The Value of Naturalness of Urban Green Spaces: Evidence from a Discrete Choice Experiment

1. Introduction

Globally, biodiversity change is accelerating (IPBES 2019), and concerns about biodiversity loss are widespread. In recent decades, cities with their fragmented structure of gardens and parks have received increasing attention as places for biodiversity conservation (Lepczyk et al. 2017; Staude et al. 2021). Urban green spaces (UGSs) provide a wide range of benefits for humans; for example, for health and recreation (Aronson et al. 2017). Moreover, UGSs are an important component of urban sustainability and can affect cities in several ways; for example, by improving air quality through filtration of polluted air (Janhäll 2015) or by offering shade and cooling (Dimoudi and Nikolopoulou 2003).

In the past decades, several stated preference studies estimated a positive willingness to pay (WTP) for UGSs close to the place of residence (Del Saz Salazar and García Menéndez 2007; Bernath and Roschewitz 2008; Bullock 2008). For instance, Tu, Abildtrup, and Garcia (2016) used choice experiments (CE) to investigate preference heterogeneity for distance to peri-urban forests and parks in France. Results show that the WTP for having UGSs in the vicinity are heterogeneous among income classes and private garden ownership. The authors found that tenants have a particularly high WTP for living close to a UGS, whereas having a house with a private garden may be a substitute for being close to a UGS.

However, little is known whether and to what extent people value and are willing to pay for the naturalness of UGSs and the urban biodiversity maintained by UGSs. The few existing studies found mixed results. The results of a CE in Dublin, Ireland, conducted by Bullock (2008) suggested that naturalness is considered more important in larger UGSs than in smaller ones. However, Stessens et al. (2020) pointed out that the naturalness of UGSs is perceived as a less important quality factor in Brussels, Belgium. Giergiczny and Kronenberg (2014 used a CE to calculate the value of street trees in Poland and suggest how the findings would improve UGSs. They found that respondents indeed have a WTP for increasing the number of street trees. Hwang et al. (2019) elicited preferences for natural growth in UGSs in Singapore. They used binominal logistic regression models to show that people prefer wild and natural urban environments.

However, neither revealed nor stated preference studies so far have identified and quantified the value of naturalness of UGSs. Using survey data, this study investigates the WTP for the naturalness of UGSs and walking distance to them in Germany using a discrete choice experiment (DCE). Of interest is the relative valuation of the naturalness of UGSs and the walking distance to them, as visitors rate the naturalness of urban parks, which describes biodiversity-related characteristics (e.g., plant species richness and animal richness) as one of the most crucial attributes of park characteristics besides cleanliness and low crime level (Bertram and Rehdanz 2015). For urban planners, it is crucial to know if values are universal or if they systematically differ between cities.

With our study, we aim to give a better understanding of preferences for natural and biodiverse UGSs. Thus, to the best of our knowledge, we provide the first DCE study estimating the WTP for biodiverse UGSs using a sample of 22 large cities across Germany. We hypothesize that Germans have a higher preference for a more natural and biodiverse UGS compared with a UGS with less natural elements. However, this is not clear a priori; biodiversity can also be perceived as “bad,” for example, because naturalness can be associated with untidiness or because it comes at the cost of lost space in parks for alternative uses such as sport. The results of the study are particularly relevant because almost three-quarters of the EU population lives in urban areas (Eurostat 2016), and cities might therefore be the place where most people experience biodiversity. Moreover, we hypothesize that people have a higher WTP for a short walking distance to a UGS.

To address our research questions, we designed a scale on which participants are asked to subjectively assess the biodiversity of their closest UGS that they use most often. We operationalize perceived biodiversity by drawing on the term “naturalness” or “nearness to nature” previously used in surveys (Bertram and Rehdanz 2015) and measure it with a graphical five-point Likert scale. Because biodiversity is a complex concept with which most Germans are not familiar, we use naturalness as a bridging concept.

Stated preference studies have to find an appropriate payment vehicle in a hypothetical market. Here we use changes to housing rent as the payment vehicle, as numerous hedonic pricing studies have shown that changes in UGS characteristics affect the housing market, making this a credible payment vehicle with which participants are familiar.

