E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ l oca te /eco l econ

ANALYSIS

Valuing future development rights: The costs ofconservation easements

Kathryn Andersona, Diana Weinholdb

aUniversity of Wisconsin, 1137 Elizabeth Street, Madison, WI 53703, USAbLondon School of Economics, Houghton Street London WC2A 2AE, UK

A R T I C L E I N F O

E-mail addresses: anderson.kathryn@gmai

0921-8009/$ – see front matter © 2008 Elsevidoi:10.1016/j.ecolecon.2008.04.015

A B S T R A C T

Article history:Received 20 July 2007Received in revised form9 January 2008Accepted 22 April 2008Available online 12 June 2008

While theory strongly suggests that restricting development rights should reduce landprices, empirical evidence of this effect has been hard to obtain. Largely based on thisdifficulty, legislation governing the tax benefits that easement donors may receive isperiodically challenged. We collect data on 131 land transactions in South CentralWisconsin, including 19 cases of development-restricted parcels. When we use the wholesample to estimate the impact of conservation easements, we replicate the results ofNickerson and Lynch [Nickerson, C.J., Lynch, 2001. The effect of farmland preservationprograms on farmland prices. American Journal of Agricultural Economics 83(2):341–351],finding a negative but statistically insignificant effect. However we then show that when thesample is appropriately restricted to amore homogenous group of land parcels, our ability todetect an effect increases dramatically. In particular, for vacant agricultural land we find astatistically significant negative impact of conservation easements that ranges up to 50% ofland values.

© 2008 Elsevier B.V. All rights reserved.

Keywords:Land useLand priceConservation easements

JEL classification:Q24; Q51; R52

1. Introduction

The contribution of future development potential to currentland prices is well understood theoretically while being quitedifficult to measure empirically. Although such informationcould have great value for public policy, the value of develop-ment rights is not directly observable, remaining roughly‘guesstimated’ by appraisers.

One area where this question is directly applicable is inthe pricing of land encumbered with conservation ease-ments. A conservation easement is a legal contract thatprevents current and future landowners from engaging incertain activities on their land and/or compels the land-owner to maintain certain attributes of the land. The

l.com (K. Anderson), d.we

er B.V. All rights reserved

conservator (government agency or land trust) is respon-sible for monitoring and enforcement of the easementcontract for perpetuity. Easements are often referred to as“partial interests” because they do not transfer the full titleto the property, only the right to enforce restrictions ondevelopment.

Conservation easements are often used as a tool toconserve scenic landscapes, wildlife habitat, places to recre-ate, and biodiversity. Among the common justifications foreasement programs are arguments that conversion of unde-veloped land is the primary cause of the habitat loss re-sponsible for the status of most of the species listed under theEndangered Species Act (Boyd et al., 1999), that undevelopedland enables natural water purification, flood control, and

inhold@lse.ac.uk (D. Weinhold).

.

438 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

groundwater recharge, and that the public enjoys the ruralcharacter of farmland.1

Since a conservation easement restricts the menu of pos-sible alternative uses for the affected land, simple economictheory would suggest that the value of the encumbered landshould fall. As future development potential is not directlyobservable, the degree to which the land value falls must beestimated, and this estimate of the value of the easement inturn determines any cash transfers or subsequent tax benefitsto the landowners.

While basic economic theory strongly suggests conserva-tion easements should reduce land values, there is surpris-ingly little empirical evidence to support this claim. Hedonicland price models have yielded mixed results, finding bothpositive and negative price effects that are mostly statisticallyinsignificant (one notable exception is Lynch et al., 2007).Furthermore, there are some cases in which theory suggestsan easement could in fact increase land values. For example, ifa development restriction raises the land values of nearbyparcels (and there is empirical evidence to suggest that thismay often be the case — see Irwin and Bockstael, 2001,Geoghegan et al., 2003; Geoghegan, 2002), either throughincreased amenity value or because prices are bid up by thedecrease in supply of developable land, this general neighbor-hood increase in value could spill back over to the home parcel(see Plantinga, 2007 for a formal theoretical exposition of theseeffects).

The lack of robust empirical evidence on systematic priceeffects combined with anecdotal accounts of non-effects inthe field have left many wondering whether conservationeasements do in fact reduce land values after all. This doubtwas brought into the public eye in February, 2005 when theJoint Committee on Taxation (JCT) of the U.S. Congressrecommended severely limiting the deductions landownerscan take for donating a conservation easement. In theirdiscussion of the issues, the JCT notes,

“Valuation is especially problematic because the measureof the deduction (i.e. generally, the difference in fairmarketvalue before and after placing the restriction on theproperty) is highly speculative, considering that, in general,there is no market and thus no comparable sales data forsuch easements.” (Joint Committee on Taxation, p. 281)

After the release of the initial JCT proposals, conservationgroups around the country rallied to oppose them. TheUmbrella conservation group Land Trust Alliance campaignedto protest the possible changes, claiming they would “destroy”the good work of the nonprofits (Stephens, 2005).

Political action was effective in this case, and in August of2006, Congress actually expanded, rather than contracted, thefederal conservation tax incentives for conservation easement

1 Surveys, referendums, and contingent valuation studies haveshown that farmland and open space preservation has wide-spread support among the public, especially programs that targetenvironmental resources, wildlife habitat, natural places, farm-land, and rural character (see, for example, Pruckner, 1995; Drake,1992; Beasly et al., 1986; Bergstrom et al., 1985; Halstead, 1984;Foster et al., 1982).

donations. In particular, the change (which was temporary,but as of December 2007 was passed as permanent by thesenate and will come up for approval by the full congress inJanuary of 2008) increased the amount of deduction thatmay be taken for charitable contributions of qualified con-servation easements from 30% to 50% of income, among othermeasures.

The purpose of this paper is to revisit the question ofpricing development rights using data on land prices fromSouth Central Wisconsin. Although an active market indevelopment rights per se does not yet exist, the growingpopularity of conservation easement programs in this regionhas resulted in an increasing number of sales of land withrestricted development potential. By observing competitiveprices and parcel characteristics from land sales of a sufficientnumber of parcels with and without restrictions on develop-ment, we use regression analysis to tease out any effect onmarket values that an easement confers.

