Do Local Landscape Patterns Affect the ... - Wiley Online Library

Journal of Agricultural Economics ¾ Volume 54, Number 1 ¾ March 2003 ¾ Pages 21-35 Ó Agricultural Economics Society

Do Local Landscape Patterns Affect the Demand for Landscape Amenities Protection? Felix Schläpfer and Nick Hanley1 (Original received Feb 2002; revision received Oct 2002; accepted Dec 2002.) Agricultural support is increasingly flowing into the maintenance and improvement of landscape quality. While variations in the demand for landscape protection have been successfully attributed to variations in socio-economic characteristics, the effect of the local landscape setting as a potential determinant of environmental preferences has received less attention. A framework is formulated that describes public support for regional landscape protection as a function of socio-economic variables and land use patterns. Models are then estimated using detailed local land use statistics and voting records from a referendum on increasing public funding for local landscape amenities protection in the Swiss canton of Zurich. The land use variables represent proportions of open landscape and landscape features that are viewed as particularly valuable for aesthetic and other reasons. Crosssectional estimation results suggest that attitudes towards public landscape protection are indeed strongly associated with the local landscape. Approval for protection programmes increased with local scarcity of open space and with the presence of high-amenity landscape features. Comparison with referendum outcomes on a national-level environmental issue suggests that the positive association with high-amenity landscape features may be partly attributable to individuals’ self-selection of residency.

1. Introduction In industrialised societies, the public good character of rural landscapes is becoming increasingly important. Landscapes near urban centres in particular are an important source of recreation opportunities, whilst landscape quality itself generates utility both to rural and urban dwellers. Efficient allocation of land between competing uses deserves attention, due both to market failure and government intervention failures. Until recently, few of the enormous public subsidies to agriculture in highly industrialised countries such as Switzerland, the UK or the US went into securing these landscape benefits (OECD, 1996; Brunstad et al., 1999); indeed, the pattern of intervention was argued to produce a loss of landscape quality. However, subsidies with undesired side-effects on production or on international trade are currently being scaled down and there is an increasing use of alternative instruments to 1 Felix Schläpfer is with the Institut für Umweltwissenschaften, University of Zürich, Switzerland. Nick Hanley is with the Department of Economics, University of Glasgow, UK. The authors would like to thank the Swiss Federal Office of Statistics, the Canton Zurich Office of Statistics, and the NHP fund administration for providing data; and Ulrich Woitek for valuable discussion. FS acknowledges support through the Royal Society (European Science Exchange Programme). We also thank two anonymous referees for their valuable comments. Address of corresponding author: Felix Schläpfer, Institut für Umweltwissenschaften, Universität Zürich, Winterthurerstrasse 190, 8057 Zürich, Switzerland. E-mail: [email protected] for correspondence.

22 Felix Schläpfer and Nick Hanley

increase the supply of landscape amenities. One way to secure these benefits is through public payments to farmers who agree to provide these landscape amenities through appropriate land management (Hanley et al., 1998). In the case of the traditional European agricultural landscapes, payments are typically linked with the management of special landscape elements such as orchards, hedgerows, flower meadows and wetlands. In order to devise efficient compensation regimes for the management of such landscape amenities, it is important to understand individuals’ preferences. Moreover, considering the fast pace of landscape change, a particularly interesting question is whether and how this change may affect the demand for recreational landscape amenities in the future. A promising approach to answering this question may be to analyse how ambient landscape settings and their recent dynamics influence individuals’ preferences for landscape protection. The objective of this study is thus to investigate the links between preferences for landscape protection programmes and aspects of local landscapes. This objective requires that information on preferences be obtained from a geographic area with sufficient variation in the relevant landscape variables. Voting data provide one source of information on these preferences. Where voting districts coincide with the spatial units at which land use statistics are available, a straightforward crosssectional approach can be pursued. We analyse voting data from a 1996 referendum on increasing public financing for the protection and management of valuable landscape elements in the canton of Zurich, Switzerland. In this instance, a relevant voting decision was coincidentally held in the same year that the data of a 12-yearly aerial photography-based land use statistic were gathered, yielding a convenient data set. We examine how local landscape patterns, as obtained from detailed land use statistics, are associated with voter support for landscape quality protection and improvements, and whether urban development and landscape change may thus predictably affect the demand for landscape quality. Our specific hypotheses are that measures for the local scarcity of (i) ‘open landscapes’ and (ii) ‘high-amenity landscape elements’ both significantly affect the support for landscape protection as evidenced in voting behaviour. While we expect a positive effect of the former scarcity, there is no unambiguous theoretical expectation or prior empirical evidence regarding the latter. Comparison with referendum outcomes on public goods issues unrelated to the local landscape is used to evaluate the issue of causality in the patterns obtained. 2. Previous Literature Research on landscape preservation has predominantly framed the issue as a question of the optimal proportional allocation of land to agricultural uses as opposed to urban use or forestry (e.g. Kline and Wichelns, 1994; Lopez et al., 1994; Brunstad et al., 1999). However, when special subsidies for the management of landscape amenities are considered, additional degrees of freedom are introduced. Subsidies can be paid in ways to induce land managers to provide or maintain special landscape features that enhance the recreational and amenity value of their land.