Our study reveals that the mean monthly WTP for naturalness in UGSs is €20.25 per month for a one-step increase of naturalness on the five-point scale that ranges from “hardly natural” to “very natural.” Moreover, the mean respondent has a negative monthly WTP of −€2.47 per month for an additional minute of walking distance to the closest UGS. However, we find higher negative WTPs for respondents with a short walking distance (−€5.79 per month) compared with people with a walking distance of more than 15 minutes (−€1.74 per month). The mean WTP values vary between cities. The mean WTP for naturalness range from €12.44 per month in Dresden to €35.88 per month in Bremen. The highest WTP for walking distance, in absolute value, is in Bremen (−€3.62 per month), and respondents from Dresden have the lowest WTP (−€1.64 per month). One possible explanation is that people sort according to their preferences for local public goods and move to the city that offers the bundle of amenities that best suits their preferences, as suggested by Tiebout (1956). People with a high WTP for natural green space would tend to move to a city that offers a lot of natural green space. Based on that theory, we expect a positive correlation between the WTP for natural green space and the amount of natural green space in the city.

Figure 1

Graphical Likert Scale for the Naturalness of the Closest and Most Often Used Urban Green Spaces

2. Experimental Design

DCEs have become a standard method of revealing determinants of people’s behavior and investigating the WTP for specific attributes. From June 16, 2020, to June 29, 2020, as well as from July 20, 2020, to July 28, 2020, we conducted an online survey to elicit preferences for access to and the naturalness of UGSs. The survey addressed respondents who rent a flat in the capitals of the 16 federal states of Germany (Berlin, Bremen, Dresden, Düsseldorf, Erfurt, Hamburg, Hanover, Kiel, Magdeburg, Mainz, Munich, Potsdam, Saarbrucken, Schwerin, Stuttgart, and Wiesbaden) and six additonal major German cities (Cologne, Dortmund, Essen, Frankfurt, Leipzig, and Nuremberg). We ran an additional survey in 14 of the initially 22 cities from November 16, 2021, to January 17, 2022, to investigate the preferences at the city level in more detail.

The survey consisted of four parts. First, we asked questions regarding the housing situation of the respondents. Then we asked questions about the perceptions and attitudes toward biodiversity in the neighborhood and their use of UGSs. The third part asked (incentivized) the respondent to upload a photo of the immediate neighborhood of the flat and contains a DCE. In the last part of our survey, we collected the sociodemographic characteristics and personality traits of the respondents.

Johnston et al. (2017) note that the valuation scenario should be seen as credible by respondents to derive valid value estimates from stated preference surveys. Consequently, the design was discussed in three independent two-hour online focus group discussions with five participants to improve our questionnaire. A professional moderator from a marketing agency facilitated the focus group discussions. The survey was revised according to the feedback. One issue of particular importance in the focus group discussions was to gauge our graphical representation of the naturalness of the UGS, shown in Figure 1.1 In addition, the survey was pretested from April 15, 2020, to April 21, 2020 with 520 participants, of which 264 respondents answered the pretest completely. We used this pretest sample to assess the suitability of our survey, assess the comprehensibility of the questions, and design the choice sets.

A total of 17,109 respondents were invited to participate in the online surveys by a marketing agency, of which 5,533 respondents answered the survey completely.2 Among the remaining, we excluded responses with implausible answers and obvious misstatements.3 This procedure left us a total of 4,913 responses, which were included in the analyses. The initial sample was commissioned to be nationally representative for age (18–70), gender, and income, with some deviations of the final sample from national figures (see Appendix Table A1). On average, the survey required approximately 17 minutes to complete (median 13 minutes). The number of respondents in each city included in the survey as well as the spatial distribution in Germany is presented in Appendix Figure A1.

In the DCE, the respondents were asked to consider that the closest UGS that they state they use most often will be restructured in terms of naturalness, and that the walking distance to it will be changed by modifying roads or walks. In the following, we refer to this UGS by using the term “closest UGS.” The cost of the rebuilding was supposed to be charged through the monthly rental payment, which can result in additional costs or savings. In the DCE, the participants had the choice between two alternative programs for rebuilding their closest UGS and their current situation (status quo). The attribute levels for the status quo were computed in the online survey based on the respondents’ answers to previous questions in the questionnaire, where they were asked to indicate their monthly rental and additional payments for the flat as well as the walking distance from their flat to the closest UGS.