While a few studies have attempted to estimate hedonicvalues of conservation easements with other data, very fewhave found statistically significant effects. This could be be-cause in fact there is very little effect on prices, however, ourprimary empirical contribution is to show that this is likely notthe case in general. Instead, the value of future developmentrightsmay vary dramatically with the type of land encumberedwith an easement. In particular, we argue that for improvedland, especially parcels with existing residences, the effects ofa conservation easement may be negligible, or at least verydifficult to measure. Residential plots selected into easementprograms may have greater than average unmeasured envir-onmental amenities that will increase market values, mitigat-ing any price effect of the easement in a hedonic regression.On the other hand, unimproved parcels with conservationeasements face a much more limited market and mayexperience a much greater decline in value. We show thatwhen we estimate the reduction in land value for our wholesample we replicate the results in most of the literature,finding a negative but statistically insignificant effect. How-ever when we restrict our sample to unimproved land, andespecially unimproved agricultural land, we find a muchlarger, robust and significant negative impact of up to 50% ofland values.

This paper proceeds as follows. In Section 2 we describeand discuss the importance of conservation easement pro-grams nationally and in Wisconsin, and briefly review thelimited existing literature on measuring the value of develop-ment rights. Section 3 describes the data. The estimationmethodology and results are presented in Section 4, withSection 5 concluding.

2. Conservation easement programs and thevalue of development rights

Rural lands were converted to rural residential and urban landat a rate of 2.1 million acres per year from 1992–97, a signifi-cant increase from the 1950–92 period (Vesterby and Krupa,2001). The associated rise in public concern about landconversion has led to the introduction of a menu of landconservation schemes designed to protect open spaces from

439E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

development, including conservation easement programs. By2005, over 6.2 million acres had been preserved by conserva-tion easements and the number of land trusts had grown toover 1600 (Land Trust Alliance, 2006). As of 2001, nineteenstate-level and 34 county-level public Purchase of Develop-ment Rights (PDR) programs in eleven states had preserved819,490 ac of farmland at a cost of $1.2 billion (Heimlich andAnderson, 2001).

Landowners benefit from direct payments for the develop-ment rights and/or tax savings, as well as from any personalsatisfaction from knowing their land will be protected in thefuture. Although the only tax benefit received by a landownerwho sells a conservation easement is the direct effect of thereduction in the recognized land value, for those donatingeasements, several additional tax benefits are available. Theseinclude charitable donation tax deductions, and estate taxdeductions. Crucially, the value of all of these benefits de-pends on the value assigned to the donated developmentrights. Some states automatically reduce the assessed landvalue by as much as 75%, but in many cases the value of aneasement must be estimated by local assessors.

While the magnitude of the price effect may be underdebate, at first glance most economists would predict that thesign of the change should be an unambiguous fall in themarket price of restricted parcels. The price of an unrestrictedproperty reflects the current and future stream of income andthe monetized value of utility from the current use (e.g.,agriculture or bird watching), the expected future stream ofincome from some alternative use (e.g., residential subdivi-sions), and the value of the option to convert the land to analternative use at a future date when more information onconditions is available. Thus, as long as the expected value of a(now restricted) alternative use is greater than the expectedvalue of current use, or the conversion option value is positive,the effect of an easement should be to reduce market price.2

A simple model illustrates the concept. In each period thata property remains undeveloped (and unrestricted), land-owners decide whether to convert their land or maintaincurrent use. Landowners make the conversion decision thatmaximizes the price of their land (the decision to convert isassumed to be irreversible).

The price of unrestricted land =

P 0; x0ð Þ ¼ max ID x0ð Þ þ dE P 1; x1ð Þ½ �� �; IU x0ð Þ þ dE P 0; x1ð Þ½ �� �� �

ð1Þ

where P is a function of development status (0 = undeveloped;1 = developed) and parcel characteristics; x0 is the vector ofparcel characteristics in the current period; x1 is a vector ofparcel characteristics in the next period, as predicted by astate equation x1 = G(x0, ε), in which ε is a vector of randomvariables; ID is development income (or money value of utility)as a function of parcel characteristics in the current period; δ isthe discount rate 1

1þr

� �, and IU is income (or monetized utility)

from current undeveloped use. This is a recursive formula that

2 The market price of a conservation easement, would there tobe a complete market, would reflect not only this change in thenet present value of returns to the land owner, but also the valueto the general public (as represented by the land trusts), ofconserving the land.

reflects the opportunity to develop in the future when moreinformation is available.

The price of restricted land ¼PR x0ð Þ ¼ EXlt¼0

dtIU xtð Þ( )

; ð2Þ

where PR is different from P.

Thus; the price of a conservation easement ¼ Peas¼P 0; x0ð Þ�PR x0ð Þ:

ð3ÞCorresponding to our initial economic intuition, under

these simple assumptions, with no spatial spillover effects,the price of restricted land will always be less than or equal tothe price of unrestricted land. However, the magnitude of thisreduction could vary considerably with the type of land underconsideration. Many income or utility-generating amenitiesfrom land rely on some kind of improvement such as a resi-dence or shelter of some sort. The range of possible usesbesides wholesale development remains very large withimproved land; a shelter facilitates greatly the enjoyment ofenvironmental amenities that substitute for money income.Thus the difference Peas = P(0,x0) − PR(x0) may be relativelylimited for land with improvements and substantial naturalamenities (exactly those that are most likely to be selectedinto conservation easement programs). On the other hand,vacant parcels which cannot be improved have a much morelimited range of possible uses. Vacant agricultural land inparticular can be cultivated economically only by neighboringfarms, limiting future values to agricultural income.