Do Local Landscape Patterns Affect the Demand for Landscape Amenities? 23

Analyses on whether particular land management regimes are desired by the public are therefore of substantial interest for the design of agri-environmental policies. For this reason, increasing efforts are under way to estimate the value of enhancing landscape amenities. Applications of the contingent valuation method to farm landscapes in Europe demonstrate substantial willingness to pay for maintaining amenities linked with traditional agricultural practices (Willis and Garrod, 1991; Drake, 1992; Pruckner, 1995; Hanley et al., 1999). A contingent valuation study performed in a region of the canton of Zurich (the case study area in this paper) suggests a mean willingness to pay for agricultural landscape protection in the order of SFR 240 per person per year (Roschewitz, 1999). Potential influences of residential landscape settings on individual demand have so far not been examined in studies using stated preferences. Existing studies on landscape valuation based on surveys have not allowed detailed analysis of ambient landscape effects on demand due to the limited variation (or limited comparability in the case of landscape visitor surveys) of respondents’ residential areas in terms of the landscape pattern. However, the fact that local landscape features are important is evident from studies on hedonic prices of residential locations (Garrod and Willis, 1994; Earnhart, 2001). An alternative way to investigate the determinants of preferences for landscape conservation is with voting outcomes from referenda. Preferences derived from referenda have been found by others to be consistent with the basic economic premises of individual behaviour. Deacon and Shapiro (1975) developed a framework to examine aggregate voting behaviour on public goods issues and used their model to characterise preferences for coastal zone preservation and investments in public transportation. Fischel (1979) applied the Deacon and Shapiro model to individual voter choices over air quality improvements, and found that the ex post level of environmental quality of the community had a significant influence on voter behaviour. More recently, aggregate (county-level) voting data were used to characterise the demand for various environmental goods based on 16 initiatives in California (Kahn and Matsusaka, 1997) and to analyse voter support for an initiative to promote sustainable forestry practices in Oregon (Kline and Armstrong, 2001). Although most of the above voting issues related to local or regional land use only, Kline and Wichelns (1994) included more detailed aspects of the local land use pattern among the variables used to explain voter behaviour. In a study of public support for purchasing development rights to farmland they showed that approval for maintaining open space in Pennsylvania and Rhode Island (USA) was affected by several variables related to the local environment including the recent growth of the population, proportion of land in farms, recent loss of agricultural land and recent increases in property values. They found some further evidence that the presence of broadly defined ‘sensitive resources’, including important agricultural soils, groundwater resources and wildlife habitats, increased approval, although only in their Rhode Island dataset. Apparently, the local environment had influenced the perception of the land preservation issue. In contrast to Kline and Wichelns (1994), our data allow an analysis of the support for qualitative improvements of the landscape. Kline and Wichelns’ results are relevant for


