1# RECORD 1196e55c7401009f4e60d7c8db96469c1e13b00e2# TITLE: Capitalization of Open Spaces into Housing Values and the Residential Property Tax Revenue Impacts of Agricultural Easement Programs3# YEAR: 2003.0 VENUE: Agricultural and Resource Economics Review DOI: 10.1017/s10682805000024834# AUTHORS: J. Geoghegan; L. Lynch; S. Bucholtz5# PAGES: 13 CANDIDATE TABLE PAGES (1-based): [6, 8, 9, 10]6# ABSTRACT: nan789===== PAGE 1 (score 16, numeric-paren 0, table-words 0, var-words 16) =====1011Capitalization of Open Spaces into12Housing Values and the Residential13Property Tax Revenue Impacts of14Agricultural Easement Programs15Jacqueline Geoghegan, Lori Lynch, and Shawn Bucholtz16Using a unique spatial database, a hedonic model is developed to estimate the value to nearby17residents of open space purchased through agricultural preservation programs in three Maryland18counties. After correcting for endogeneity and spatial autocorrelation, the estimated coefficients are19used to calculate the potential changes in housing values for a given change in neighborhood open20space following an agricultural easement purchase. Then, using the current residential property tax21for each parcel, the expected increase in county tax revenue is computed and this revenue is22compared to the cost ofpreserving the lands.23Key Words: agricultural preservation, hedonic models, open space, spatial econometrics24The preservation of open space has become an25important policy issue in the United States as con-26version ofland in forestry and agricultural uses into27residential and commercial uses continues. From281982 to 1992, approximately 6.2 million acres of29agricultural land and 5.1 million acres offorest land30were converted to urban and other developed uses31in the United States (Vesterby et aI., 1997). In the32Washington, DC, metropolitan area,therate atwhich33land is being consumed exceeds the population34Jacqueline Geoghegan is assistant professor, Department of Economics,35Clark University, Worcester, MA; Lori Lynch is associate professor and36Shawn Bucholtz is graduate student, both in the Department of Agricul-37tural and Resource Economics, University ofMaryland.38The authors thank Michael Fragkias, Laura Schneider, and Suchandra39Basu for their assistance with this research, as well as two anonymous40reviewers and participants at the NAREA 2002 Land Use Policy Work-41shop for helpful comments and suggestions. This research was supported42by U.S. EPA Grant No. R825309, NASA New Investigator Program in43Earth Sciences Grant No. NAG5-8559, Center for Agricultural and44Natural Resource Policy, and the National Center for Smart Growth45Research and Education at the University of Maryland. The geographic46data in the analysis are from a project coordinated by Professor Nancy47Bockstael, Department of Agricultural and Resource Economics, the48University of Maryland, with funding by EPA Cooperative Agreement49No. CR-821925010, EPA STAR Grant No. R-82801201, and a Maryland50Sea Grant.51This paper was presented at the Land Use Policy Workshop of the52Northeastern Agricultural and Resource Economics Association annual53meetings, Harrisburg (Camp Hill), PA, June 9-11, 2002.54growth rate by a factor of almost 2.5. If current55trends continue, 800,000 additional acres ofresource56lands will be developed by 2030 in the greater57Washington, DC, region (Chesapeake Bay Founda-58tion, 2002).59Open space provides many potential public goods60with aesthetic, recreation, and biodiversity values.61It also offers associated ecosystem services such as62flood control and waterpurification. Farmland pres-63ervation programs can help maintain a land base for64the agricultural economy, provide the amenities of65open space and rural character, slow suburban66sprawl, provide critical wildlife habitat, and reduce67pollution in areas where suburban development is68occurring (Bromley and Hodge, 1990;Fischel, 1985;69Gardner, 1977; McConnell, 1989; Wolfram, 1981).70Because the private land market does not71recognize these public goods (i.e., there is a market72failure), private and governmental entities have in-73troduced a variety ofmechanisms to preserve open74space or slow its conversion, including cluster and75exclusive agricultural zoning, purchase of and76transfer ofdevelopment rights programs (PDRs and77TDRs), use ofagricultural easements, and the out-78right purchase of open spaces for parks using tax79dollars or private donations. The general public80Agricultural and Resource Economics