The respondents were also asked to assess the naturalness of the closest UGS on a five-point Likert scale, ranging from “hardly natural” to “very natural,” which we designed for our survey. The Likert scale was described graphically, as shown in Figure 1. It depicts five iconic states of a UGS, ranging from a sealed playground with non-native plants and artificial light at the lower end to a pond with diverse vegetation and close to wilderness at the upper end. At the latter, several insect and bird species are visible, including birds, such as the kingfisher, that require a habitat in a good ecological state. The graphical scale was originally developed by several expert panel meetings of biologists and economists to reflect an increasing index of biodiversity and ecosystem services (e.g., water purification and carbon sequestration). It was validated and slightly revised in the focus group discussions, where respondents were asked the following questions: (1) What do you understand by the term “naturalness”? (2) What characteristics does a near-natural green space have for you? (3) Rank the five illustrations by the degree of naturalness.

The focus group discussions and our pretest confirmed that our baseline (status quo) condition as well as the proposed change(s) relative to the status quo were well and consistently understood and viewed as credible by the respondents.

In our second survey, we asked respondents to mark their most often used UGS on a map. This enabled us to detect who is using which UGS and test how different respondents perceive the same UGS. We found a high correlation of the perceived naturalness for the same UGS among the respondents. Thus, we assume that the perception of a respective UGS is homogeneous among the respondents.We calculated a highly significant (p < 0.01) Cramér’s V coefficient of 0.46 for the perception of naturalness of respondents in the same UGS if four or more respondents use this UGS.

The status quo is defined using the information provided by the participants in the survey. For the other alternatives, we define levels of the attributes, which can be found in Table 1.

The levels of the attribute walking distance were calculated following Kolbe and Wüstemann (2015), who estimate a mean distance to UGS in German cities of 300 m, with a standard deviation (SD) of 300 m. Other studies also suggest that 300 m is an appropriate buffer zone for a UGS (Kong, Yin, and Nakagoshi 2007; Liebelt, Bartke, and Schwarz 2018; Grunewald, Richter, and Behnisch 2019).4 In one survey question, we ask the respondents to state their walking distance to the closest UGS in minutes. We define a 100% change from this walking distance as 300 m (1 SD). Changes in the original rental payments are derived from previous hedonic price studies, examining the price premiums for the distance to the closest UGS on the rent (Kolbe and Wüstemann 2015; Schläpfer et al. 2015; Liebelt, Bartke, and Schwarz 2018).

For the two programs and the status quo, the levels for the attribute naturalness were described graphically, as shown in Figure 1.

Once the attributes and levels were determined, they were combined into choice sets. A full factorial design included 150 profiles. Because not all alternative choices are equally informative, we selected a subset of 30 choice sets using a fractional factorial Bayesian D-optimal design computed using the NGene software. To build the choice sets for the final experiments, a multinomial logit model was estimated using the pretest data. The estimates served as priors to generate 30 choice sets, creating an efficient design to maximize the D-efficiency measure. The final design had a D-error of 0.051. Following Loureiro and Umberger (2007), the 30 choice sets were randomly allocated between the respondents to mitigate any potential ordering effects. Each respondent was faced with 10 choice sets. Figure 2 shows an example of a choice set, in which the status quo shows the average attribute levels in the status quo over all respondents.

Descriptive analyses show that of the 4,913 respondents, 317 (6%) respondents always accepted one of the rebuilding schemes, 1,371 (28%) never accepted a scheme, and a majority of 3,225 (66%) decided selectively.

3. Data Description

The median respondent is a 40-year-old female, who is married and lives in a two-person household without children. She has an academic degree, works full-time with an average of 38 hours a week, and has a monthly net income of €2,252. The descriptive statistics of the socioeconomic characteristics are shown in Appendix Table A2.

In the second part of the survey, we asked questions regarding the closest UGS as well as perception questions related to the biodiversity of this UGS. We are interested in whether the two public environmental amenities, namely the access to the UGS and the degree of UGS biodiversity, have some value for the respondents and, if so, to what extent.

The average walking distance from the rented flat to the closest UGS is 11.50 minutes (Table 2). Most of the respondents visit this UGS weekly (38%), followed by daily (31%) and monthly (22%). Just 3% of the respondents stated that they visit this UGS yearly or never. Many (43%) indicated that they stay there between 31 and 60 minutes, followed by some who stay there less than 30 minutes (27%). Scored on a five-point Likert scale with categories ranging from “fully agree” (1) to “fully disagree” (5), the relative majorities of respondents fully agree with the statement that the UGS is relaxing (45%), and that they feel safe when visiting it (46%). Forty percent of the respondents agree with the statement that the closest UGS is clean. Furthermore, 27% of the respondents stated that they strongly disagree that having a UGS in the vicinity was an important factor for choosing their flat.

