCRP rental rate and the groundwater vulnerability index is nega-tive, but is close to zero (with a correlation coefficient of −0�02).In another study, Heimlich (1989) compared the distributions ofcrop yields on highly erodible land and other land and found thatthe two distributions are almost identical, indicating that retiringhighly erodible, physically marginal cropland is not synonymouswith retiring less productive, economically marginal cropland.
by considering a farmer’s land use decisions.Suppose a farm has three types of land:high-quality, medium-quality, and low-qualityland. Suppose all high-quality land is devotedto crop production, and all low-quality landis devoted to a non-cropping activity. Themedium-quality land is divided between cropproduction and the non-cropping activity, andthe allocation depends on the output pricesand production costs. Let AH , AM , and AL
be the total acreage of high, medium, and lowquality land, respectively. Let πC(AH�AMC)and πN (AL�AMN ) be the total profit func-tions for crop production and the non-cropping activity, where AMC and AMN areacres of medium-quality land allocated tocrop production and the non-cropping activ-ity. It is assumed that ∂2πC/∂A
2MC ≤ 0 and
∂2πN /∂A2MN ≤ 0. Without the CRP, the
medium-quality land would be allocated suchthat
∂πC(AH�AMC)
∂AMC
= ∂πN (AL�AMN )
∂AMN
�(6)
If the left-hand side is greater than the right-hand side, more land would be allocated tocrop production, and vice versa. This condi-tion is shown in figure 2, where the distancebetween the two vertical axes equals the totalacreage of the medium quality land. Thus,graphically,AMN equals the distance between0’ and AMC .Now, suppose ACRP acres of medium-
quality cropland are enrolled into the CRP.As a result, ∂πC(AH�AMC − ACRP )/∂AMC >∂πN (AL�AMN )/∂AMN . Because the marginalprofit from crop production is greater thanthe marginal profit from the non-croppingactivity, the farmer would have an incen-tive to convert some medium-quality landfrom the non-cropping activity into crop pro-duction.Without government restrictions, thefarmer would switch AS acres of medium-quality land from the non-cropping activityto crop production such that
∂πC(AH�AMC −ACRP +AS)
∂AMC
(7)
= ∂πN (AL�AMN −AS)
∂AMN
�
Thus, in equilibrium, the total croplandacreage would be reduced by (ACRP −AS) instead of ACRP . Graphically, the landbetween (AMC − ACRP ) and AMC is retired,land between AMC and (AMC + AS) is con-verted to crop production, and the land
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
Wu Slippage Effects of the Conservation Reserve Program 983
Figure 2. Slippage effects from land substitution
between (AMC −ACRP +As) and AMC is thenet reduction in total cropland acreage (seefigure 2).Up to now, we have shown that both out-
put price increases and substitution effectscan cause slippage effects. The question iswhether a slippage effect exists in the CRP,and if so, how large is the effect.
An Empirical Analysis ofSlippage Effects in the CRP
This section discusses the data and empiricalmodel that we used to analyze the slippageeffects in the CRP. The empirical analysisfocuses on twelve states in the Corn Belt,Lake States and Northern Plains (figure 3), aregion that accounted for over 50% of totalCRP acreage and experienced large landuse changes in many areas during 1982–92(figure 4). Overlaying figure 3 upon figure 4reveals that counties with large CRP acreagetend to have more non-cropland convertedto cropland. In this section, we use regres-sion analysis to estimate the acreage of non-cropland converted to cropland as a result ofland retirements under the CRP.