agencies concerned with land zoning issues, whereas the present study focuses on preferences more relevant for the design of agri-environmental incentive schemes or small-scale nature reserves targeted at increasing the recreational benefits of the landscape. 3. Choice of Conceptual and Analytical Framework In the case of environmental pollution, the demand for pollution control as a public good is typically conceived as a function of actual levels of pollution exposure (e.g., Poe and Bishop, 1999). There is much reason to expect that demand for positive externalities should be similarly dependent on perceived local abundance or scarcity of the public good. Thus, the benefits of ‘landscape amenities protection’ should be dependent on the perceived local abundance or scarcity of landscape quality. Support for landscape protection measures could thus be framed as a function of land use descriptors related to the quantity and quality of undeveloped local landscapes, along with socio-economic variables such as income. Choosing appropriate indicators of the state and recent dynamics of the landscape that may affect individuals’ demand for enhanced landscape services is not, however, straightforward. The issue of how important quantitative and qualitative attributes of the landscape should be measured is not well understood (Deller et al., 2001). In particular, perceived qualitative aspects of the local landscape which may influence citizens’ demand for landscape protection or improvement are likely to depend on complex spatial patterns that are difficult to capture in variables constructed from land use statistics. Given a limited theoretical background for model specification, we decided a priori to consider land use variables that relate to (a) the state and (b) the recent dynamics of variables for ‘open space’ and ‘landscape quality’. Assuming that ‘open space’ may not be a pure public good in the sense of non-rivalry in consumption, we chose to construct a variable for ‘open space per capita’ instead. Qualitative aspects of the open landscape are hypothesised to be measurable by the extent of selected landscape features, such as high-growth orchards, hedgerows, and wetlands, that are perceived to enrich the landscape for aesthetic, ecological and other reasons. Using the basic framework developed by Deacon and Shapiro (1975), we take the perspective that an individual voter has a utility function with a vector of public goods and a vector of private goods as arguments. The level of collectively chosen public goods determines the individual’s tax liability for the provision of these goods. The individual’s vector of private goods is subject to a budget constraint that is equal to the difference between income and tax liability. The individual reaches a voting decision in his or her perceived self-interest by comparing attainable utilities under alternative policies. In a general model of voting, the individual would weight the expected benefit from voting with the costs of going to the polls at all. In such a model the voting turnout itself could theoretically yield important information on the demand for the public good in question. In the context of the referendum data analysed here, such additional information is not useful due to the presence of several other issues in the same ballot. Deacon and Shapiro estimated two joint equations, one describing


approval of the voting proposition and the other describing voter participation among those going to the polls. Unlike in the Deacon/Shapiro referenda, which were held together with presidential elections, there were only a few citizens at the polls not casting a vote regarding the nature and heritage protection issue analysed here. Thus the present analysis, like that in Kline and Wichelns (1994), takes a single equation approach. A conceptual weakness in many regression studies concerns the issue of unclear causalities between significantly correlated variables. In the present case it could be argued that citizens with particularly high preferences for environmental quality tend to live in places with high landscape quality, which would render the direction of causality ambiguous. To allow for this possibility we also apply the model to data from votes on other public goods which are not related to the local landscape. If the state of the local landscape, and not the self-selection of residence, produces the observed pattern in voting outcomes, this pattern should be unique to the landscape protection referendum. When, however, individuals with high willingness to pay for environmental quality tend to take residence preferentially in municipalities with high amenity values, then these municipalities are likely to display particularly high approval for environmental expenditures in voting issues with no conceivable connection to the local landscape state. Recent voting decisions within the same constituency on national energy taxes and insurance schemes for pensioners are used for this comparison. These referenda were selected (1) to concern (perceived) additional expenditures and (2) to be strongly controversial in terms of nearly equal numbers of ‘yes’ and ‘no’ votes to ensure sufficient voter interest and information. 4. Empirical Analysis Our empirical analysis combines municipality-level land use and income statistics with municipality-level voting records for a referendum on landscape protection expenditures held in 1996 in the canton of Zurich, Switzerland. In the canton of Zurich, there were, in 1996, roughly ten thousand hectares (or five per cent of the land) allocated to managed landscape elements which have particular importance for landscape amenities according to our definitions below. This total area was composed of thousands of mostly very small parcels of orchards, hedgerows, wetlands, and other landscape features. From 1984 to 1996 this total area had declined by 11 per cent. To reverse this trend, many of these landscape features are today eligible for compensation payments to land managers (who may be farmers or nature protection societies). According to the significance of the features, payments have the character of compensations for mandatory land use restrictions in the case of some 800 small nature reserves, or of contractual cost-sharing incentives to farmers for a total of roughly 400 hectares of hedgerows, orchards, and meadows. The latter payments are a ‘bonus’ on particularly valuable amenities supplementing the national environmental scheme of the Federal Office of Agriculture. The payments flow from a cantonal Fund for Nature and Heritage Protection established in 1974 (Parliament of the Canton of Zurich, 1974; Executive Council, 1977; Canton of Zurich, 1997). There have been several national referenda on agricultural policy reform in Switzerland in the recent years, in which the future management of the landscape