We asked for respondents’ approval for a set of potential reasons for visiting the closest UGS, also on five-point Likert scale. The respondents mostly indicated the following reasons for the last visit: “to get some fresh air,” “to switch off and get some distance from everyday life,” or “to enjoy nature.” Appendix Figures A2 and A3 illustrate respondents’ agreement to a set of statements regarding the closest UGS, each scored on a five-point Likert scale with categories ranging from “fully agree” (1) to “fully disagree” (5).

Figure 2

Example of a Choice Set

Based on the graphical Likert scale shown in Figure 1, 28.68% of respondents rate the naturalness of their closest UGS as “nearly natural,” followed by “partly natural” (28.37%) and “very natural” (20.82%). However, 13.76% state that their closest UGS is a “little natural,” and 8.37% think it is “hardly natural.” As shown in Table 2, the mean respondent defines the closets UGS as partly natural (mean Likert scale value of 3.40). The distribution of the variables is shown in Appendix Figure A4.

The average monthly rental payment is approximately €590, and the average additional utility costs the respondents pay are €192 per month.

4. Econometric Approach

DC models are based on the argument that attributes of goods determine the utility they provide (Lancaster 1966) and random utility theory (McFadden 1974). It is assumed that individuals choose an alternative that provides the highest level of utility. The utility U_njt of an individual n from an alternative j in a choice situation t is described by cost (rent) and noncost attributes x, which are observable to the researcher, and a random component ε_njt that is unknown: 1 where α_n is the cost coefficient, x_njtis a vector of variables describing goods or attributes of goods (naturalness and walking disrance), and ε_njt is assumed to be independently and identically distributed (i.i.d.) with an extreme value distribution, also known as Gumble distribution (Greene 2012). Because the variance of this distribution is π²/6, we are implicitly normalizing the scale of utility. To account for preference heterogeneity, the vector of taste β_n varies across individuals. Thus, we derive the mixed logit model (random parameter model) in preference space. To derive the WTP of the preference space model, we follow Mariel and Meyerhoff (2018) using Monte Carlo simulations as follows: 2 where β_k is the normal distributed random coefficient of any of the attributes k ∈ {naturalness, walkingdistance} and α_rent is the log-normal distributed random coefficient of the monetary attribute; in our case, the rent. The estimated SD of the attributes k and rent are indicated as and . The standard normal distributed random variables we denote as ϑ_k and ϑ_rent.

In addition, we estimate the model in WTP space (Train and Weeks 2005; Scarpa, Thiene, and Train 2008). The advantage of the WTP space model is that the estimated coefficients can be directly interpreted as WTP measures. Thus, we are able to compare the two applications and the resulting WTP estimates: 3 where rent_njt is the cost according to the payment vehicle, w_n is a vector of WTP for each noncost attribute (naturalness and walking distance), and λ_n is a random scalar. The scalar λ_n = α_n / k_n, where α_n is the cost coefficient in preference space, k_n is the scale parameter of individual n, and w_n = β_n / α_n, where β_n is the vector of the noncost coefficients in preference space. Finally, ε_njt is the random component. With homogeneous preferences (i.e., α_n and β_n, identical for all individuals), models 1 and 3 are fully equivalent. The difference comes about because the different assumptions about the type of preference heterogeneity. In utility space, model 1, the coefficients of the utility function follow normal or log-normal distributions; in WTP space, model 3, the corresponding assumptions on the distribution of heterogeneity are imposed directly on the WTP parameter.

Equations [1] and [3] are estimated using a simulated maximum likelihood estimation of a mixed logit model with 1,000 Sobol draws. We used the Apollo package in R for this purpose (Hess and Palma 2019). In our models, we included the alternative specific constants (ASC1 and ASC2), which show the preferences for programs 1 and 2 over the status quo. The attributes naturalness and the walking distance are assumed to be normally distributed, whereby the price coefficient (rent) is assumed to be log-normally distributed.

5. Empirical Results

Two empirical specifications, for both the WTP space and the preference space specification, were estimated to elicite the WTP for the naturalness of and walking distance to the closest UGS in 22 German cities. In Table 3, the results of the WTP space models are indicated with (a), and the calculated WTP measures of the preference space models are indicated with (b).5 Figures 3 and 4 show the estimated WTP values for naturalness and walking distance obtained from the WTP space models.