In addition, there are circumstances in which we mightobserve thepriceof restricted landbeinggreater thanunrestrictedland. For example such a pattern could be caused either byselection bias or spatial spillover effects. In the former case, thedecision to conserve a parcelmay be correlatedwith unobservedfactors that are simultaneously causing the price to increase.This could be the case for an areawhere land trusts are targetingparcelswithhighamenityvalueand/ordevelopmentpressure. Inthe second case of spatial spillovers, several studies have foundthat conservationeasementshave thepotential to raise thevalueof nearby properties (Geoghegan et al., 2003; Farja and Rausser,2007). If the change in neighborhood property values causesother characteristics to change (e.g.,more affluent residentsmayhave nicer gardens), thismay result in an increase in the price ofthe restricted land — a spill-over-and-back-again effect.

Although theory does thus suggest the possible existenceof mechanisms through which conservation easements mayhave a negligible or even positive impact on land prices, thepreponderance of theory leads most economists' priors toremain that the effect should be significant and negative (see,for example, Plantinga and Miller, 2001). As theory is am-biguous as to both the order of magnitude and the sign of anyprice effect of easements, the question must be determinedempirically. However, conservation easements are not boughtand sold in an open market and thus do not have observableprices.3 Furthermore, appraisers determining the value of a

3 Although some easements are sold, it is not in a competitiveenvironment; demand derives primarily from government agenciesthatact asa singlebuyer inanyonemarketandmayhaveasubjectiveand inaccurate conception of the easements' value (Boyd et al 1999).

Table 1 – Continuous variables

Variable Definition No. Obs. Mean Std. Dev. Min Max

Easement=0sale_price Sale price 112 157,675.6 122,785.6 5,000 600,000Ppa Price per acre 112 5,032.166 3,796.439 400 14,751Acres Acres 112 36.09821 25.56707 10 154Improvement Value of improvements 112 51,033.49 72,856.81 .01 317,850Road_den Road density 112 2.64039 .8062855 1.2593 5.2165Dist Distance to Madison 112 28.25607 10.65498 3.383239 45.32708

Easement=1Sale_price Sale_price 19 204,273.7 115,727.4 78,000 485,000Ppa Price per acre 19 3,124.901 3,094.605 629.0323 12,034.74Acres Acres 19 100.0474 71.80127 37 357Improvement Value of improvements 19 25,270.53 52043.89 .01 158,340Road_den Road density 19 2.308711 .7180318 1.2689 3.4244Dist Distance to Madison 19 21.62735 13.98221 3.912329 37.42102

440 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

parcel's development rights explicitly factor in past valuationsof “similar” land, and an inaccurate conception of these valuesis thus self-perpetuating. For example, Florida appraisers fac-tor in easement values of similar parcels located in Pennsyl-vania, even though the determination of these valuesthemselves had no empirical basis (Boyd et al., 1999).

Any sort of systematic data on sales of protected land isvery difficult to collect. Land trusts and conservation agenciesare scatteredwith various quality of record keeping on sales ofeased parcels. Land registries often do not include easementinformation in the easily accessible, digital records of landsales. Furthermore, due to self-selection, landowners witheasements on their propertymay be less inclined than averageto sell and the sale is less likely to be “arms-length” (e.g. wherethere is no relationship between buyer and seller). Finally,since many conservation easement programs have only beenin existence for less than two decades, very few restrictedproperties have come on the market.4

Despite these difficulties, Nickerson and Lynch (2001) col-lected a sample of 244 land transactions, including 24 re-stricted properties and data on many plot characteristics.They tested the effect of development restrictions and foundthat development restrictions lowered sales price, but notstatistically significantly. Lynch et al. (2007) re-examined thequestion using a substantially expanded data set of 3554observations in 22Maryland counties, including 249 preservedproperties. Their hedonic method does find a statistically sig-nificant reduction in parcel price due to preservation, rangingfrom 11% to 17%. However, using a propensity score method,which matches preserved and control parcels in part by thedistance between them, yields only a small and statisticallyinsignificant effect, suggesting that perhaps landowners onlyenroll in preservation programs in areas of low developmentpressure and low land values.

Plantinga et al. (2002) suggest an innovative alternativeestimation strategy to identify the value of developmentrights indirectly from county-level data. They find that the

4 A 2001 survey by Gathering Waters, Wisconsin's umbrella landtrust, found that 97% of restricted parcels are still owned by theeasement grantor (Gathering Waters Conservancy, 2001).

share of total land value due to development potentialincreases with highway density, proximity to urban centers,and the rate of change in population density, and decreaseswith the amount of county land base that is undeveloped. Inparticular, they predict that for the contiguous U.S. as awhole, the present value of future development on agricul-tural land represents about 10% of the total value of agri-cultural land, whereas in counties near urban centers, futuredevelopment potential often accounts for over 50% ofagricultural land value. While intuitive, the results are oflimited value to local planners assessing individual parcelswithin a county.

Finally, two as yet unpublished Masters Theses and anunpublished staff paper have attempted to estimate the valueof conservation easements directly from land purchase data.Blakely (1991) finds that sale prices of agricultural parcelsenrolled in a Washington PDR program were significantlylower on average than unpreserved farmland (with a meandifference of $1217, but controlling only for improvements).Zhang (2004) uses a limited data set of 85 land transactions(including 6 restricted parcels) and finds that easements had apositive but insignificant effect on sales price in HowardCounty, MD. With a sample of 34 long-term easement en-cumbered land transactions, Taff (2004) finds a statisticallysignificant negative effect of 32%, but finds that the range ofthe easement effect is so large as to be untrustworthy.