Review 32/1 (April 2003): 33-4581Copyright 2003 Northeastern Agricultural and Resource Economics Association82https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press838485===== PAGE 6 (score 87, numeric-paren 54, table-words 2, var-words 18) =====868738 April 200388Agricultural and Resource Economics Review89Table 2. Descriptive Statistics and Variable Definitions by County90Calvert Co.91Carroll Co.92Howard Co.93Mean94Mean95Mean96Variable Name97Variable Description98(Std. Dev.)99(Std. Dev.)100(Std. Dev.)101LN_PRICE102Dependent variable: Naturallogarithm offull10311.942310411.909810512.3072106transaction price (in dollars)107(0.4143)108(0.3471)109(0.3457)110(P_)LGPERMOPENa111Percent ofpermanent open space within large radius1120.05021130.12141140.1245115(1,600-meter radius buffer)116(0.0404)117(0.0777)118(0.0549)119(P_)SMPERMOPENa120Percent of permanent open space within small radius1210.01101220.04191230.0269124(100-meter radius buffer)125(0.0797)126(0.1650)127(0.1130)128(P_)LGAGFORESTa129Percent ofagricultural and forest land within large1300.50541310.53031320.3594133radius (1,600-meter radius buffer)134(0.1653)135(0.0838)136(0.0932)137(P_)SMAGFORESTa138Percent of agricultural and forest land within small1390.20171400.19931410.1375142radius (100-meter radius buffer)143(0.2724)144(0.2727)145(0.0780)146LN_ACRE147Natural logarithm ofparcel size (in acres)148-0.3632149-0.4869150-0.8688151(1.0826)152(1.0801)153(0.9082)154LN_AGE155Natural logarithm ofage ofstructure (in years)1561.64011572.26541581.7306159(1.5266)160(1.5149)161(1.3778)162LN_SQFT163Natural logarithm ofsize ofstructure (in square feet)1647.27931657.38921667.6759167(0.4341)168(0.3869)169(0.4302)170AVERAGE171Dummy variable for a structure ofaverage quality1720.57521730.28311740.5807175(0.4945)176(0.4506)177(0.4935)178GOOD_VGOOD179Dummy variable for a structure of good or very good1800.02271810.02201820.3894183quality184(0.1489)185(0.1468)186(0.4877)187AVSTORY188Number ofstories of structure1891.46571901.60881911.8098192(0.4733)193(0.4712)194(0.3862)195BASEMENT196Dummy variable for presence ofa basement in the1970.59311980.95241990.9326200structure201(0.4914)202(0.2129)203(0.2507)204FRAME205Dummy variable for framed structures2060.46122070.08752080.6842209(0.4986)210(0.2825)211(0.4649)212ALUM213Dummy variable for aluminum structures2140.47552150.70382160.2186217(0.4996)218(0.4567)219(0.4133)220LN_POPDEN221Natural logarithm ofpopulation density for census2224.76082235.19152245.8756225tract (in households per square mile)226(0.8077)227(0.9585)228(1.1733)229%EDUCB230Percent ofbachelor's degree education attainment2310.06972320.09202330.1753234(0.0247)235(0.0361)236(0.0543)237MEDHSINC238Median household income of census tract (in dollars)23947,66524043,64524162,181242(9,533)243(7,256)244(13,623)245LN_CITYDIST246Natural logarithm of distance to Washington, DC,24711.113324810.671124910.1559250or Baltimore,:MD (in 100 kilometers)251(0.2367)252(0.2233)253(0.2362)254LN_TOWNDIST255Natural logarithm of distance to nearest town2567.66652577.86542588.4256259(in 100 kilometers)260(0.7375)261(0.7486)262(0.8121)263a Either actual value or predicted value (see regression results, tables 3, 4, and 5).264agricultural or forest use, (b) are not factors that265explain housing prices, and (c) are not highly corre-266lated with the observed explanatory variables that267influence housing prices.268In addition to the soil attribute and slope variables269used in Irwin and Bockstael (2001), we also include270the distance to nearest transportation node and271whether the area is currently or is planned to be con-272nected to a sewer system. The physical attributes273measuring soil quality are clearly exogenous to land274values and to the other variables, as they are pre-275determined and do not change with a change in land276use. Transportation node and city distance variables277were tested for correlation, which was found to be278small (p ranged from 0.02 to 0.23).279Sewer service is expected to influence decisions280in the future for these currently undeveloped281land uses. Using the estimated coefficients from282these regressions, instruments for P_SMPERMOPEN,283P_SMAGFOREST, P_LGPERMOPEN, and P_LGAGFOREST284https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press285286287===== PAGE 7 (score 14, numeric-paren 0, table-words 4, var-words 6) =====288289Geoghegan. Lynch. and Bucholtz290Capitalization ofOpen Spaces into Housing Values 39291where u = AWU + e, and e ''IN(O, 0' 21). The weight292matrix W reflects the relationship between each i,j293pair of observations. For the models reported294below, a row-standardized inverse distance weight295matrix W was used, which assumes the spatial296effects decrease as the distance between observa-297tions increases. The elements ofW are defined such298that wij = 11dij if dij < C, and wij = 0 if i = j or if299d.