by farmers has been an important issue. However, consent to or rejection of these proposals is difficult to interpret because important landscape-preservation objectives were coupled with regional policy objectives and equity issues. The best information on public demand for landscape protection should therefore be sought in referenda that are not, or at least not obviously, linked with farm income. The 1996 referendum on increasing the public funding for nature and heritage protection in the canton of Zurich meets this requirement. The modest payments from these funds contribute only a little to total farmer incomes, a fact best highlighted by the farming union’s disapproval of the voting proposition, which reflected the critical attitude of the farmers towards subsidies that are paid explicitly for landscape benefits. An amendment to the canton of Zurich’s ‘act on the financing of measures for nature and heritage protection and for recreation areas’ (NHP finance act) was submitted to the voters (and adopted with 57 per cent approval) in September 1996 (Office of the Parliament, 1996). The proposition envisaged increasing the annual instalments into the public (cantonal) Fund for Nature and Heritage Protection from SFR 10-20 million to SFR 20-30 million. Further, the amendment empowered the parliament to decide on an additional yearly instalment of SFR 10 million into the fund to pay off debts. In the NHP finance act the NHP fund is described as follows (Parliament of the Canton of Zurich, 1974): i) A fund is created for the financing of measures to create, maintain, enable access to, improve or manage land and townscapes that are worthy of protection, natural and cultural objects and recreation areas. ii) The funds are designated (a) for the purchase of properties and the establishment of restricted property rights for the objectives of (i); (b) for the compensation of land owners who are for these objectives burdened with restrictions similar to expropriation, (c) for the financing of other measures according to (i), such as the management of objects, the construction of walking trails, recreation areas, and the like, as far as there are no other available sources for financing. In the summary of the roughly 1500-word text of the voter information booklet the change to the act was justified by ‘increasing expenditures for the maintenance, restoration and management of the objects of nature and heritage protection’ (Executive Council, 1996). The allocation of the fund’s money is budgeted by the administration and controlled by the parliament. Precise regional allocation of the additional funding was not specified in the information available to the voters. However, because the basic objective was to continue the existing policy, voters could perhaps expect that municipalities with a large endowment of natural and cultural amenities would attract most of the additional funds. By 82 votes in favour and 74 against, the cantonal parliament recommended adopting the proposition. The expected annual income tax increase due to the referendum for a taxpayer with average income amounted to an estimated SFR 12-24 , or £ 5-10. 4.1 Data and Analysis Since individual ballot decisions are not disclosed, only the aggregate responses of each municipal constituency are observed. As in Deacon and Shapiro (1975) and in


Kline and Wichelns (1994), the dependent variable we use is a logit transformation of P(yes), the percentage of votes approving the referendum:

Logit ( yes ) = ln

P ( yes ) 1 - P ( yes )