In the first model specification, model Ia and Ib, the variables naturalness and walking distance are treated as continuous. We extend this model specification in model IIa and model IIb and included the different level of naturalness as dummy variables. Moreover, we followed Mariel et al. (2021) and applied a piecewise linear approach for walking distance. We choose the quartiles of the walking distance as thresholds and generated the following variables: WD_1 = min(x,3); WD_2 = max(0,min(0-3,3)); WD_3 = max(0,min(x-6,9)); WD_4 = max(0,x-15). The statistically significant standard deviations in all models reveal unobserved preference heterogeneity among the respondents (Table 3).

Looking at the WTP space models, Ia and IIa, we see that the estimated mean WTPs are statistically significant at the 1% level. We found a negative WTP for both ASC variables, which indicates that on average, choosing any rebuilding option leads to a lower WTP compared with the status quo. In model Ia, we estimated a mean WTP for naturalness of €20.25 per month for a one-step increase on the Likert scale. The WTP for the choice attribute walking distance shows a negative WTP of −€2.47 per month for an additional minute of walking to the closest UGS.

Figure 3

Mean Willingness to Pay for an Increase in Naturalness as Continuous Variable (by 1) and Dummy Variables (Compared with Nearly Natural Urban Green Spaces)

Model IIa shows that the mean WTP inceases with the level of the naturalness of the UGS. The estimated mean WTP ranges from −€59.21 per month for hardly natural to €12.32 per month for very natural compared with the base category nearly natural (step 4 on the Likert scale). Furthermore, the results of the piecewise linear specification of the attribute walking distance show significant negative WTP (Table 3; Figure 4). When the walking distance to the closest UGS is less than three minutes, respondents have a negative mean WTP of −€5.79 per month per minute, whereas when the walking distance exceeds 15 minutes, the mean WTP is −€1.74 per month per minute.

Preference heterogeneity implies that only a share has a preference that goes in the same direction as the mean estimate. The share of individual-level coefficients that is positive is shown in the column “share” in Table 3. Model Ia indicates that most respondents (90%) have a positive WTP for UGSs with a higher level of naturalness. Furthermore, a share of 6% prefer a longer walking distance to the clostest UGS, whereas the majority of the respondents (94%) have a positive WTP for a shorter walking distance from their flat to the closest UGS.

Model IIa shows that around 90% of the respondents have a negative WTP for a hardly, little, or partly natural UGS compared with a nearly natural UGS. Furthermore, our results reveal that a share of 92% of the respondents have a positive WTP for a very natural UGS compared with a nearly natural UGS. For walking distance, the picture is similar. We find that 80% of the respondents with a walking distance of less than three minutes have a positive WTP for a short walking distance to their closest UGS. Also, most respondents with a walking distance of 3–6 minutes, 6–15 minutes, and more than 15 minutes have a positive WTP for a shorter walking distance (90%, 84%, and 94%, repectively).

Figure 4

Mean Willingness to Pay for Walking Distance by One Minute as Continuous and Piecewise Variables

The simulated median WTP (equation [2]) from the preference space models are presented in the last two columns of Table 3 (models Ib and IIb). In accordance with Train and Weeks (2005), the results of the preference space and WTP space specification are similar. However, the WTP space models yielded in a more plausible distribution of WTP, with fewer respondents having very high WTP than in the preference space models (see also Appendix Table A4 for detailed results and Appendix Figures A5 and A6 for the distributions of WTP of the preference space models).

We also tested the robustness of the results and included sociodemographic variables in the model, which we interacted with the naturalness variable (Appendix Table A5). Again, the WTP for a higher walking distance is negative, and the respondents prefer a UGS that is more natural. For these variables, the standard deviations are significant and similar to the magnitudes of our basic model Ia, indicating heterogeneity among respondents. Regarding the interaction terms, we find that more educated people have an higher WTP for more natural UGSs. The number of children and the size of the flat have a negative effect on respondents’ mean WTP for naturalness. As we expected intercity differences in monthly WTP values, we estimated the WTP space models for all cities with more than 100 observations separately to estimate the WTP of the mean respondent for a marginal increase in naturalness and walking distance and for the factor and piecewise variables specifications, as shown in Table 4.6

The estimated mean WTP values vary between cities. The mean WTP for naturalness ranges from €12.44 per month in Dresden to €35.88 per month in Bremen. The highest absolute value for the (negative) WTP for walking distance is shown for Bremen (−€3.62 per month), and the lowest WTP can be attributed to Dresden.