Empirical Specification and Data
A multivariate regression model is usedto examine the impact of the CRP on
the acreage of non-cropland converted tocropland:6
Ai = α+ βCRPi +Xiγ + εi�(8)
where i is an index of Agricultural StatisticsDistricts (ASD) as defined by the NationalAgricultural Statistics Service, Ai is theacreage of non-cropland converted to crop-land between 1982 and 1992 as a proportionof the total land area in ASD i, CRPi is theacres of cropland enrolled in the CRP by1992 as a proportion of the total land area inASD i, and Xi is a vector of other land andfarm characteristics that affect land use con-versions, including changes in farm size andpopulation. The empirical model was speci-fied at the ASD level because reliable esti-mates of land use changes for the whole studyregion were not available at a more disaggre-gate level, as will be discussed.The land use data for this analysis were
derived from the 1982 and 1992 NationalResource Inventories (NRI). The NRI isconducted every five years by the NaturalResource Conservation Service (NRCS) todetermine the status, condition, and trend ofthe nation’s soil, water, and other relatedresources at more than 800,000 sites across
6 This specification precludes us from separating the substitu-tion effect from the output price effect. Ideally, we would like toexamine how the CRP affected output prices and how changesin output prices in turn affected land conversions. However,because of lack of time-series data on land conversion and lit-tle variations in output prices across counties, we are unable toinclude output prices in our land conversion model.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
984 November 2000 Amer. J. Agr. Econ.
Figure 3. Total CRP acres enrolled from 1986–92 by county
Figure 4. Acres of non-cropland converted to cropland from 1982–92 by county
the continental United States. Each NRI siteis assigned a weight (called the expansionfactor) to reflect the acreage it represents.For example, the summation of expansionfactors for all sites planted to crops in aregion gives an estimate of cultivated crop-land acreage in the region. The samplingdesign of NRI ensures that inferences at thenational, regional, state, and substate levelsare made in a statistically reliable manner.There were 336,462 NRI sites in the study
region. At each NRI site, information onnearly 200 attributes was collected, which
include land use and cover, cropping history,irrigation type, tillage and conservation prac-tices, topography, and soil type. NRI dividesland use into twelve major categories (culti-vated cropland, non-cultivated cropland, pas-tureland, rangeland, forest land, urban andbuilt up land, and six other categories). NRIrecorded land use at each NRI site in 1982and 1992. By comparing land use in 1982and 1992 at each NRI site, I estimated acresof non-cropland converted to cropland from1982 to 1992.As shown in figure 4, there werelarge variations in the number of acres con-
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
Wu Slippage Effects of the Conservation Reserve Program 985
verted across the study region. A comparisonof figure 4 with figure 3 indicates that coun-ties with more converted cropland tend tohave more CRP acres.Although the NRI data can be used to
show spatial patterns of land use change atthe county level, point estimates from theNRI (e.g., acres of non-cropland convertedto cropland) are generally reliable only ata more aggregate geographic level. For thisreason, I aggregated the NRI data to Agri-cultural Statistics Districts (ASD). Each statein our study region has nine ASDs, exceptNebraska (which has eight). On average, eachstate falls into ten Major Land ResourceAreas (MLRAs). Thus, the average size of anASD is comparable to the average size of aMLRA within a state, a level of aggregationreliable for point estimates according to theNRCS.The county-level CRP summary data from
the Economic Research Service were usedin this analysis. This data included CRPacreage by enrollment year, annual rentalrates, soil erosion reduction, corn and wheatbase acres retired by the CRP, and a numberof other variables. The percentage of crop-land enrolled into the CRP varies from zeroto over 80% among counties, with an averageof 8.6% for all counties in the study region.There were also large variations in CRPrental rates and soil erosion reduction amongcounties (see figures 5 and 6). The averageCRP rental rate varied from $12 to $113 per
Figure 5. CRP rental rates by county
acre. On average, CRP rental rates in theCorn Belt were the highest in the nation, withalmost all counties having an average annualrental rate above $70 per acre. In contrast,CRP rental rates in the western Dakotas andnorthern Lake States were the lowest, withalmost all counties having an average annualrental rate below $40 per acre. Although theaverage soil erosion reduction per CRP acrewas generally higher in the Corn Belt than inthe Dakotas, the average soil erosion reduc-tion per dollar expended was relatively lowin the Corn Belt.The 1982 and 1992 Census of Agriculture
provides information on farm characteristicsin 1982 and 1992. From these census data,several variables that may affect land usechanges were derived, including changes infarm size and average market value of landand buildings from 1982 to 1992. Some ofthese variables were not included in the finalmodel because they were statistically insignif-icant. Equation (8) was estimated using theordinary least square procedure in SAS.