Regression coefficients thus describe the change in the natural logarithm of the ratio of the per cent of voters approving the referendum to the per cent rejecting the referendum, given a one-unit change in the value of an explanatory variable. The voting aggregates used in the empirical analysis were the 171 political municipalities of the canton of Zurich. Data provided by the cantonal office of statistics were used to calculate the percentage of approving votes among valid votes. Census data were used to construct relevant socio-economic variables. Observations on average net income of taxpayers in the municipalities in 1996 were obtained from the cantonal office of statistics. The percentage of the population in different age segments and a variable for educational achievement were both strongly correlated with income (r > 0.55) and were thus dropped from the analysis. The landscape descriptors are derived from the aerial photography-based, municipalitylevel Swiss area statistics. In the canton of Zurich, these data were gathered in the years 1982/1984 and 1994/1996 using a 100m by 100m grid sampling procedure (FSO, 1979/85 and 1992/97). As our variable for ‘open space per capita’ we used the proportion of cultivated land in a municipality, divided by municipal population size in 1996 (OPENSPACE). This is suggested to be an adequate measure for the remaining open landscape since it comprises virtually all non-forested and non-urban land in the canton of Zurich2. Changes in the proportion of cultivated land between 1984 and 1996 were strongly correlated with the 1996 proportions (r = obvious problems for interpreting regression results. Since the explanatory power of OPENSPACE was very high, and indeed much larger than could reasonably be expected from an effect of ‘recent change’ in this variable, the variable for ‘recent change’ was omitted from the regression. Municipality-level land area in managed landscape features with high amenity value (AMENITY) was derived from the detailed version of Swiss area statistics (FSO, 1979/85 and 1992/97). The variable represents the summed area with each municipality of 15 out of 74 distinct land use categories. These categories are (with hectares for 1996): orchards (row and scattered) (2460), hedgerows (1905), groups of trees (796), small woods (521), shrubs (80), wetlands (1247), shore vegetation (75), thin forest on unproductive land (27), thin forest on agricultural land (21), forest strips (1818), meadows and pastures in early stage of succession (31), alpine pastures (158), herbal vegetation (51), and rocks, sand, or rubble (24). These 2

Cultivated land includes the categories ‘fruit trees, vinyards and horticulture’; ‘meadows, fields and pastures’; and ‘alpine pastures’ in the coarse grained version of the area statistics (distinguishing 15 land use variables).


landscape features with high amenity value sum to 9214 hectares or 5.3 per cent of the total land area. The related dynamic measure (DAMENITY) was defined as the (net) relative change of AMENITY over the period between 1984 and 1996. State and dynamics of AMENITY were much less correlated (r = analogues for ‘open space’. Thus both variables could be included in the model. An additional variable for the presence of local architectural heritage was included to account for the fact that historical architectural heritage, as an integral part of the rural landscapes, is also supported from the NHP fund. The cantonal inventory of townscapes ‘of supra-municipal significance’ yielded a dummy variable indicating whether a municipality has a particularly valuable visible historical heritage (HERITAGE). All explanatory variables were log-transformed, substantially improving model fit, except for DAMENITY and HERITAGE, where zero and negative values prevented log-transformation. 5. Results Table 1 sets out our expectations regarding coefficient signs. Although environmental quality is typically expected to be a normal good, the expected coefficient sign for income is ambiguous, since the expected tax increase is higher for the higher income groups. Due to decreasing marginal utility, the quantitative static measures of open landscape (OPENSPACE) should negatively influence demand. With the measures for landscape and townscape quality (AMENITY, HERITAGE) the sign expectations are ambiguous. On one hand, there may be an effect of decreasing marginal utility when more land is in high-quality landscape elements. Alternatively, the presence of high quality landscapes may for some people initially establish or increase the perception of benefits from these public goods. A rather different interpretation of any positive association of AMENITY with approval for landscape protection expenditures may be due to self-selection of residency if the same pattern is also evident in votes on other, non-local, public goods. Change (reduction) in high quality landscape elements (DAMENITY) may be relatively difficult to perceive but would be expected to have positive effects on voting decisions. Table 1. Descriptive Statistics of the (Untransformed) Independent Variables Variable INCOME OPENSPACE AMENITY DAMENITY HERITAGE

Description

Mean (SD)

Sign expectation

Average net income of taxpayers in 1000 CHF Area of open (cultivated) land per capita (ha) Percentage of near-natural and traditional landscape elements area on non-forest land Percentage change (reduction) of AMENITY over 12-year period Dummy variable for historical heritage townscape: 1 = present; 0 = absent

58.7 (11.8) 0.228 (0.224) 7.6 (3.8)

?

Note: There are 171 observations in the dataset.

12.7 (11.5) 0.46

– ? (+) ?