Figure 5

Correlation between the Offer of Green Space and Willingness to Pay for Naturalness of Green Space in Per Capita Terms (left) and Absolute Values (right)

As discussed, a sorting mechanism as proposed by Tiebout (1956) may play a role here. If this is the case, we expect a positive correlation between the amount of UGS the city offers and the WTP for natural green space. We tested this in two ways. First, we correlated the WTP for naturalness of a UGS with the UGSs per capita in the respective city (model IIIa). In a second specification, we correlated the aggregate WTP (population size times mean WTP) with the aggregated UGSs in the city, using GDP as control variable (model IIIb). Results from model IIIb are consistent with the Tiebout mechanism, as illustrated in Figure 5 and detailed in Appendix Table A6. However, this effect is not present in model IIIa.

6. Discussion and Conclusion

In this article, we studied how citizens value the naturalness of the closest UGS they use most often. While it is well known that public amenities like UGSs generate price premiums in housing markets (Cho et al. 2009), much less is known about individual preferences for characteristics of UGSs. Our study contributes to a better understanding of preferences for natural and biodiverse UGSs and proximity to them. We introduced a graphical measurement scale for the naturalness of a UGS that varies in five steps between hardly natural and very natural. To estimate the WTP for changes in the naturalness or proximity of a UGS, we used changes to the housing rent as a payment vehicle. This seems appropriate, as the attractiveness of a neighborhood commonly affects urban housing rents; however, we cannot exclude that some doubts about the consequentiality of their choices led to some noise in responses.

We found that German citizens on average hold preferences for a higher naturalness of their closest UGS and value a short walking distance to this UGS. The mean respondent is willing to pay €20.25 per month for an increase in the naturalness of their closest UGS by one step on the naturalness scale. Thus, on average, respondents would benefit if their closest UGS would become more natural.

On the other hand, respondents receive on average a loss in terms of a negative WTP of −€2.47 per month for an additional walking minute to their closest UGS. Regarding the UGS vicinity, our findings are in line with previous revealed preference studies (Palmquist 1992; Plant, Rambaldi, and Sipe 2017; Łaszkiewicz, Czembrowski, and Kronenberg 2019). Park et al. (2017) show that an increase in the distance to the green space by 1 m causes a decrease in the expected house value by $309. In the few studies that consider rental prices, Zhang et al. (2020) found that the presence of a park within 500 m vicinity leads to a rent increase of 1.39% in Beijing, China, whereas Donovan and Butry (2011) found that an increase in the distance to a park by 1 km increases the price by 3.3% in Portland, Oregon, United States.

Using our five-point scale to measure the naturalness of the UGS, we were able to elicit preferences and WTP measures for rebuilding the closest UGS in terms of different levels of naturalness. This is important because UGSs with different characteristics provide different types of benefits (e.g., sports facilities, areas for social and cultural interactions), and the acceptance of potential rebuilding measures depends on how this will affect the utility of the UGS. Our results reveal that the mean respondent receives the highest benefit, measured in WTP, from rebuilding schemes that increase the naturalness of their closest UGS. We find a mean WTP of €12.32 per month for very natural UGS relative to near natural. The average respondent indicates a loss of −€59.21 for a decrease in naturalness from near natural to hardly natural.

In general, urban development is facing significant societal challenges. In particular, demographic changes are different in different areas (Martinez-Fernandez et al. 2012). Thus, we were interested in analyzing intercity differences between important German cities. Indeed, WTP measures differ between these cities. For distance, they range between a monthly mean WTP for an extra minute walking to the closest UGS of −€1.63 in Dresden to −€3.62 in Bremen. Future research should be devoted to investigating this spatial preference heterogeneity, which should be of key interest for urban planners.

Overall, our results highlight the importance of urban nature for city life. Considering the rapid urbanization (United Nations 2014), our insights might be of use for urban planning and management because they evidence the importance of preserving and improving biodiversity in cities, particularly those areas where citizens use green spaces around their places of residence in the daily life. The appreciation of a high level of naturalness suggests that many urban residents support nature-oriented rebuilding schemes of UGSs in Germany.

Future research may use our multisite CE as a starting point for gaining a better understanding of what drives the intercity differences in the median WTP for biodiverse urban greenery. These differences might depend on the level and spatial distribution of urban greenery in cities as well its correlation with income (Meya 2020) or on the availability of substitutes, such as environmental amenities outside a city or private gardens. Understanding heterogeneity in the WTP for biodiversity in cities on the respondent level could inform benefit transfer and add to an emerging literature on spatial heterogeneity in the WTP for environmental public goods (Czajkowski et al. 2017; Liu, Hanley, and Campbell 2020).