Results and Implications
Three steps are involved in the estima-tion of the slippage effects of the CRP.The first one involves a statistical evalua-tion of the relationship between the CRPand the converted cropland acreage based onthe specification presented in equation (8).The relationship allows us to estimate the
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
986 November 2000 Amer. J. Agr. Econ.
Figure 6. Annual soil erosion reduction per CRP acre
acreage of non-cropland converted to crop-land as a result of land retirements under theCRP. The second step involves estimation ofenvironmental benefits offered by CRP landand the converted cropland. In the final step,we combine results from the first two stepsto determine the magnitude of the slippageeffects. The results from these three steps arereported in tables 1–3, respectively.Table 1 presents the statistical results from
the estimation of the relationship betweenthe CRP and the converted cropland acreage.As is evident from the table, all variablesexcept changes in population and averagefarm size are statistically significant at the10% level. These variables explain 45% ofthe variation in converted cropland acreageacross ASDs. Coefficients of particular inter-est to this analysis are those on CRP acreageas a proportion of the total land area andits interaction terms with regional dummyvariables. These coefficients indicate that landretirements under the CRP had a strong andhighly significant impact on the acreage ofnon-cropland converted to cropland. In theCorn Belt, each one hundred acres of crop-land retired under the CRP resulted in thirtyacres of non-cropland converted to cropland.The slippage effect is smaller in the LakeStates and Northern Plains, but still signifi-cant. Each one hundred acres of land retiredunder the CRP resulted in sixteen and fifteenacres of non-cropland converted to croplandin these two regions, respectively.
In addition to the CRP acreage, the char-acteristics of CRP land also affected the slip-page effects. The larger the corn and wheatbase acres retired under the CRP, the morelikely the farmers will replace them by con-verting non-cropland to cropland. However,both the CRP rental rate and the soil erosionreduction per CRP acre are negatively corre-lated to the converted cropland acreage. Thisresult may reflect that areas with higher CRPrental rates tend to have few acres of non-cropland suitable for crop production. Thecorrelation coefficient between CRP rentalrates and non-cropland acreage across coun-ties is −0�52. The coefficient on the change inthe average farm size indicates that the largerthe increase in the average farm size, themore likely non-cropland will be convertedto cropland. Likewise, the larger the increasein population in a county, the more likely thaturbanization is proceeding, and the less likelynon-cropland will be converted to crop pro-duction. Although these results make sense,the coefficients on both changes in farm sizeand population are statistically insignificantat the 10% level.To determine the magnitude of slippage
effects of the CRP, we also need to esti-mate the total environmental benefits offeredby CRP land and the converted cropland.Table 2 shows the use of CRP land andthe converted cropland before their retire-ments or conversions and the changes inwater and wind erosion rates on these lands
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
Wu Slippage Effects of the Conservation Reserve Program 987
Table 1. The Impact of the Conservation Reserve Program on the Acreage ofNon-Cropland Converted to Cropland from 1982–92 in the Central United States
Variablea Coefficientb T -Statistic
Intercept 0�0361∗∗∗ 6�70Acres of CRP land enrolled by 1992 as a 0�3041∗∗∗ 6�81percent of the total land area (CRP)
CRP * a dummy variable for lake states −0�1355∗∗ −2�23CRP * a dummy variable for northern plains −0�1514∗∗∗ −2�70CRP rental rate ($/acre) −0�0003∗∗∗ −3�54Average annual erosion reduction per CRP −0�0003∗ −1�86acre (tons/acre/year)
Corn base acres retired per CRP acre 0�0321∗∗∗ 2�86Wheat base acres retired per CRP acre 0�0256∗∗ 2�42Percent change in population 1982–92 −0�0119 −0�92Change in average size of farm 1982–92 0�00003 1�41Total land area (million acres) −0�0019∗∗∗ −4�02R-square 0�45
aThe dependent variable is the acreage of non-cropland converted to cropland from 1982-92 as a proportion of the total land area. Theunit of the cross-sectional analysis is Agricultural Statistics Districts (ASD) as defined by the National Agricultural Statistics Service.bOne, two and three asterisks indicate statistical significance at the 10%, 5% and 1% level, respectively.
