Green gentrification: a fuzzy-set qualitative comparative analysis of greened neighbourhoods in Berlin

Abstract

To contribute to the understanding of green gentrification, this study aims to elucidate the factors that contribute to its occurrence or non-occurrence following urban greening, by considering necessary and sufficient factors. The investigation examines several potential characteristics, including the distribution, size, function, transportation connectivity, neighbourhoods’ centrality, and pre-existing availability of green spaces within neighbourhoods. The research employs an embedded case study approach, focusing on Berlin as the primary case. A fuzzy-set qualitative comparative analysis (QCA) is conducted to analyse the notions of necessity and sufficiency. The analysis identifies several sufficient pathways to gentrification, indicating that the neighbourhood context, particularly the initial availability of green spaces in the neighbourhood, plays a crucial role in determining the occurrence of green gentrification. By contrast, no specific type of greening, such as scattered greening approach, park size, function, or transportation connectivity, consistently led to green gentrification.

Keywords:

• green gentrification
• urban green spaces
• berlin
• qualitative comparative analysis

1. Introduction

Green spaces have been shown to be associated with both ecological and (physical and mental) health benefits (Cole et al. 2017; Mensah et al. 2016; Sharifi et al. 2021). In effect, urban greening initiatives (UGIs) are often portrayed as an all-around win-win approach (Anguelovski et al. 2019; Anguelovski and Connolly 2022; Cole et al. 2017). ‘Going green’, planners and policymakers assume, will necessarily improve a city’s sustainability and general prosperity (Anguelovski et al. 2018; Curran and Hamilton 2012; Gould and Lewis 2017). However, researchers have noted a frequently unequal distribution of green spaces, wherein predominantly white and higher-income residents tend to enjoy greater access to parks and greenways (Anguelovski et al. 2018; Triguero-Mas et al. 2022). Consequently, individuals who are already vulnerable and politically marginalised often find themselves excluded from the ecosystem services that parks and greenways offer. Systematic reviews have revealed a strong correlation between an individual’s quality of life and their access to green spaces (Mensah et al. 2016).

Partly in response to the observed environmental inequalities, cities worldwide are implementing numerous greening initiatives specifically targeted at historically underserved neighbourhoods. For instance, Berlin aims to increase the availability of public green spaces to a minimum of 6 square metres per resident (Senatsverwaltung für Stadtentwicklung und Wohnen Berlin [SenSW] 2017). However, researchers have identified a phenomenon known as the ‘green space paradox’ (Anguelovski et al. 2018; Anguelovski and Connolly 2022; Curran and Hamilton 2012; Draus et al. 2019; Wu and Rowe 2022): despite cities’ efforts to create more, larger, and higher-quality green spaces as a means of promoting environmental justice, these UGIs may contribute to social and economic inequalities through processes reminiscent of gentrification. These observations have led to the assertion that greening initiatives are causally linked to gentrification, giving rise to the concept of green gentrification (Anguelovski et al. 2018; Dooling 2009; Gould and Lewis 2017; Rigolon and Németh 2020).

However, some studies suggest that UGIs do not always lead to gentrification outcomes (Rigolon and Németh 2020, Stuhlmacher, Kim, and Kim 2022). In addition, the factors that determine the occurrence and extent of green gentrification remain poorly understood (Triguero-Mas et al. 2022). One such factor is the size of the park, which has been suggested to affect the likelihood and intensity of gentrification. According to the ‘just green enough’ (JGE) approach, introducing many small parks in an urban area may prevent gentrification (Curran and Hamilton 2012; Wolch, Byrne, and Newell 2014). Conversely, other researchers argue that park size is just one among multiple determinants of gentrification outcomes, and even small parks can contribute to gentrification in attractive neighbourhood contexts (Gould and Lewis 2017). Other studies, carried out in different geographical contexts, have yielded partly different results on the effect of park size (Chen et al. 2021, Kim and Wu 2022). Furthermore, some scholars challenge the JGE hypothesis and emphasise the role of factors like park location (Rigolon and Németh 2020). It has been observed that parks in closer proximity to city centres generally exhibit more noticeable gentrification outcomes, although these claims are not conclusive (Rigolon and Németh 2020). Additionally, public transport connectivity is proposed to be influential in gentrification outcomes (Gould and Lewis 2012; Gould and Lewis 2017; Zukin 1987). Finally, the functions or potential uses of the park are also likely to be determinants of the impact of the greening initiative on gentrification in a neighbourhood (Rigolon and Németh 2020).