> c, where dijis the distance between parcel i and300parcel j, C is the distance after which no spatial301correlation is expected, and c is set to 1,000 meters.302That is, only those "neighbors" within 1,000 meters303are assumed to have the potential for spatial effects.304Becausehedonic models arereduced-form models305of an equilibrium locus of offers and bids for306parcels ofland, economic theory cannot inform the307empirical specification [see Cropper, Deck, and308McConnell (1988) for further discussion]. There-309fore, following previous research (Bockstael, 1996;310Geoghegan, Wainger, and Bockstael, 1997), flexible311were created. If a Haussman test indicated the312presence of simultaneity in the price equation with313the land use variables, the predicted value from the314instrumental variable step, rather than the actual315value for the land use buffers, was used in the final316regression for each county.317In addition to the endogeneity issue involved318with estimating the hedonic model, the use of319spatial data leads to the potential for bias and320inefficiency if spatial effects are not taken into321account. One ofthe most relevant ofthese effects is322spatial autocorrelation, which results in inefficient323estimates and biased standard errors. Spatial auto-324correlation is likely to occur when important vari-325ables common to all observations in a neighborhood326are not observable by the researcher, and thus not327included in the model. Spatial autocorrelation in the328residuals was tested using Moran's I statistic, the329likelihood-ratio test, and a Lagrange multipliertest.330In each county, these tests were conducted to indi-331cate if there was spatial autocorrelation in the332residuals. Ifspatial autocorrelation existed, a spatial333model was estimated to address this issue. Similar334to serial correlation, corrections require specification335of a pattern to the errors. In this analysis, spatial336weight matrices based on the inverse distance are337employed.338Specifically, assume X is a matrix ofthe explan-339atory variables S, L, and G of equation (1). Then340the spatial model discussed above is specified as:341(2)342R=Xp+u,343function forms are used here, specifically a log-log344specification.345Estimation Results346The results from estimating equation (1) (the OLS347model) and equation (2) (the spatial model) for348Calvert, Carroll, and Howard counties are reported349in tables 3, 4, and 5, respectively. The test statistics350results ofthe spatial autocorrelation in the residuals351for all three counties are provided in table 6. While352for each county the results from the tests on spatial353autocorrelation in the residuals demonstrated the354need to use the spatial model, we found no major355differences when comparing the OLS results to the356spatial model results. Estimated coefficients are357generally ofthe same size, and the more efficient358test statistics from the spatial model do not result in359qualitative or statistical significance changes for the360estimatedcoefficients. Therefore, the following dis-361cussion of the parameter estimates is based on the362results from the spatial model estimations.3363As seen from tables 3, 4, and 5, the estimated364coefficients on the lot size and structural character-365istics all meet a priori expectations and are statisti-366cally significant for the three counties, except for367insignificant coefficients on NUSTORY for Carroll368County and BASEMENT for Calvert County. The369estimated coefficients for other characteristics varied370both qualitatively and quantitatively by county. For371example, the coefficient on LN_CITYDISTis negative372and statistically significant in the Calvert and373Carroll county regressions; the closer a parcel in374these counties is to eitherWashington or Baltimore,375the higher the sales price, as would be predicted by376urban economic theory. LN_CITYDIST is not statisti-377cally significant in Howard County. In explaining378this lack of significance, we argue that because379Howard County is situated between Washington380and Baltimore, it is possible Howard residents value381a location affording a