from 1982 to 1992. These measures were esti-mated based on the comparison of water andwind erosion rates in 1982 and 1992 at eachCRP and converted cropland site in the NRIsample.Before their retirements, 19.4% of CRP
acres were planted to corn in the region,10.8% to soybeans, 27.9% to wheat, and41.9% to other crops or summer fallowin 1982. The average water and wind ero-sion rates on these lands were 6.56 and4.51 tons/acre/year, respectively. However,conservation practices under the CRP onthese lands reduced the average water andwind erosion rates to 0.52 and 0 tons/acre/year, respectively, in 1992.In contrast, before their conversions, 63.4%
of the converted cropland acres were pas-tureland, and 22.3% were rangeland in1982.7 Cropping practices on these landsincreased the average water and wind ero-sion rates for the whole region from 1.44and 0 tons/acre/year in 1982 to 4.12 and3.11 tons/acre/year in 1992. The largestwater erosion increase occurred in theCorn Belt (from 1.34 ton/acre/year in 1982to 6.99 tons/acre/year in 1992), while thelargest wind erosion increase occurred inthe Northern Plain (from zero in 1982 to4.10 tons/acre/year in 1992).
7 Since a large percentage of the converted cropland waspastureland or rangeland before the conversion, an interestingextension of this study is to look at how the CRP affected thecattle industry.
The average water erosion rate on othercropland was slightly reduced in all threeregions, with the largest reduction in the CornBelt (by 1.65 tons/acre/year). The reduc-tion may reflect the increased adoption ofconservation tillage and other conservationpractices from 1982 to 1992. For example, thepercentage of cropland cultivated with con-servation tillage was increased from 13.5%to 23% in the Corn Belt from 1982 to 1992.One important factor that contributed to theincreased adoption of conservation tillageis the reduced production costs associatedwith the use of conservation tillage (Martinet al.). In addition, the conservation com-pliance and lower commodity prices mightalso be a factor. The conservation compliancerequired adoption of conservation tillage andother conservation practices as a conditionof continued receipt of farm program pay-ments. Lower commodity prices tended tocause producers to reduce the intensity ofcrop production, primarily through reducedtillage and less intensive crop rotations.8The last three rows of table 2 show the
average land capability class of CRP landand converted and non-converted croplandas well as the percentage of each type ofland in capability classes I–II. On average, theconverted cropland has a slightly lower landcapability class (higher land quality) than theCRP land, but both have lower land qualitythan the other cropland.
8 I thank a referee for pointing this out.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
988 November 2000 Amer. J. Agr. Econ.
Table 2. Land Use, Soil Erosion Rates, and Land Quality of CRP Land and ConvertedCropland Before and After Conversions, by Region
Variables Corn Belt Lake States N. Plains The Region
Use of CRP Land in 1982 (Before Retirements) (Percent)Corn 45�5 46�0 11�5 19�4Soybeans 22�8 2�4 8�2 10�8Wheat 6�9 0�5 34�5 27�9Other 24�8 51�1 54�2 41�9
Use of Converted Cropland in 1982 (Before Conversions) (Percent)Pastureland 80�0 81�1 56�9 63�4Rangeland 0 0 30�5 22�3Forestland 8�8 5�5 3�3 4�1Other 10�4 13�4 9�3 10�2
Soil Water Erosion Rate in 1982 (Tons/Acre/Year)CRP land 12�02 6�77 5�18 6�56Converted cropland 1�34 0�42 1�56 1�44Other cropland 6�09 4�00 3�17 4�07
Changes in Soil Water Erosion Rates 1982–92 (Tons/Acre/Year)CRP land −11�42 −6�09 −4�69 −6�04Converted cropland 5�65 3�68 1�75 2�68Other cropland −1�65 −0�81 −0�67 −0�97