To date, there has been limited empirical research investigating the link between green-space creation initiatives and gentrification, extending beyond the analysis of park distribution patterns and park size (Cole et al. 2017; Rigolon and Németh 2020; Chen et al. 2021, Kim and Wu 2022). Consequently, a significant knowledge gap remains regarding the specific factors related to the type of greening and the surrounding urban context that contribute to the occurrence or absence of gentrification following urban greening (Stuhlmacher, Kim, and Kim 2022). Given this knowledge deficit, the objective of this article is to identify the necessary and sufficient conditions, if any, or combinations thereof, that are associated with the emergence or absence of green gentrification. To do so, this article makes use, for the first time, of qualitative comparative analysis (QCA). As further explained in the methodology section, QCA can be used to define necessary and sufficient conditions by systematically examining combinations of factors that are either present or absent in cases, helping to identify the critical conditions that are consistently associated with an outcome, that is green gentrification which is the focal point of the article. Identifying necessary and sufficient conditions is crucial when studying green gentrification, a complex phenomenon, as it provides nuanced insights into the specific factors that consistently drive this environmental transformation. By employing QCA to discern the essential conditions, researchers can unravel the intricate web of elements contributing to green gentrification. This precision not only enhances our understanding of the phenomenon but also enables the development of targeted strategies to address its drivers and consequences.

2. Green gentrification: processes and conditions

Quinton, Nesbitt, and Sax (2022) have argued that the ways in which urban greening contributes to gentrification are not thoroughly understood and this is also partly due to the complex interactions between greening and non-greening factors. Furthermore, the spatiotemporal relationship between greening and gentrification can be interpreted as a cycle in which gentrification may precede or follow greening, depending on the influence of different actors and processes and their interplay (Rigolon and Collins 2023). Similar to other forms of gentrification, the process of green gentrification is often triggered by a perceived increase in a neighbourhood’s desirability (Anguelovski et al. 2018, 2019; Anguelovski and Connolly 2022; Connolly 2019; Gould and Lewis 2012; Gould and Lewis 2017; Gould and Lewis 2018; Rigolon and Németh 2020; Sharifi et al. 2021). In the context of green gentrification, this enhanced attractiveness is partially attributed to the introduction of new parks or greenways (Ali, Haase, and Heiland 2020; Gould and Lewis 2012; Wu and Rowe 2022). The arrival of a new green space tends to attract individuals belonging to what Gould and Lewis describe as the ‘sustainability class’ (Gould and Lewis 2017, 107; Gould and Lewis 2018, 12). The sustainability class consists of individuals who prioritise ecological sustainability and perceive green spaces as integral to a sustainable urban lifestyle (Gould and Lewis 2018). Typically, members of this class have middle or high incomes, enabling them to demonstrate their demand for green spaces through financial means (Gould and Lewis 2018). As more residents from the sustainability class enter the local real estate market, lower-income residents are often priced out and ultimately displaced due to their lack of competitive purchasing power (Rigolon and Németh 2018). Over time, the influx of higher-income residents and the subsequent outmigration of lower-income residents lead to a social upgrade of the neighbourhood. However, it is important to note that the causal relationship between greening initiatives and gentrification is uncertain (Cole et al. 2017; Gould and Lewis 2017), and green gentrification remains a complex phenomenon that is challenging to fully comprehend (Sharifi et al. 2021). A study conducted in Berlin examined the monetary value individuals assign to green spaces within their residential environment and found that, on average, individuals attribute EUR 27 per month to each hectare (10,000 m²) of green space (Bertram and Rehdanz 2015). Consequently, an increase in the availability of green space in a neighbourhood is likely to be associated with an overall increase in the monetary value attributed to the neighbourhood, as reflected in real estate prices. However, it is important to note that the proximity or mere presence of newly created urban green spaces (UGSs) does not necessarily result in gentrification outcomes, even though land closest to new parks tends to experience the highest value increases. Recent research on park size, location, and function confirms that not all UGIs lead to measurable gentrification in the surrounding area (Rigolon and Németh 2020). The availability and proximity to green spaces are necessary but not sufficient conditions for green gentrification. It should be clarified that the necessity of greening does not imply that it is a universal prerequisite for gentrification; rather, in the context of green gentrification, a UGI must have preceded the observed urban shifts. The hypothesised classification of non-sufficiency suggests that other variables play a role in determining whether, and to what extent, greening contributes to gentrification. Thus, green gentrification can be characterised as a multifactorial process.