dual-commute opportunity,382rather than simply being closer to one city versus383the other.384We originally hypothesized that the closer a par-385cel is to a town, the higher the price. However, our386findings show the distance to the nearest town387(LN_TOWNDIST) is not statistically significant in388Carroll or Howard counties, but is positive and389significant in Calvert County. Calvert residents390pay more to be farther away from towns, possibly3913 Semi-annual time dummy variables were included in the regressions392reported below and were all statistically significant.393https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press394395396===== PAGE 8 (score 87, numeric-paren 53, table-words 6, var-words 6) =====39739840 April 2003399Agricultural and Resource Economics Review400Table 3. Regression Results for Calvert County401OLS402Spatial Model403Asymptotic404Variable405Coefficient406t-Statistic407Coefficient408t-Statistic409CONSTANT41010.8992*41122.0441210.7624*41315.78414P_LGPERMOPEN4150.6303*4163.594170.7118*4182.83419SMPERMOPEN420-0.1134421-1.72422-0.1117423-1.67424P_LGAGFOREST425-0.3492*426-7.42427-0.3907*428-6.03429SMAGFOREST430-0.0482431-2.38432-0.0496433-2.42434LN_ACRE4350.0982*43613.834370.1002*43812.87439LN_AGE440-0.0226*441-5.31442-0.0287*443-6.28444LN_SQFT4450.3894*44620.464470.3765*44819.49449AVERAGE4500.2106*45113.884520.2020*45312.94454GOOD_VGOOD4550.4859*45612.294570.4699*45811.56459NUSTORY4600.0790*4615.994620.0655*4635.14464BASEMENT465-0.0126466-1.12467-0.0008468-0.07469FRAME470-0.0052471-0.22472-0.0311473-1.36474ALUM475-0.0422476-1.82477-0.0464478-2.07479LN_POPDEN480-0.0396*481-3.67482-0.0393*483-2.85484%EDUCB4850.9934*4863.944871.0015*4882.83489MEDHSINC4900.0000491-0.304920.0000493-0.18494LN_CITYDIST495-0.1669*496-4.27497-0.1432*498-2.59499LN_TOWNDIST5000.0241*5012.675020.02655032.02504Lambda5050.4067*50612.15507Adjusted R!5080.75495090.7785510* Denotes statistical significance at the 5% level.511Table 4. Regression Results for Carroll County512OLS513Spatial Model514Asymptotic515Variable516Coefficient517t-Statistic518Coefficient519t-Statistic520CONSTANT52111.4720*52237.4452311.6311*52432.47525P_LGPERMOPEN526-0.1181527-1.86528-0.1171529-1.57530SMPERMOPEN5310.04435322.015330.03175341.33535P_LGAGFOREST536-0.1932*537-3.39538-0.2074*539-3.26540SMAGFOREST5410.00275420.205430.00275440.20545LN_ACRE5460.1334*54731.685480.1338*54931.06550LN_AGE551-0.0641 *552-21.80553-0.0610*554-20.15555LN_SQFT5560.2692*55724.295580.2629*55924.00560AVERAGE5610.1546*56217.515630.1551*56416.76565GOOD_VGOOD5660.4036*56717.175680.3973*56916.69570NUSTORY571-0.0212572-2.52573-0.0155574-1.83575BASEMENT5760.1324*5778.745780.1330*5798.87580FRAME581-0.0528*582-4.05583-0.0488*584-3.79585ALUM586-0.0015587-0.16588-0.0007589-0.08590LN_POPDEN5910.0286*5924.755930.0274*5944.02595%EDUCB5960.08145970.625980.00835990.06600MEDHSINC6010.0000*6026.826030.0000*60423.02605LN_CITYDIST606-0.1581 *607-6.45608-0.1687*609-5.80610LN_TOWNDIST611-0.0044612-0.76613-0.0048614-0.75615Lambda6160.2149*61721.15618Adjusted R"6190.73946200.7458621* Denotes statistical significance at the 5% level.622https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press623624625===== PAGE 9 (score 58, numeric-paren 30, table-words 5, var-words 10) =====626627Geoghegan, Lynch, and Bucholtz628Capitalization ofOpen.Spaces into Housing Values 41629Table 5. Regression Results for Howard County630OLS631Spatial Model632Asymptotic633Variable634Coefficient635t-Statistic636Coefficient637t-Statistic638CONSTANT6399.8834*64059.9864110.1107*6421,661.01643P_LGPERMOPEN6440.4950*6456.406460.5869*6473.91648SMPERMOPEN6490.03676501.696510.0489*6522.02653P_LGAGFOREST654-0.1612*655-2.89656-0.2035657-2.31658P_SMAGFOREST6590.2066*6603.186610.14826621.75663LN_ACRE6640.1045*66526.826660.1114*66723.79668LN_AGE669-0.0559*670-20.79671-0.0566*672-18.56673LN_SQFT6740.251867534.226760.2031 *67729.46678AVERAGE6790.2258*68014.676810.2300*68214.07683GOOD_VGOOD6840.4152*68524.966860.4120*68723.06688NUSTORY6890.0675*6908.046910.0734*6928.97693BASEMENT6940.0847*6958.136960.0731 *6977.15698FRAME699-0.0736*700-7.85701-0.0538*702-6.00703ALUM704-0.0919*705-8.58706-0.0601 *707-5.65708LN_POPDEN7090.0089*7102.627110.00417120.74713%EDUCB7140.2816*7154.867160.16017171.62718MEDHSINC7190.0000*72013.317210.0021 *7225.70723LN_CITYDIST724-0.0033725-0.247260.01547271.68728LN_TOWNDIST729-0.0059730-1.327310.00047320.05733Lambda7340.6339*735167.09736Adjusted R!7370.73907380.7741739* Denotes statistical significance at the 5% level.740because not all