Soil Wind Erosion Rate in 1982 (Tons/Acre/Year)CRP land 0�56 0�05 5�72 4�51Converted cropland 0 0 0 0Other cropland 0�48 0�08 3�81 2�65
Changes in Soil Wind Erosion Rates 1982–92 (Tons/Acre/Year)CRP land −0�56 −0�05 −5�72 −4�51Converted cropland 0�51 0�17 4�10 3�11Other cropland 0 0 0 0
Average Land Capability Class (% in Land Capability Classes I–II)CRP land 3�2 3�5 3�3 3�3
(33%) (27%) (30%) (30%)Converted cropland 2�9 3�3 3�2 3�1
(49%) (38%) (38%) (40%)Other cropland 2�3 2�7 2�5 2�5
(71%) (58%) (61%) (64%)
Table 3 presents estimates of the slip-page effects of the CRP and some of thedata used in the estimation. The first rowof table 3 shows total CRP acreage enrolledin each region by 1992 based on the ERS’scounty-level CRP summary data. The reduc-tion in water and wind erosion on CRP landwas estimated by multiplying the total CRPacreage by the change in average water andwind erosion rates on CRP land as reportedin table 2.The fourth row of table 3 shows thetotal acreage of non-cropland converted tocropland from 1982 to 1992. The acreage con-verted as a result of the CRP was estimatedusing the estimated relationship between theCRP and the converted cropland acreageas reported in table 1. Specifically, I first
estimated the acreage of non-cropland thatwould be converted without the CRP byassuming a zero CRP acreage in the rela-tionship and then subtracted it from the totalconverted acreage. The result was then mul-tiplied by the change in average water andwind erosion rates on the converted crop-land to give an estimate of the increase intotal water and wind erosion on the land con-verted by the CRP. The percentage of CRPsoil erosion benefit offset by the slippageeffect was then estimated based on the reduc-tion in soil erosion on CRP land and theincrease in soil erosion on the cropland con-verted by the CRP.The results in table 3 show that a slip-
page effect was present in the CRP. For the
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
Wu Slippage Effects of the Conservation Reserve Program 989
Table 3. Slippage Effects of the CRP on Cropland Acreage and Soil Erosion Benefits
Variables Corn Belt Lake States N. Plains The Region
Total acres of cropland 5,204,923 2,847,611 9,576,646 17,629,180enrolled in CRP by 1992a
Reduction in soil water erosion on −59� 440� 221 −17� 341� 951 −44� 914� 470 −106� 480� 247CRP land (tons/year)b
Reduction in soil wind erosion on −2� 914� 757 −142� 381 −54� 778� 415 −79� 507� 602CRP land (tons/year)b
Total acres of non-cropland 2,117,100 544,000 7,493,600 10,254,700converted to cropland 1982–92
Acres of non-cropland converted 1,582,817 480,107 1,462,354 3,525,278to cropland by the CRP
Increase in water erosion on converted 8,942,917 1,766,795 2,559,119 9,447,745cropland by the CRP (tons/year)
Increase in soil erosion on converted 807,237 81,618 5,995,651 10,963,615cropland by the CRP (tons/year)
Acres of non-cropland converted 0.30 0.17 0.15 0.20for each CRP acre
Percent of soil water erosion benefit 15% 10% 6% 9%offset by the slippage effect
Percent of soil wind erosion benefit 28% 57% 11% 14%offset by the slippage effect
aEstimated based on the Economic Research Service’s county-level CRP summary data.bThe water and wind erosion reductions are estimated by multiplying total CRP acreage by the average water and wind erosion reduction rates as shownin table 2. The sum of the water and wind erosion reductions is lower than the estimate of soil erosion reduction based on ERS’s county-level CRPsummary data. The estimates of water and wind erosion reduction rates are used here for consistency because ERS’s county-level CRP summary data donot separate water and wind erosion and include no information on soil erosion on the converted cropland.