Thus, economic theory is somewhat ambiguous about theimpact of conservation easements on land values, and em-pirical work to date has not been definitive in determiningeither the sign or the order ofmagnitude of any effect. Becauseof its great policy importance and this theoretical and em-pirical ambiguity, the extent to which restricting developmentrights changes land values continues to be the focus oflegislative debate. In the following sections we examine datafrom South Central Wisconsin to show that the lack of anyrobust empirical evidence may be due to the failure of pre-vious research to differentiate between vacant and improvedland. When we pool the sample of parcels we replicate thepreponderance of the published results to date, finding anegative but statistically insignificant effect of conservationeasements on land prices. However when we restrict thesample to vacant land only, and especially vacant agricultural

Table 2 – Qualitative (dummy) variables

Variable Definition Easement=1 Easement=0

Count Share Count Share

Col Columbia County 8 0.42 48 0.43dane Dane County 9 0.47 32 0.29Jeff Jefferson County 2 0.11 32 0.29year1 1999 and 2000 6 0.32 31 0.28year2 2001 and 2002 10 0.53 66 0.59year3 2003 and 2004 3 0.16 15 0.13City City 0 0.00 2 0.02Town Township 19 1.00 109 0.97Village Village 0 0.00 1 0.01Zone3 Minimum of 1–5 ac 1 0.05 29 0.26Zone4 Minimum of 8–16 ac 0 0.00 8 0.07Zone5 Minimum of 35 ac 17 0.89 63 0.56Zone_oth Other zoning 1 0.05 12 0.11Vacant No buildings on parcel 15 0.79 57 0.51Agric Primarily agricultural use 15 0.79 64 0.57

441E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

land, we are able to detect an economically and statisticallysignificant negative impact on prices despite a limited numberof observations and control variables.

5 Note that ln(ppa)= ln(price/acre)=ln(price)− ln(acre) which isequivalent to a regression of ln(price)=a+b ln(acre) in which thereis a restriction that b=1.6 In an earlier version of this paper we used level Price as the

dependent variable, with broadly similar qualitative and quanti-tative results. Using Price without logs imposes the constraintthat the effect of an easement on price per acre will decline as thenumber of acres increases, which is clearly an undesirableassumption.

3. Data

Southern Wisconsin provides a promising arena in which toassess whether conservation easements affect market landprices. According to the American Farmland Trust, DaneCounty is in the third most threatened farm area in thecountry (Sorenson et al., 1997) with about 5000 ac of farmlandbeing converted each year. South CentralWisconsin also has aproactive Department of Natural Resources (DNR) and avibrant land trust community, both of which have conservedland with conservation easements. The USDA NaturalResource Conservation Service has also bought many con-servation easements on Wisconsin wetlands.

The final cleaned data set includes 19 restricted propertiesand 131 unrestricted properties with complete data, sold be-tween 1999 and 2004 in three contiguous Wisconsin countiesnear the capital city of Madison — Dane, Jefferson, andColumbia. The detailed information on easement characteris-tics comes from files in the DNR real estate office, from therecords of local land trusts, and from county courthouse in-teractive databases. All 19 easements specified non-develop-ment, agricultural preservation, or wetland reserve restrictions.Control parcels come froma real estate transfer return databasemaintained by the Wisconsin Department of Revenue (DOR).

The sample was restricted to a small, geographicallyhomogenous area and to land sales that were confirmed full,“arm's length” (no relationship between buyer and seller),market transactions of greater than 10 ac. Control parcelswerelimited to those within a nine-section area around a restrictedparcel. Outlier parcels of unusually small acreage or unusuallyhigh price-per-acre value, waterfront property, parcels witheasements that covered only a small portion of the land,and parcels with atypical easements, such as hunting ease-ments and easements with unique constraints, were alleliminated.

The dependent variable is captured in the market saleprice, but the literature on hedonic price functions provides noguidance in choosing between sale price (price), log of saleprice (ln(price)), price per acre (ppa), or ln(ppa). Nickerson andLynch (2001) use ln(ppa), Geoghegan et al. (2003) use ln(price),and Plantinga et al. (2002) use ppa. Using level sales pricerather than price per acre as the dependent variable has theadvantage that it is intuitive and does not impose any fixedlinear restriction on the relationship between price andacreage.5 The disadvantage of using level price as thedependent variable is that a conservation easement wouldbe expected to have an effect on price per acre or perhaps as apercentage discount on the price, but likely not as a singleintercept-shift, suggesting that using ln(price) may yield themost intuitively interpretable specification. Thus we adopt ln(price) as our dependent variable to test for sign/significanceof a conservation easement effect.6

Economic theory and the vast literature on hedonic landprice models (as well as practical limitations) informed thechoice of RHS control variables collected. Tables 1 and 2 pro-vide detailed descriptions and summary statistics for allvariables. The main variable of interest, easement, is whethera parcel was sold encumbered with a development restriction.Easement = 1 if there is a restriction, 0 otherwise. Other controlcharacteristics of each parcel include parcel size, assessedvalue of improvements (house, barn, etc.), distance toMadison, road density nearby, whether the parcel is vacant,whether it is predominantly used for agriculture, countydummies, zoning dummies, village, town or city location,and year of sale. From Tables 1 and 2 we observe that many of

Table 3 – Dependent variable=Ln(Price)

Full sampleΔ

1 2 3 4

OLS OLS 2S treatmenta OLS

Easement −0.0569 (−0.37) −0.0513 (−0.34) 0.0289 (0.09)Hard easement − .172 (−1.27)ln_acres 0.743⁎⁎⁎ (7.25) 0.739⁎⁎⁎ (8.03) 0.724 ⁎⁎⁎ (6.88) .754⁎⁎⁎ (7.67)ln_improv 0.0237 (1.56) 0.0237 (1.61) 0.0235 (1.57) .0236 (1.55)Ln_dist −0.212 (−1.24) −0.212⁎⁎ (−2.01) −0.21 (−1.22) − .207 (−1.23)Road_den 0.0957 (1.56) 0.0964⁎ (1.68) 0.0957(1.38) .0928 (1.52)Jeff 0.169 (0.61) 0.176 (1.11) 0.176 (0.47) .188 (0.70)Col 0.0077 (0.04) 0.0128 (0.06) − .00444 (−0.02)Year2 −0.0508 (−0.40) −0.0441 (−0.36) − .0462 (−0.36)Year3 −0.047 (−0.24) −0.0408 (−0.24) − .0277 (−0.14)City 0.0743 (0.46) 0.0703 (0.16) .0777 (0.49)Village 0.769⁎⁎⁎ (3.28) 0.793⁎⁎⁎ (4.33) 0.77 (1.25) .764⁎⁎⁎ (3.30)Zone3 0.0138 (0.06) 0.00529 (0.02) − .0264 (−0.11)Zone4 0.517⁎ (1.83) 0.491⁎ (1.89) 0.503 ⁎ (1.75) .505⁎ (1.79)Zone5 0.13 (0.81) 0.122 (0.97) 0.119 (0.61) .124 (0.78)Vacant −0.73⁎⁎⁎ (−2.84) −0.734⁎⁎⁎ (−2.99) −0.746 ⁎⁎⁎ (−3.06) − .734⁎⁎⁎ (−2.88)Constant 9.7⁎⁎⁎ (13.64) 9.69⁎⁎⁎ (16.39) 9.75 ⁎⁎⁎ (14.30) 9.66⁎⁎⁎ (13.54)Lambda −0.0657 (−0.30)Kennedy's Estimateb −7.24% −6.59% – −17.73%No. Obs. 131 131 131 123Easements 19 19 19 11R-square .6471 .6465 .6487Pseudo Rsq (1st stage) .5208