of Calvert towns provide desirable741amenities or employment opportunities. Towns be-742tween and within these counties vary enormously,743For both Carroll and Howard, we find a high degree744of correlation between distance to towns and745distance to city (p ranges from -0.65 to -0.45), but746dropping this variable from the estimation did not747change any ofthe results appreciably.748Hypothesis tests on the estimated coefficients for749the block-group census variables resulted in some750unexpected outcomes. In Howard County, the esti-751mated coefficient on LN_POPDENis not statistically752significant, in Carroll it is positive and statistically753significant, and in Calvert it is negative and statis-754tically significant. However, population density755could have two opposite effects on housing prices.756Our a priori hypothesis was that population density757is a measure ofcongestion. Nevertheless, it is con-758ceivable that population density acts as a proxy for759the density ofother goods and services which might760attract people and bid up the housing prices-761although we hope to have captured these locational762amenities that are capitalized in the value of the763home through inclusion ofother right-hand-side764variables. MEDHSINchas the expectedpositive coef-765ficient in Carroll and Howard counties, but is not766statistically significant in Calvert County. The co-767efficient on %EDUCB is positive and statistically768significant only in Calvert County. These census769variables are averages over the block group rather770than for the micro-neighborhoods surrounding the771transaction parcels comprising the data set. Unfor-772tunately, these are the most disaggregated variables773available for these neighborhood attributes.774Thepolicy-relevant coefficientson the four open775space land use variables in each ofthe counties also776revealed mixed results. The Haussman test for777endogeneity indicated the predicted values for the778large buffer P_LGPERMOPEN and P_LGAGFOREST779measures shouldbe used for both Calvert and Carroll780counties, while in Howard County, predictedvalues781were used for these two variables as well as for the782small buffer P_SMAGFORESTmeasure. Given the783functional form used for the hedonic estimation, the784estimated coefficients on these variables can be785interpreted as elasticities. In Calvert County, the786estimated coefficient on the large permanent open787space (p_LGPERMOPEN) is positive and statistically788significant, with an estimated elasticity of 0.71. In789Howard County, the estimated coefficients for both790the large (p_LGPERMOPEN) and small (SMPERMOPEN)791open space variables are positive and statistically792https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press793794795===== PAGE 10 (score 23, numeric-paren 6, table-words 1, var-words 14) =====79679742798April 2003799Table 6. Error Term Spatial Dependence Test800Statistics, by County801County802Test Statistic a803Calvert804Carroll805Howard806Likelihood Ratio80789.7480858.34809606.39810(0.0000)811(0.0000)812(0.0000)813Lagrange Multiplier814109.8481577.548161,093.29817(0.0000)818(0.0000)819(0.0000)820a The Moran's I test statistic was positive and significant for Calvert821and Carroll counties. It was not computationally feasible for Howard822County due to the large number ofobservations.823significant with estimated coefficients of 0.59 and8240.05, respectively. In Carroll County, these two esti-825mates are not statistically significant.826Carroll County residents could value open forest827and agricultural land less because there is a greater828amount of these lands in this county. However, in829the large buffer, Carroll County had a mean of 12%830permanently preserved open space and 53% unpre-831served agricultural and forest land around the parcels832examined. Howard County had a mean of 13%833permanently preserved open space and 36% unpre-834served agricultural and forest land. Calvert County835had only 5% permanently preserved open space,836but 51% unpreserved agricultural and forest land.837Similarly, in the small buffer, Carroll had a mean of8384% permanently preserved open space and 20%839agricultural and forest land, with corresponding840values for Howard County of3% and 14%, and for841Calvert County, 1% and 20%. Thus, while Carroll842has more open space around each parcel (65%843compared to 56% in Calvert and 49% in Howard),844it seems unlikely the differences in open space845immediately surrounding the parcels alone explain846the