entire study region, each one hundred acresof land retired under the CRP resulted intwenty acres of non-cropland converted tocropland. This slippage effect offset 9% ofwater erosion benefits and 14% wind ero-sion benefits offered by the CRP. The slip-page effect on acreage was the largest in theCorn Belt; each one hundred acres of landretired under the CRP resulted in thirty acresof non-cropland converted in the region. Thisslippage offset 15% of CRP water erosionbenefits and 28% of CRP wind erosion bene-fits. The slippage effect on acreage was small-est in the Northern Plains; each one hundredacres of land retired under the CRP resultedin fifteen acres of non-cropland converted inthe region, offsetting 6% of CRP water ero-sion benefits and 11% of CRP wind erosionbenefits. The slippage effect on water erosionbenefits was smaller than the slippage effecton acreage in all three regions because theincrease in the average water erosion rate onthe converted cropland was smaller than thereduction in the average water erosion rateon the CRP land. A large portion of CRPwind erosion benefit (57%) was offset by theslippage effect in the Lake States becausethe increase in the wind erosion rate on theconverted cropland was much larger than the
decrease in the wind erosion rate on the CRPland. Fortunately, the average wind erosionrate was minimal on all types of land in theLake States.
Concluding Remarks
Each year, billions of dollars of public andprivate funds are expended to purchaseeasements on farmland for resource conser-vation and environmental protection. Oneunintended impact of these conservation pro-grams is that they may bring non-croplandinto crop production. Such a slippage effectcan be caused by increased output pricesand/or by substitution effects. In this study,we found that substantial slippage effectswere present in the CRP. For each one hun-dred acres of land retired under the CRPin the central United States, twenty acresof non-cropland were converted to cropland,offsetting 9% and 14% of CRP water andwind erosion benefits, respectively.The USDA estimated that the CRP pro-
vided $9.7–14.5 billion of net social benefits(Osborn and Ribaudo, p. 294). The results ofthe present analysis suggest that these ben-efit estimates may be over stated because
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
990 November 2000 Amer. J. Agr. Econ.
the slippage effects were not incorporatedinto the estimation. For the twelve states inthe central United States, ignoring slippageeffects would result in an overestimation ofacreage control benefits by 20% and an over-estimation of water and wind erosion relatedbenefits by 9% and 14%. The overestimationof other environmental benefits may be evenlarger if the converted cropland, prior to con-version, provides the same or higher level ofthese benefits per acre than the CRP land.The results of this analysis have impli-
cations for the design and implementationof land retirement programs. The substan-tial slippage effect identified in this analy-sis suggests that it should be considered inthe targeting of conservation efforts. Cur-rently, the CRP targets only cultivated crop-land. Non-cropland is ineligible for the CRP.Relaxing program rules to allow targeting ofnon-cropland that offers large environmen-tal benefits but would otherwise be convertedinto crop production would reduce the slip-page effects and improve the environmentalperformance of conservation programs (Wu,Zilberman and Babcock). Although enrollingnon-cropland would not provide any supplycontrol benefit, it would be consistent withthe objective of maximizing environmentalbenefits per dollar expended, as specifiedfor the Environmental Quality IncentivesProgram (EQIP) and other conservationprograms.A good part of the rationale for land retire-
ment as a national policy since the 1930s isthat it achieves both conservation/environ-mental objectives and supply control/farmincome support objectives. But if conver-sions of non-cropland to cropland as a resultof land retirement are not recognized andprevented, the effect of land retirement onsupply control is reduced, and the objec-tive of maintaining commodity prices, farmincome and asset values would also becompromised. Thus, reducing slippage effectsnot only improves the environmental per-formance of a land retirement program, itincreases supply control although supply con-trol has not been universally agreed upon tobe beneficial from a societal standpoint.The advent of the 1985 Food Security