ΔRobust t-statistics in parentheses. ⁎significant at 10%, ⁎⁎significant at 5%, ⁎⁎⁎significant at 1%.a For 2-step treatment regressions, z-scores in parentheses.b Kennedy's estimate of the effect of an Easement on land prices, as given in Eq. (5).

442 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

the characteristics and distributional properties of the re-stricted and unrestricted parcel groups are fairly similar.

Thus our primary estimating model is:

ln pricei� ¼ aþ b1ln acresið Þ þ

Xk

bkcontrolki þ beeasementi þ ei

ð4Þ

4. Results and discussion

We begin by running simple regressions controlling for all ourcollected land characteristics and a dummy variable for re-stricted parcels (easement). Results are presented in Table 3,columns 1–3. Our model explains about 65% of the variation inthe log of sales prices and the signs of the control variablecoefficients are sensible. The coefficient of easement is slightlynegative, but statistically insignificant. Following Kennedy(1981), we calculate the effect of an easement on land prices as

bk ¼ 100 exp be �12br2b

�� 1

� ð5Þ

where βe and σ̂β2 are the estimated coefficient and variance of

easement, respectively. Kennedy's estimate is presentedat thebottom of Table 3 for each regression where it is calculable.7

7 We have not presented Kennedy's estimate for the two-stagetreatment effect models, as the standard errors are calculatednonlinearly and we are not certain whether Kennedy’s formulawould apply.

Regression (1) includes all the control variables, while re-gression (2) represents a simplified specification from system-atically deleting the control variables that were statisticallyinsignificant in the general model, a so-called ‘general-to-simple’ reduction to yield a final, parsimonious specificationthat economizes on degrees of freedom. In bothmodels (1) and(2), however, easements have a very small negative impact onprice that is not statistically significantly different from zero.

While theory does leave open the possibility that ease-mentsmay have zero (or even positive) effects, the analysis sofar cannot justify such a conclusion. There are several reasonswhy wemay not observe a negative effect (if it exists) with thedata as it stands. First, it could be that we have too few ob-servations to pick up an effect. Another possibility is that thereis a selection effect going on. In general, easement restrictionsare not a randomly assigned characteristic of a parcel, forexample there may be unobserved characteristics of easedparcels (including both characteristics of their owners as wellas unobserved characteristics of the land itself) that increasethe chances that they will be restricted. In that case the errorterms in the regressions presented in columns (1) and (2) ofTable 3 may include unobserved characteristics that deter-mine the probability that a parcel will be selected for a con-servation easement, creating a treatment bias in the results.

To address this possibility, we assume that a parcel will berestricted if the returns (in terms of total utility) to the land-owner are greater than leaving it unrestricted. The actual netreturn to the landowner from a restriction is an unobservedlatent variable, say zi⁎. While we cannot observe zi⁎ directly,

Table 4 – Dependent Variable=Ln(Price)

Vacant parcels only sampleΔ

5 6 7 8

OLS OLS 2S treatmenta OLS

Easement −0.391 (−1.59) − .388⁎ (−1.72) −0.551⁎ (−1.66)Hard easement − .600⁎⁎⁎ (−2.91)ln_acres 0.994⁎⁎⁎ (6.05) .98⁎⁎⁎ (6.49) 1.05⁎⁎⁎ (6.94) .982⁎⁎⁎ (6.81)ln_dist −0.0905 (−0.32) − .311⁎ (−1.84) −0.123 (−0.48) .0139 (0.05)Road_den 0.136 (1.18) .0736 (0.70) 0.124 (0.98) .123 (1.06)Jeff −0.0804 (−0.21) −0.0107 (−0.02) − .0682 (−0.18)Col −0.242 (0.83) −0.211 (−0.66) − .335 (−1.14)Year2 −0.202 (−0.94) − .156 (−0.90) −0.235 (−1.24) − .169 (−0.78)Year3 −0.153 (−0.50) −0.167 (−0.64) − .0317 (−0.10)Village 0.708⁎⁎ (2.57) .655⁎⁎⁎ (2.74) 0.708 (1.04) .712⁎⁎ (2.63)Zone3 −0.141 (−0.41) −0.167 (−0.34) − .293 (−0.90)Zone4 0.798 (1.38) .719 (1.29) 0.917⁎⁎ (2.01) .781 (1.37)Zone5 0.157 (1.01) 0.163 (0.69) .137 (0.91)Constant 7.82⁎⁎⁎ (6.82) 8.62⁎⁎⁎ (9.79) 7.79⁎⁎⁎ (8.26) 7.6⁎⁎⁎ (6.67)Lambda 0.229 (0.79)Kennedy's Estimateb −35.00% −34.50% – −47.78%No. Obs. 72 72 72 65Easements 15 15 15 8R-square .5664 .5549 .5781Pseudo Rsq (1st stage) .7913

ΔRobust t-statistics in parentheses. ⁎significant at 10%, ⁎⁎significant at 5%, ⁎⁎⁎significant at 1%.a For 2-step treatment regressions, z-scores in parentheses.b Kennedy's estimate of the effect of an Easement on land prices, as given in Eq. (8).