results.847If we consider open spaces beyond those sur-848rounding individual houses, we find there is more849agricultural land available in Carroll County. In8501997, Carroll had 178,000 acres ofagricultural land851compared to 40,000 acres in Howard and 45,000852acres in Calvert. In addition, while the rate ofhous-853ing development is increasing in Carroll County,854the county has not experiencedthe conversion rates855seen in Howard and Calvert counties during the last85625 years. Thus, Carroll County residents may have857a lower willingness to pay for open space due to a858perception that there is a sufficient quantity ofopen859space and they do not perceive it is disappearing860quickly.861Another possibility is that Carroll County fanners862employ different production practices than Howard863Agricultural and Resource Economics Review864or Calvert fanners. For example, they might start865farming earlier in the morning, have a higher density866ofanimals with the accompanying odors and insects,867spray insecticides more frequently, or employ some868other practice which neighboring residents find869objectionable. Therefore, although residents receive870some positive externalities from the presence of871nearby open space, they also receive negative exter-872nalities as well-negating the positive effects.873Additionally, we hypothesized that open space is874a luxury good and will have more value in counties875where incomes are higher. While Howard County's876median income is much higher on average than in877either of the two other counties, Carroll's average878median income of $44,000 is not much lower than879Calvert'smedianincome of$48,000. In conclusion,880further analysis is needed to explain the lack ofsig-881nificance in Carroll. While the results for Howard882County were robust, results for the Carroll and883Calvert county models show the estimated coeffi-884cients were sensitive to changes in the variables and885specification.886Interestingly, the agricultural and forest land887measures (AGFOREST) were either negative and sta-888tistically significantor not statisticallysignificant in889the three counties, suggesting individuals either do890not wish to live near these lands or are not willing891to pay a premium to do so. We remain perplexed by892this result. While permanent open space did have an893impact in two counties, open space in agricultural894or forest use that may be developed in the future was895not important or decreased the value of the prop-896erty. As mentioned above, perhaps these farms use897production practices that generate negative exter-898nalities to the neighbors. Perhaps the insecurity of899not knowing what changes might occur next to a900parcel some time in the future fosters a lack of901willingness to pay to be next to the open space.902In addition, the forests abutting these properties903might be unsightly or people might prefer to have904a view which the trees block. Deer have become an905issue in these counties as they eat homeowners'906landscape plants, cause car accidents, and increase907the incidence ofLyme disease-all ofwhich could908make forest land less desirable. However, we still909find it surprising that homeowners are not willing910to pay more to be near agricultural and forest land.911The central policy focus relates to the effect of912permanent open space on housing values. Based on913the regression results, Carroll County will have no914additional property taxes from increased preserved915agricultural land, as this land use variable had no916effect on the housing value. Consequently, we now917https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press918919920===== PAGE 11 (score 17, numeric-paren 0, table-words 2, var-words 13) =====921922Geoghegan, Lynch, and Bucholtz923focus exclusively on Calvert and Howard counties924to conduct the simulations.925Simulations926How much additional land can agricultural easement927programs finance? To answer this question, we928used the estimated elasticities from the regression929models for Howard and Calvert counties to930simulate the increase in property taxes associated931only with houses located near current agricultural932easements for a 1% increase in the neighboring933agricultural easements. As both the large and small934permanent open space measures were statistically935significant for Howard County, the larger measure936is used for the simulations."937We increase permanent open space by 1% and938calculatethe increase in tax revenue associatedwith939the