All independent variables described in Table 1 were incorporated in a regression equation to explain logit(yes). To account for heteroscedasticity we used White’s robust covariance matrix (Greene 1997) in calculating the standard errors. Regression results are shown in Table 2. As may be seen, 60 per cent of the variability in voter approval for NHP is explained by the model. In the case of OPENSPACE, a negative coefficient was obtained, suggesting a diminishing marginal utility effect. The effect of INCOME was positive and highly significant. With the progressive tax system taken into consideration, this suggests that landscape protection in the canton of Zurich is a luxury good. A scatter plot of mean taxpayer income and approval rates in the individual municipalities is provided in Figure 1. But income variability alone accounted for only 18 per cent of the variability in voter approval. A much larger proportion of the variation could be attributed to the landscape variables, where OPENSPACE in particular turned out to be highly influential. The effect of local landscape amenities (AMENITY) was positive and also highly significant. Thus, there was no positive effect of scarcity of high amenity landscape features on the demand for their protection and increase. On the contrary, a possible effect of decreasing marginal utility from additional landscape protection appeared to be outweighed by a generally positive effect these landscape elements may have on the perception of the landscape protection issue or alternatively on the choice of residency by individuals who value them highly. It is not possible to decide between these two alternative explanations based on the voting records for the NHP issue alone. The positive tendency observed for the dummy variable HERITAGE for aesthetically significant townscapes can be similarly interpreted. The change of high-amenity landscape elements (DAMENITY) was not significant in explaining voter behaviour, but this may be due to problems for people in perceiving what are, after all, relatively small changes in the local landscape. Table 2. OLS Estimates for the Nature and Heritage Proposal Dependent variable: logit of percentage of votes in favour of proposal Variable Intercept log INCOME log OPENSPACE log AMENITY DAMENITY HERITAGE Adjusted R2 Partial R2 of INCOME

a

Coefficient estimate -4.767 0.354 -0.213 0.141 0.002 0.068

t-Statistica -4.480*** 3.671*** -14.146*** 3.624*** 0.972 1.684(*)

0.60 0.18

: Statistic based on White’s corrected variance matrix Note: (*),*,**,***: Significant at 0.1, 0.05, 0.01, and 0.001 levels, respectively.


Figure 1. Mean Net Income and Percentage Approval of the Nature and Heritage Proposition 70

Percentage approval

65 60 55 50 45 40 35 30 25 20 40

60

80

100

120

Mean net income (1000 SFR)

5.1. Comparison with referenda on other public goods We might anticipate that local environmental variables, and changes in these, would have a greater influence on preferences for local environmental goods, such as with the nature and heritage fund, than would be the case in voting over a national or even global environmental good. As an example of voting on such a (non-local) public good, we use results from a referendum proposing a national tax on nonrenewable energy use3. The left part of Table 3 shows results for this referendum, using the same independent variables as in Table 2. Table 3. OLS Estimates for Energy Tax and Old Age Insurance Proposals Dependent variable: logit of percentage of votes in favour of proposal Non-renewable energy proposal (national environmental public good) Variable Intercept log INCOME log OPENSPACE log AMENITY DAMENITY HERITAGE Adjusted R2 Partial R2 of INCOME

Coefficient estimate -2.282 0.156 -0.068 0.100 0.001 0.097

t-Statistica -2.055* 1.549 -3.749*** 2.785** 0.628 2.620**

0.18 0.04

Insurance proposal (nonenvironmental public good) Coefficient estimate

t-Statistica

-9.487 0.837 -0.037 0.035 -0.000 0.016

-7.629*** 7.260*** -2.445* 1.117 -0.330 0.608

0.48 0.45

a

: Statistic based on White’s corrected variance matrix Note: *, **, ***: See Table 2.

3

The proposition envisaged raising a refundable tax on non-renewable energy use to protect environment, landscape, and climate. Voter turnout in the canton of Zurich was 47 per cent. The approval rate was 52 per cent. This and the following issue were both voted on in September 2000.