Act signaled a significant change in theUSDA’s agricultural and environmental pol-icy (Johnson, Wolcott, and Aradhyula). Thesodbuster and swampbuster provisions wereincluded in this and all subsequent farm billsto restrict conversion of wetland and highly
erodible land. The incarnation of the CRP inthe farm bill represents a significant improve-ment over previous land retirement programsby limiting enrollment to highly erodibleland. The evidence presented in this article,however, suggests that both the environmen-tal and supply control benefits of the CRPcould be increased if the slippage effects wereexplicitly recognized and controlled. We rec-ognize the conceptual and practical limita-tions on accounting for the secondary effectof land retirement and the considerable dif-ficulties imposed by budget constraints andinstitutional strictures. However, we are opti-mistic that the environmental and economicperformance of conservation programs willcontinue to improve with the increasing avail-ability of data and economic analysis and thegrowing public demand for more accountablegovernment policies.This cross-sectional analysis is more likely
to uncover the substitution effects causingslippage than the price effects and thus mayunderestimate the slippage effects. An impor-tant extension of this study is to determinethe magnitude of the price-related slippage.If substitution effects are the main cause,then focusing on participating farmers shouldbe sufficient to prevent the slippage effect.On the other hand, if output price increasesare the main cause of the slippage effect,focusing on participating farmers alone willnot solve the slippage problem. Because cropprices are determined in national and inter-national markets, the CRP would affect cropprices by similar propositions across the studyregion. Thus, time-series data on CRP par-ticipation and land use changes are neededto identify the price-related slippage. Finally,this study addresses only one issue asso-ciated with the optimal design of conser-vation programs. Many other issues havebeen identified in the literature, including thechoice of targeting criteria, the importanceof offsite benefits, problems of asymmetricinformation, interrelations between alterna-tive environmental benefits, and cumulativeeffects. Optimal design of conservation pro-grams must consider these issues as well.In addition, U.S. conservation policy is ata crossroads between more coercive regula-tory policies, more costly voluntary programs,and more facilitative market-oriented poli-cies (Heimlich and Claassen). More researchis needed to identify the pitfalls, advantages,and tradeoffs along these different paths.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
Wu Slippage Effects of the Conservation Reserve Program 991
Appendix: Derivation of Equation (5)
Let f1(y) and f2(b) denote the marginal densityfunctions of y and b. Under the assumption that(y� b) follows the bivariate normal distributionN(µy� µb� σy� σb� ρ),
f1(y) = 1σyϕ
(y − µy
σy
)�
f2(b) = 1σbϕ
(b − µb
σb
)�
where ϕ(·) is the probability density function ofthe standard normal distribution. Similarly, thecumulative distribution function of y and b are
F1(y) ≡ +
(y − µy
σy
)�
F2(b) ≡ +
(b − µb
σb
)�
where +(·) is the cumulative distribution functionof the standard normal distribution.The changes in total acreage of cropland, envi-
ronmental benefits and total output equal
A =∫ y∗
c/p0
∫ +∞
−∞s(y� b)dydb(A1)
−∫ c/p0
c/p1
∫ +∞
−∞s(y� b)dydb
=∫ y∗
c/p0
f1(y)dy −∫ c/p0
c/p1
f2(y)dy
= [F1(y∗)− F1(c/p0)]
− [F1(c/p0)− F1(c/p1)]
= [F1(y∗)+ F1(c/p1)− 2F1(c/p0)]�
B =∫ y∗
c/p0
∫ +∞
−∞bs(y� b)dydb(A2)
−∫ c/p0
c/p1
∫ +∞
−∞bs(y� b)dydb
=∫ y∗
c/p0
f1(y)∫ +∞
−∞bs(y� b)
f1(y)dydb
−∫ c/p0
c/p1
f1(y)∫ +∞
−∞bs(y� b)
f1(y)dydb
=∫ y∗
c/p0
f1(y)
[µb + ρ
(σbσy
)
× (y − µy)
]dy
−∫ c/p0
c/p1
f1(y)
[µb + ρ
(σbσy
)
× (y − µy)
]dy
= {µb[F1(y
∗)− F1(c/p0)]
− ρσb[f1(y∗)− f1(c/p0)]
}− {
µb[F1(c/p0)− F1(c/p1)]
− ρσb[f1(c/p0)− f1(c/p1)]}
= µb A− ρσb[f1(y∗)
+ f1(c/p1)− 2f1(c/p0)]�
Y =∫ c/p0
c/p1
∫ +∞
−∞ys(y� b)dydb(A3)
−∫ y∗
c/p0
∫ +∞
−∞ys(y� b)dydb
=∫ c/p0
c/p1
yf1(y)dy −∫ y∗
c/p0
yf1(y)dy
= {µy [F1(c/p0)− F1(c/p1)]
− σy [f1(c/p0)− f1(c/p1)]}
− {µy [F1(y
∗)− F1(c/p0)]
− σy [f1(y∗)− f1(c/p0)]
}= −µy A+ σy [f1(y
∗)
+ f1(c/p1)− 2f1(c/p0)]�
Multiplying (A2) by σy and (A3) by ρσb and thenadding them together gives equation (5).