increase in housing values. Taking information940on land use around preserved agricultural parcels in941Howard and Calvert, the area of low density (0.4942units per acre) and medium/high density residential943land (8 units per acre) for a one-mile (approxi-944mately 1,600 meters) radius buffer was calculated.945This area was multiplied by the relevant housing946density to compute the expected number ofhouses947within this one-mile distance from each preserved948parcel.949Using the average value ofhouses in each ofthe950counties ($134,245 for Calvert and $227,963 for951Howard), the total value ofthe expected housing in952each buffer was calculated. The increase in the953property values was then calculated using the esti-954mated elasticityassuming currentpreserved agricul-955turalland has increased by 1%. Next, this increase956in the parcel value was multiplied by the county957property tax rate ($0.89 per $100 for Calvert and958$1.04 per $100 for Howard) to determine how much959tax revenue would be generated each year by the960simulated increase in preserved agricultural land.'961Finally, using the average county easement pay-962ment over 1993-1997 ($2,855 for Calvert and9634 The small open-space buffer is part of the large open-space buffer;964thus, using both would be double-counting.965S The increased value will not be immediately evident in the assessed966value of the property on which the property tax is based. However, in967Maryland, these assessments are updated at least once every three years.968The assessments are based on the fair market value using standard969appraisal approaches. A recent transaction will be used in computing the970fair market value of that property as well as the price of adjoining and971comparableproperties. Thus, weestimate the modelusing the mostrecent972transaction price, knowing that within 0-3 years of the transaction, the973property's assessment will be comparable. The property tax rate for974Howard County is $1.044 for each $100 of assessed value, so we used a975rate of 0.01044 for ease of exposition.976Capitalization ofOpen Spaces into Housing Values 43977Table 7. SimulationResults for One-Mile Buffer978of Each. Agricultural Easement: Howard and979Calvert Counties980Description981Howard County982Calvert County983[1] Sum oftotal housing with-984in 1 mile ofeach easement98541,63198629,526987[2] Average housing price988$227,963989$134,245990[3] Expected housing value991(= [1] *[2])992$9,490,251,998993$3,963,690,359994[4] I% increase in open space995181 acres996148 acres997[5] Estimated elasticity from998spatial model9990.586910000.71181001[6] Expected housing value1002increase for a 1% increase1003in open space (= [3] *[5])1004$55,695,4421005$28,214,6181006[7] Additional property tax1007collected on increased value1008$579,2331009$251,6741010[8] Average easement price1011per acre1012$5,2741013$2,8551014[9] Additional acres ofease-1015ment that could be acquired1016(= [7]+[8])1017110 acres101888 acres1019$5,274 for Howard), the total number ofnew acres1020ofagricultural easement that could be purchased1021with the increased tax revenue from just one year of1022property tax revenue was computed. Simulation1023results are presented in table 7.1024For a 1% increase in preserved agricultural land1025(148 acres) in Calvert County, the increase in hous-1026ing values within a one-mile radius ofpreserved1027parcels generated sufficienttax revenue to purchase1028an additional 88 acres in the first year. Assuming1029no real change in housing prices and no change in1030the property tax rate, the county could increase the1031number of acres preserved by 2,640 acres in 301032years. For a 1% increase in preserved agricultural1033land in Howard County (181 acres), the increase in1034housing values within a one-mile radius of these1035preserved parcels generated sufficient tax revenue1036to purchase an additional 110 acres in one year, or1037in 30 years increased the preserved acreage by10383,300 acres.1039Conclusions1040Many agricultural land preservation programs have1041inadequate resources to preserve all the parcels the1042general public may wish protected and that agri-1043cultural landowners would be willing to enroll.1044Programs are looking for innovative funding mech-1045anisms or new sources offunds. Beyond the social1046benefits provided by permanently preserved agri-1047cultural land to the region's residents as a whole,1048https://doi.org/10.1017/S1068280500002483 Published online by Cambridge University Press1049