As may be seen from the coefficient estimates 4, levels of open space are negatively associated, and local environmental amenities, and historical heritage are positively associated with approval in the vote on a non-renewable energy tax, although effects, except for the historical heritage, were smaller. The persisting significance of OPENSPACE is not surprising given a generally higher awareness of environmental issues in more urbanised areas. Regarding the AMENITY and HERITAGE variables, however, there is little theoretical support for a direct causal relationship with approval for the energy tax proposition. As there was, further, no correlation of AMENITY with OPENSPACE (r = positive association of AMENITY and HERITAGE with approval for landscape amenities protection was likewise not directly causal. In other words, high approval rates may not reflect a possible expectation that high-amenity and high-heritage municipalities would attract more of the funds. Rather, the association may be an effect of self-selected residence in high-amenity landscapes and townscapes by individuals with a particularly high demand for environmental quality and hence public environmental expenditure in general. Finally, for illustrative purposes, we compare the results from the environmental public decisions with those for a referendum on a non-environmental public good, namely a proposal to reduce the cantonal supplements to the national old age insurance (Table 3, right)5. As would be expected, income had a very strong effect on voting patterns. OPENSPACE was still (weakly) significant, which may be the net effects of urban-rural differences in a variety of factors including age distribution. As would be expected given the ‘self-selection of residence’ hypothesis, the landscape and townscape features, unlike in the environment-related voting decisions, were no longer significantly associated with voting outcomes. 6. Discussion and Conclusions In this paper we have presented evidence that the support for public efforts to protect landscape amenities is conditional not only on socio-economic characteristics but also on the current state of the landscape itself. The unusual combination of municipality-level voting data on landscape protection expenditures and fine-grained land-use statistics allowed us to include, for the first time, qualitative aspects of the landscape as explanatory variables in a model explaining the demand for landscape protection. Our main result is that per capita availability of open space alone explained a very large part of cross-sectional variation in the demand for increased landscape protection, as measured by voting behaviour. The effect indicates that scarcity of open space in one’s locality increases demand for landscape protection. We also found that wealth in landscape amenities was positively associated with support for increased public spending on landscape protection. Higher incomes also increased the approval rate for the NHP proposition, with an income elasticity of voting

4

5

Although, due to the logit transformation, coefficients are not the marginal effects they can be directly compared across models because of similar average approval in the three votes. Voter turnout was 47 per cent. The rejection rate (favouring higher expenditures) was 56 per cent.


approval greater than one6. Similar results in terms of explanatory patterns in the canton of Zurich and in Kline and Wichelns’ (1994) study indicate that voter motives may be comparable in spite of referendum propositions relating to landscape amenities protection in one case and agricultural land preservation in the other. In particular, the negative coefficient of OPENSPACE in Zurich is consistent with a negative coefficient of %FARMLAND (in the Pennsylvania model). The positive association of approval rates with the AMENITY variable in Zurich seems to be consistent with the positive coefficient of %RESOURCES, which was Kline and Wichelns’ variable for ‘sensitive resources’ in their Rhode Island model. Although urban-rural differences in the perception of many environmental issues are well known, these differences seem to be particularly pronounced in the case of landscape amenities protection. The identified pattern of support for improving landscapes at the cantonal level implies that the areas with most landscape amenities and those with most of their beneficiaries are geographically very distinct. This finding suggests that local decision-making about land use issues, such as municipal zoning, may usually not be economically optimal from a cantonal perspective. Cantonal decision-making and the appropriate design of incentive schemes to gain local support for landscape protection deserve high attention. As is well known, use of aggregated cross section data requires great care in econometric specification to yield interpretable results. To avoid ambiguities in attributing effects, we limited the number of independents variables to a small set with little pairwise correlation. The obvious cost of such a procedure was that we could not, for instance, obtain a separate effect of recent changes in ‘open landscape’ on voting behaviour. By comparing voting patterns with those in other selected voting issues, we devised a way to address the issue of causality in the regression results. Comparison with voting outcomes in a non-local environmental issue strongly suggests that the positive association between local landscape quality and the support for increased landscape protection expenditures may be partly attributable to a generally higher demand for environmental quality in municipalities with high landscape quality (the selection effect). Results demonstrate that amenity characteristics of cultivated landscapes can be organised into consistent and meaningful empirical measures that move beyond a mere distinction of developed versus undeveloped land. Given the continued pace of land development in highly urbanised regions, our results suggest that the demand for landscape protection will increase in the near future as the local public good ‘open landscape’ gets scarcer. Our results suggest that a benefits function approach accounting for citizens’ residential landscape setting could help to derive future demand functions which could be useful in setting priorities for landscape preservation and zoning.

6

Marginal income taxes in the canton of Zurich range from zero to thirteen per cent across incomes. At the average income, a one per cent increase of income affects income tax by 1.57 per cent. From regression results for referendum approval elasticity for the landscape good must therefore be well above 1.57.