2.1. Park function

The study by Rigolon and Németh (2020) suggests that the potential uses or functions of a park can influence green gentrification outcomes. Functionality is indirectly related to green gentrification, as parks with more potential uses tend to attract more visitors, thereby increasing the neighbourhood’s perceived attractiveness. The researchers examined the role of a park’s function in generating green gentrification outcomes by specifically analysing the use of the park for active transportation. In their sample of US parks, all parks that served as sites for active transportation were associated with more significant gentrification outcomes. However, it is important to note that active transportation is just one of many potential functions that UGSs can offer. The perceived desirability of different functions is highly dependent on the cultural context in which the park is located. According to Kabisch and Haase (2014), the perceived effectiveness of a green space as a recreational area is influenced by the specific cultural expectations and preferences of potential local visitors, as well as the degree to which these expectations are met. While migrants may prefer parks with barbecue facilities or other communal areas for picnicking, native Germans favour areas for active sports and play (Kabisch and Haase 2014). Furthermore, studies examining cultural ecosystem services in the context of green gentrification have shown that the socio-cultural association of green spaces has an impact when green gentrification occurs. In particular, parks associated with ‘aesthetics’ and ‘recreational activities’ were found to be associated with subsequent gentrification, while those associated with ‘cultural identity’ and ‘social activities’ were not (Amorim et al. 2020). To thoroughly assess the impact of functionality on gentrification outcomes, it is crucial to consider the specific desired park functions within the context of the study area.

2.2. Park size and scatter degree

Regardless of their intended purpose, empirical studies indicate that the size of a park plays a significant role in the phenomenon of green gentrification. Early investigations on green gentrification consistently observed a higher likelihood of gentrification associated with large parks compared to small ones. Notably, in a study conducted by Gould and Lewis (2017), it was found that only one relatively small park among three greening initiatives did not result in gentrification. These findings, along with similar observations, support the notion of a ‘just green enough’ approach (Curran and Hamilton 2012; Wolch, Byrne, and Newell 2014). In its most recent developments, this approach posits that smaller parks are less likely to trigger gentrification, suggesting that deliberate efforts to limit the size of green spaces can help to prevent gentrification (Wolch, Byrne, and Newell 2014). However, when researchers systematically tested the relationship between park size and gentrification outcomes, no statistically significant association was found, leading some scholars to reject the ‘just green enough’ hypothesis (Rigolon and Németh 2020). It should be noted that neither Curran and Hamilton (2012) nor Wolch, Byrne, and Newell (2014) analysed quantitatively whether smaller parks were less likely to cause gentrification than other parks. Additionally, an inherent limitation of the ‘just green enough’ approach is its potential to impede a city’s capacity to significantly improve the availability of green spaces if new parks are exclusively small. To address this concern, proponents of the ‘just green enough’ approach put forward a strategy of ‘small-scale and scattered’ greening initiatives (Wolch, Byrne, and Newell 2014, 241). Theoretically, this approach allows for a substantial expansion of green spaces in an area while still mitigating the effects of gentrification. However, the empirical testing of this aspect of the ‘just green enough’ approach remains unexplored.