443E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

we do observe its sign, zi, which takes the value 1 if zi⁎ ispositive and the parcel is restricted, and 0 otherwise. Follow-ing the standard two-step treatment estimation (see Greene,1993) we estimate a first stage probit selection equation of zi:

zi ¼ g0 þXn

gncontrolni þ Ai: ð6Þ

Ideally we would want to include in Eq. (6) at least oneexplanatory variable that influenced the landowner's decisionto put a restriction on the land, but which did not directlyimpact the market sale price. In practice there are no suchvariables available in our data set. Instead, we build a modelwhich explains selection into an easement program with dif-ferent transformations of some of the same variables that weuse in the primary land price equations. Thus, only in theprobit selection model, we use acres and acres_squared in-stead of ln(acres), and improvements and distance, instead oftheir respective logs. Other control variables include zone3,zone5, dane, year2, year3, vacant, and road_den. Zoning hasbeen found to have varying effects on land value (seeHenneberry and Barrows, 1990), however our zoning variablesdid not have a strong influence, which is not surprising as wefocused on agricultural land. The parameters from Eq. (6) areused to calculate an Inverse Mills Ratio, or hazard term(Lambda), which reflects the likelihood that a particular parcelwill be selected for an easement restriction. Lambda is thenincluded in the second stage regression to control for theendogenous treatment effects. We rely on both the transfor-mations of the independent variables as well as the nonlinearconstruction of the Inverse Mills Ratio itself for identification.Column (3) reports the results. We still do not find anysignificant effects of easements. The pseudo R-squared from

the first stage probit model (6) is a respectable 0.53, howeverthe hazard term, lambda, is not statistically significant. Thisfinding is also consistent with those of Nickerson and Lynch(2001) who also failed to find evidence of selection bias in theirdata, suggesting that endogenous treatment effects may nothave a major influence on the results.

This is not to say that there is definitely no self-selection ofparcels into restricted status – quite to the contrary, there arevery strong theoretical and anecdotally-supported reasons toexpect that endogenous selection of certain kinds of parcels isin fact a major characteristic of conservation easementprograms. However, whatever self-selection does occur eitherdoes not have a systematic correlation with the developmentvalues, or cannot be detectedwith the observed characteristicsthatwe are able to include in the first-stage selection equation.

Finally, we consider whether the fact that conservationeasements are themselves heterogeneous could explain whywe do not detect an effect on sales prices. We define a newvariable, hard easement, which is a conservation easementthat does not allow any development at all in the future (i.e.,no additional development on already improved parcels andthe preservation of vacant status on unimproved parcels). Onthe one hand, we would expect a bigger impact from moresevere restrictions, but on the other hand, fewer parcels withhard easement increases the difficulty of finding statisticalsignificance. Column (4) of Table 3 presents the results usinghard easement instead of easement. Instead of 19 restrictedproperties, we now have only 11. However, the magnitude ofour estimated effect is now much larger and is much moretightly estimated, albeit still not statistically significant.

Thus the results in columns (1)–(4) of Table 3 are quitesimilar to those found by Nickerson and Lynch (2001) — there

444 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

is some very weak suggestion that easements may have anegative effect, but the correlation is not statistically signifi-cant. However, we have yet to consider the conditions underwhich an easement really bites. As we discussed above, whilea property that is both improved (e.g., with a house or barn)and easement-restricted cannot be subdivided or developed,people can live on it and even remodel/improve the existingbuildings (and the presence of an easement suggests that thesurrounding land may be very nice). Thus, that parcel wouldretain good value for either agricultural work or as a weekendretreat/holiday house. Hence, while the easement should, intheory, lower the sales price, wemay not be able to pick up thiseffect with our data, either because the sample is too small,the effect may be modest to begin with if immediatesubdivision pressures are not looming, or because the ease-ment may be associated with a lovely setting, which is anunobserved characteristic of the plot.

On the other hand, a vacant plot with a developmentrestriction is much less attractive. If it is agricultural land it isreally only of any use to the neighboring farms. Farmersfurther away will find it inconvenient to access and urbanbuyers cannot stay there, regardless how lovely. Thus, by solimiting the effective market, an easement may have a muchmore pronounced impact on vacant land.

Followingour intuition, in Table 4we explorewhathappensif we restrict our sample to vacant land only. First, we lose 59observations, 4 of which are restricted parcels. However, nowour sample of 72 parcels is more homogenous in nature.Hopefully, by limiting ourselves to only those plots where wewould expect to see the largest impact of easements we willfurther indirectly control for many unobservable character-istics which may have confounded our analyses with moreheterogeneous data. Furthermore, sincemost easements wereon vacant land, in the earlier regressions it may have beendifficult to identify the easement effect from the vacant effect.

Table 5 – Dependent Variable=Ln(Price)

V

9

OLS O

Easement − .664⁎⁎⁎ (−2.80) − .668⁎Hard easementln_acres .997⁎⁎⁎ (7.13) .993⁎⁎⁎Ln_dist − .391 (−1.47) − .425⁎Road_den .0376 (0.38) .0268 (Col − .0404 (−0.13)Year2 − .253 (−1.36) − .242Year3 − .0335 (−0.15)Village .37 (1.57) .382⁎ (Constant 9.14⁎⁎⁎ (10.52) 9.25⁎⁎⁎LambdaKennedy's Estimateb −50.14% −50.19No. Obs. 47 47Easements 11 11R-square .6855 .6852Pseudo Rsq (1st stage)

ΔRobust t-statistics in parentheses. ⁎significant at 10%, ⁎⁎significant at 5%a For 2-step treatment regressions, z-scores in parentheses.b Kennedy's estimate of the effect of an Easement on land prices, as give

Regressions (5) and (6) of Table 4 present the results fromboth the full, general model and the reduced, parsimoniousmodel using our vacant-only sample. The coefficient ofeasement is still negative and much larger in magnitudethan we found in the full sample. Furthermore, while ease-ment is not strictly statistically significant in the generalmodel, it's p-value of just under 0.12 is very close. Once weeliminate insignificant variables the coefficient does manageto tip into 10%-significance with a p-value of around 0.09.Essentially, our finding that easement is barely significant atthe 10%-level is not particularly strong evidence but certainlysuggestive of an effect. The estimated impact of an easement(following Kennedy, as in Eq. (5) above) is about a 35%reduction in sales price. In column (7) we present the outcomefrom a two-step treatment effects model. Controlling for theprobability that a parcel was selected for an easement nowincreases the magnitude, if not the statistical significance, ofour easement variable. However, as before, the hazard term,lambda, is not statistically significant.