[Received July 1999;accepted April 2000.]
References
Babcock, B.A., P.G. Lakshminarayan, J. Wu, andD. Zilberman. “The Economics of a Pub-lic Fund for Environmental Amenities: AStudy of CRP Contracts.” Amer. J. Agr. Econ.78(November 1996):961–71.. “Targeting Tools for the Purchase of Envi-ronmental Amenities.” Land Econ. 73(August1997)325–39.
Heimlich, R.E. Sodbusting: Land Use Change andFarm Programs.Washington DC: U.S. Depart-ment of Agriculture, Economic Research Ser-vice, AER–536, June 1985.
Heimlich, R.E., and C.T. Osborn. “After the Con-servation Reserve Program: Macroeconomicsand Post-contract Program Design.” In Pro-ceedings of Great Plains Agricultural CouncilMeeting, p. 113–33, Rapid City SD June 2–4,1993.
Heimlich, R.E. “Productivity of Highly Erodi-ble Cropland.” J. Ag. Econ. Res. 41(Summer1989):17–22.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from
992 November 2000 Amer. J. Agr. Econ.
Heimlich, R.E., and R. Claassen. “AgriculturalConservation Policy at a Crossroads.” Agr.Res. Econ. Rev. 27(April 1998):95–107.
Johnson, S.R., R. Wolcott, and S.V. Aradhyula.“Coordinating Agricultural and Environmen-tal Policies: Opportunities and Tradeoffs.”Amer. Econ. Review 80(May 1990):203–7.
Martin, M.A., M.M. Schreiber, J.R. Riepe, andJ.R. Bahr. “The Economics of Alterna-tive Tillage Systems, Crop Rotations andHerbicide Use on Three RepresentativeEast-Central Corn Belt Farms.” Weed Sci.39(April/June 1991):299–307.
Miranowski, J., and M. Cochran. “Economics ofLand in Agriculture.” Agricultural and Envi-ronmental Resource Economics. G.A. Carlson,D. Zilberman, and J.A. Miranowski eds.,pp. 392–440. New York: Oxford UniversityPress, 1993.
Osborn, C.T. “The Conservation Reserve Program:Status, Future, and Policy Options.” J. Soiland Water Conserv. 48(July/August 1993):271–78.
Osborn, C.T., and M. Ribaudo. “ConservationReserve Program.” Agricultural Resourcesand Environmental Indicators, 1996–97.
U.S. Department of Agriculture, EconomicResearch Service, Natural Resources andEnvironment Division. Agricultural Hand-book No. 712. pp. 286–96. Washington DC:1998.
Reicheldefer, K., and W.G. Boggess. “Govern-ment Decision Making and Program Perfor-mance:The Case of the Conservation ReserveProgram.” Amer. J. Agr. Econ. 70(February1988):1–11.
Ribaudo, Marc O. “Consideration of OffsiteImpacts in Targeting Soil Conservation Pro-grams.” Land Econ. 62(November 1986):402–11.
Thurman, W.N. Assessing the EnvironmentalImpact of Farm Policies. Washington DC: TheAEI Press, 1995.
Wu, J., D. Zilberman, and B.A. Babcock. “Envi-ronmental and Distributional Impacts of Con-servational Targeting Strategies.” J. Environ.Econ. Manage., in press.
Young, C.E., and C.T. Osborn. The ConservationReserve Program: An Economic Assessment.Washington DC: U.S. Department of Agricul-ture, Economic Research Service, AER–626,February 1990.
at University of Iow
a Libraries/Serials A
cquisitions on Novem
ber 13, 2014http://ajae.oxfordjournals.org/
Dow
nloaded from