References Brunstad, R.J., Gaasland, I., and Vardal, E. (1999). Agricultural production and the optimal level of landscape preservation. Land Economics 75(4):538–546. Canton of Zurich (1997). Staatsrechnung des Kantons Zürich 1997. Verwaltungsrechnung. Zurich, Switzerland: Cantonal Printing Office. Deacon, R. and Shapiro, P. (1975). Private preference for collective goods revealed through voting on referenda. American Economic Review 64:943–955. Deller, S., Tsai, Tsung-Hsiu, Marcouiller, D.W., and English, D.B.K. (2001). The role of amenities and quality of life in rural economic growth. American Journal of Agricultural Economics 83(2):352–365 Drake, L. (1992). The non-market value of the Swedish agricultural landscape. European Review of Agricultural Economics 19:351–364. Earnhart, D. (2001). Combining revealed and stated preference methods to value environmental amenities at residential locations. Land Economics 77(1):12–29. Executive Council (1977). Verordnung über den Natur- und Heimatschutz und über die kommunalen Erholungsflächen. Offizielle Sammlung 46:779. Executive Council (1996). Änderung des Gesetzes über die Finanzierung von Massnahmen für den Natur- und Heimatschutz und für Erholungsgebiete. Beleuchtender Bericht des Regierungsrates. Kantonale Volksabstimmung, Issue “Abstimmungsvorlagen Kanton Zürich 22. September 1996”, August 1996, pp. 4–5. Fischel, W. (1979). Determinants of voting on environmental quality: a study of a New Hampshire pulp mill referendum. Journal of Environmental Economics and Management 6:107–118. FSO (1979/85 and 1992/97). Land use statistics 1979/85 and 1992/97. GEOSTAT. Neuchâtel, Switzerland: Federal Statistical Office. Garrod, G. and Willis, K. (1994). An economic estimate of the effect of a waterside location on property values. Environmental and Resource Economics 4(2):209–217. Greene, W. (1997). Econometric Analysis. 3rd Edition, Englewood Cliffs, New Jersey: Prentice Hall. Hanley, N.D., Kirkpatrick, H., Simpson, I. and Oglethorpe, D. (1998). Principles for the provision of public goods from agriculture: modelling moorland conservation in Scotland. Land Economics 74(1):102–113. Hanley, N.D., Whitby, M., and Simpson, I. (1999). Assessing the success of agrienvironmental policy in the UK. Land Use Policy 16(2):67–80 Kahn, M.E. and Matsusaka, J. G. (1997). Demand for environmental goods: evidence for voting patterns on California initiatives. Journal of Law and Economics 15:137–173. Kline, J. and Armstrong, C. (2001) Autopsy of a forestry ballot initiative. Characterizing voter support for Oregon’s Measure 64. Journal of Forestry 99(5):20–29. Kline, J. and Wichelns, D. (1994). Using referendum data to characterize public support for purchasing development rights to farmland. Land Economics 70:223–233. Lopez, R.A., Shah, F.A., and Altobello, M.A. (1994) Amenity benefits and the optimal allocation of land. Land Economics 70(1):53–62. OECD (1996). Agricultural Policies, Markets, and Evaluation 1996. Paris: Organization for Economic Co-operation and Development. Office of the Parliament (1996). Beschluss des Büros des Kantonsrates über die Ergebnisse der kantonalen Volksabstimmung vom 22. September 1996. Zürcher Amtsblatt 1996 (4 October):1641–1668. Parliament of the Canton of Zurich (1974). Gesetz über die Finanzierung von Massnahmen für den Natur- und Heimatschutz und für Erholungsgebiete. Offizielle Sammlung 45, 61.


Poe, L.G. and Bishop, R.C. (1999). Valuing the incremental benefits of groundwater protection when exposure levels are known. Environmental and Resource Economics 13:341–367. Pruckner, G. (1995). Agricultural landscape conservation in Austria: an application of the CVM. European Review of Agricultural Economics 22:173–190. Roschewitz, A. (1999). Der Monetäre Wert der Kulturlandschaft. Eine Contingent Valuation Studie. Kiel: Wissenschaftsverlag Vauk Kiel KG. Willis, K.G. and Garrod, G. (1991). Landscape values: A contingent valuation approach and case study of the Yorkshire Dales National Park. ESCR Countryside Change Initiative. Working Paper 21.