Our story about why easements should impact vacant landmore than improved land suggests that the strongest kind ofeasement – i.e. those that allow no future development at all –will have the biggest impact. Thus, in column (8), we againintroduce hard easement into the model. Now we have only 8restricted parcels in our sample, but the coefficient jumps to −0.60 and is statistically significant at the 1% level. Thus ourintuition seems to be justified; we find that strict no-develop-ment easements on vacant land reduce sale prices by 47% onaverage.

Finally, we focus our attention on the sample where wewould expect to observe the biggest impact of all from adevelopment easement by restricting our sample to vacantagricultural land. As discussed above, theoretically we wouldexpect a restriction on vacant agricultural land to significantlyreduce the potential market as these parcels are primarily of

acant agricultural land sampleΔ

10 11 12

LS 2S treatment a OLS

⁎⁎ (−2.87) − .782⁎⁎ (−2.56)− .509⁎⁎ (−2.37)

(7.52) 1.04⁎⁎⁎ (7.47) .865⁎⁎⁎ (6.46)⁎⁎ (−2.83) − .421⁎ (−1.90) − .207 (−0.76)0.27) .0247 (0.23) .0799 (0.76)

− .0127 (−0.05) − .235 (−0.76)(−1.65) − .282 (−1.61) − .141 (−0.69)

− .0453 (−0.20) .13 (0.57)1.80) .363 (0.78) .406⁎ (1.74)(12.79) 9.14⁎⁎⁎ (11.77) 8.9⁎⁎⁎ (10.32)

.142 (0.60)% – −42.05%

47 4211 6

.6606.7808

, ⁎⁎⁎significant at 1%.

n in Eq. (5).

445E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

interest only to farmers in relative proximity. Furthermore,given that costly data collection leaves us with a limitednumber of control variables, restricting the sample to a morehomogenous group of parcels should help to reduce theimpact of unobservable characteristics. The limitation is thatour sample size is reduced to 36 non-restricted and 11restricted, vacant parcels.

Table 5 presents the results of our analysis with the vacantagricultural land sample. In regressions (9) and (10) ourmodelsexplain just over 68% of the variation in ln(price).With a highlystatistically significant coefficient of − 0.66, the effect of ease-ment is estimated to reduce market land values by about 50%.In column (11) we again try a two-step treatment effectsestimation. This increases themagnitude of the coefficient oneasement and slightly lowers the statistical significance.However, again the hazard term is not found to be statisticallysignificant.

Lastly, in column (12) of Table 4 we estimate the impact ofhard easement on vacant agricultural land. The number ofrestricted parcels is now only 6, and the estimate of the mag-nitude of the impact on land prices falls to 42%. The coefficientis still significant, but at the 5% rather than the 1% level.

5. Conclusions

Understanding the contribution of development rights to landprices will not only inform public policy decisions regarding taxincentives, but will also help assess the effectiveness of currenteasementprogramsandguideprogramdesign. If easements arenot lowering sales price, the common goal of making agricul-tural land more affordable for farmers is not being met, taxpayers may be unfairly subsidizing the incomes of landownerswho donate or sell easements, and it may be an indication thatlandat low risk of development is beingpreserved.Additionally,understanding easement price effects will help agenciesnegotiate with private landowners over contracts and compen-sation associated with extinguishing the development rights.However, the lack of a reliablemechanism to evaluate the valueof development rights has led some to doubt whether ease-ments have a measurable effect on prices at all, and has leftconservation groups with little hard empirical evidence withwhich todefend the tax benefits associatedwith their programs.

This paper attempts to address this deficit. We providestrong and robust evidence of a significantly negative impactof conservation easements on market land prices anddemonstrate why most previous studies may not have beenable to pick up this effect. In particular, given the difficulties ofgathering data on myriad basically unobservable plot char-acteristics, we find that a measurable impact of easementsshows up empirically in only a subset of transactions. For landthat already includes a residence, the possible uses and desi-rability on the open market of an easement-restricted parcelmay be much greater than for a vacant parcel with reduced orno possibility of building a house. In fact, due to self-selection,an easement-restricted parcel with a homemay bemore likelyto contain interesting wildlife or scenic vistas than otherparcels without easements and thus may make an even moredesirable residence or vacation home. On the other hand, avacant parcel – especially a vacant agricultural parcel – would

be of use primarily to neighboring farmers or hunters and thusface amuchmore limited set of potential buyers. Theoreticallywe would still expect the former parcel (with home) to suffer aprice decline when encumbered with an easement, but in theabsence of control variables for amenities such as wildlife andviews, the reduction may be too heterogeneous and slight todetect with such a limited sample.

Indeed,whenwe limit our sample to vacant land only, evenwith a severely limited data set we find statistically significantevidence that conservation easements reduce the marketvalues of land, and we estimate that effect from around 35%(with high variance) up to just under 50% for vacant land,depending on the severity of the restrictions imposed by theeasement. For vacant agricultural land we find that develop-ment easements reducemarket land prices by 50%. Consistentwith Nickerson and Lynch (2001) we find no evidence that self-selection into easement programs has had an effect on theseestimates. However we suspect that this finding could havemore to do with our inability to collect sufficient data onamenity characteristics of the parcels and leavemore analysison the topic for future research.

Acknowledgements

We gratefully acknowledge the support of Bill Provencher,who provided constant support and advice, Greg Delwiche ofthe Wisconsin DNR, Arlin Brannstrom from the University ofWisconsin Center for Dairy Profitability, JimWelsh at the DaneCounty Natural Heritage Foundation, Pam Foster-Felt atGathering Waters Conservancy, the staff at the Dane, Jeffer-son, and Columbia County Courthouses, Andy Erdman fromthe Jefferson County Land Information Office, and KristenAndersen from the Columbia County Land Information Office.We further thank Jeremy Foltz, Dan Bromley, Rich Bishop, andtwo anonymous referees for their insightful comments andsuggestions. All errors and omissions are our own.

R E F E R E N C E S

Beasly, S.D., Workman, W.G., Williams, N.A., 1986. Estimatingamenity values of urban fringe farmland: a contingentvaluation approach: note. Growth and Change 17, 70–78 (Oct).

Bergstrom, J.C., Dillman, B.L., Stoll, J.R., 1985. Public environmentalamenity benefits of private land: the case of prime agriculturalland. Southern Journal of Agricultural Economics 17, 139–150(July).

Blakely, M., , 1991. An Econometric Analysis of the Effects ofDevelopment Rights Purchases on Land Values in King County,Washington. M.S. thesis, Washington State University.

Boyd, J., Caballero, K., Simpson, R.D., 1999. The law and economicsof habitat conservation: lessons from an analysis of easementacquisitions. Resources for the Future Discussion Paper 99–32.

Drake, L., 1992. The non-market value of Swedish agriculturallandscape. European Review of Agricultural Economics 19 (3),351–364.

Farja, Y., Rausser, G.C., 2007. Land trusts policy: is it sociallyoptimal? Selected Paper prepared for presentation at theAmerican Agricultural Economics Association Annual Meeting,Portland, OR, July 29–August 1, 2007.

Foster, J.H., Halstead, J.M., Stevens, T.H., 1982. Measuring theNonmarket Value of Agricultural Land: A Case Study. Research

446 E C O L O G I C A L E C O N O M I C S 6 8 ( 2 0 0 8 ) 4 3 7 – 4 4 6

Bulletin, vol. 672. Massachusetts Agricultural ExperimentStation, University of Massachusetts, Amherst, MA.

Gathering Waters Conservancy, 2001. The Impacts ofConservation Easements on Property Taxes in Wisconsin.www.gatheringwaters.org.

Geoghegan, J., 2002. The value of open spaces in residential landuse. Land Use Policy 19 (1), 91–98.

Geoghegan, J., Lynch, L., Bucholtz, S., 2003. Capitalization of openspaces into housing values and residential property taxrevenue impacts of agricultural easement programs.Agricultural and Resource Economics Review 32 (1), 33–45.

Greene, W.H., 1993. Econometric Analysis, 2nd ed. Prentice-Hall,Englewood Cliffs, NJ.

Halstead, J.M., 1984. Measuring the non-market value ofMassachusetts agricultural land: a case study. NortheasternJournal of Agricultural Economics 14, 12–19.

Heimlich, R.E., Anderson, W.D., 2001. Development at the urbanfringe and beyond: impacts on agriculture and rural land.USDA ERS Agricultural Economic Report No. 803. 88 pp., June.

Henneberry, D.M., Barrows, R.L., 1990. Capitalization of exclusiveagricultural zoning into farmland prices. Land Economics 66,249–258 (August).

Irwin, E.G., Bockstael, N.E., 2001. The problem of identifying landuse spillovers: measuring the effects of open space onresidential property values. American Journal of AgriculturalEconomics 83 (3), 98–704.

Joint Committee on Taxation, 2005. Options to Improve TaxCompliance and Reform Tax Expenditures. JCS-02-05, reportprepared for Congress, January 27.

Kennedy, P.E., 1981. Estimation with correctly interpreted dummyvariables in semilogarithmic equations. American EconomicReview 71, 801.

Land Trust Alliance, 2006. The 2005 National Land Trust CensusReport. http://www.lta.org.

Lynch, L., Gray, W., Geoghegan, J., 2007. Are farmland preservationprogram easement restrictions capitalized into farmlandprices? What can a propensity score matching analysis tell us?Review of Agricultural Economics 29 (3), 502–509.

Nickerson, C.J., Lynch, 2001. The effect of farmland preservationprograms on farmland prices. American Journal of AgriculturalEconomics 83 (2), 341–351.

Plantinga, A.J., 2007. The economics of conservation easements.In: Ingram, G.K., Hong, Y.H. (Eds.), Land Policies and theirOutcomes, Proceedings of the 2006 Land Policy Conference.Lincoln Institute of Land Policy.

Plantinga, A.J., Miller, D.J., 2001. Agricultural land values and thevalue of rights to future land development. Land EconomicsFebruary 77 (1), 56–76.

Plantinga, A.J., Lubowski, R.N., Stavins, R.N., 2002. The effects ofpotential land development on agricultural land prices. Journalof Urban Economics 52 (3), 561–581.

Pruckner, G.J., 1995. Agricultural landscape cultivation in Austria:an application of the CVM. European Review of AgriculturalEconomics 22 (2), 173–190.

Sorenson, A.A., Greene, R.P., Russ, K., 1997. Farming on the Edge.American Farmland Trust. Center for Agriculture in theEnvironment (March).

Stephens, Joe, Charities Fight for Easement Donors: PreservationGroups Target Legislators in Move to Save Tax Breaks.Washington Post, February 26, 2005; Page A02.

Taff, S.J., 2004. Evidence of a market effect from conservationeasements. Staff Paper 04-9.

Vesterby, M.A., Krupa, K.S., 2001. Major uses of land in theUnited States, 1997. USDAERS Statistical BulletinNo. 973. 60 pp.

Zhang, X., 2004. The Effects of Conservation Easements on LandValues. Masters Thesis. Virginia Polytechnic Institute (May).