2.3. Supply and tipping points

An alternative perspective posits that park size plays a crucial role as a “tipping point in the urban environmental amenity production process” (Gould and Lewis 2017, 170). According to this viewpoint, a threshold in park size exists below which greening initiatives do not trigger gentrification. This threshold hypothesis can be better understood by examining the relationship between the absolute provision of green space and individual satisfaction. For instance, a study conducted in Berlin revealed a U-shaped effect between the supply of UGSs and life satisfaction, indicating that a provision of 35 hectares (350,000 m²) within an individual’s immediate living environment (within a 1 km buffer) led to the maximum increase in self-reported life satisfaction (Bertram and Rehdanz 2015). Therefore, when investigating the impact of varying park sizes on neighbourhood attractiveness, as well as the resulting social and real estate structures, it is crucial to consider both the relative change in the supply of UGSs and the initial per resident supply of green spaces, referred to here as green space availability (GSA). Such relational measures are already commonly employed in the field of environmental valuation, where the perceived value of life satisfaction derived from green spaces accounts for the relative change in their provision (e.g. Ambrey and Fleming 2014). It is worth noting that the perception of adequate green space is likely to be affected by other factors including the overall level of greenness of a neighbourhood, such as the presence of trees and private gardens. Thus, the perceived level may differ from the actual supply of public green space.

2.4. Centrality

The proximity of a neighbourhood to the city centre emerges as another influential factor in determining gentrification outcomes (Anguelovski et al. 2018). Rigolon and Németh (2020) identified the proximity to the downtown area as a statistically significant predictor of green gentrification in their study on gentrification drivers. Among the cases examined, parks situated within a distance less than the median for gentrification-eligible tracts in each city were found to be more likely to contribute to gentrification outcomes (Rigolon and Németh 2020). Even small UGSs located in central areas can trigger gentrification (Rigolon and Németh 2020). Furthermore, as pointed out by Pearsall and Eller (2020), also areas located near previously gentrified or gentrifying neighbourhoods are more susceptible to green gentrification. Consequently, neighbourhood centrality may not only serve as a determining factor for gentrification but also one of greater significance than park size.

2.5. Transportation connectivity and overall development

In addition, the connectivity of parks to the city’s transportation system can significantly enhance a neighbourhood’s appeal and contribute to gentrification, even if the parks are located further away from the city centre. This factor is particularly influential in cities with well-developed public transportation systems, where proximity to a train station, for instance, implies extensive accessibility. Moreover, the overall transportation connectivity of a neighbourhood has been identified as a cause of gentrification independent of greening initiatives (Gould and Lewis 2012; Zukin 1987). It has been observed that many underdeveloped neighbourhoods characterised by inadequate green space provision were also disconnected from the rest of the city and lacked accessible public transportation options (Matheney, Pérez del Pulgar, and Shokry 2021). Consequently, it is plausible to suggest that good transportation connectivity is associated with more advanced neighbourhood development. Furthermore, neighbourhood development itself emerges as another key determinant of green gentrification outcomes. Neighbourhoods that are generally more developed and possess other desirable features such as commercial districts (Gould and Lewis 2017) or a historic housing stock (Anguelovski et al. 2018; Gould and Lewis 2017) tend to experience accelerated gentrification in response to park-building initiatives (Anguelovski et al. 2018). On the contrary, neighbourhoods characterised by industrial activities, high population density, and overall lower attractiveness appear to generate less significant or no gentrification following greening efforts (Anguelovski et al. 2018; Gould and Lewis 2017).

To summarise, in this study the phenomenon of green gentrification is conceptually divided into two main components: the outcome of the green gentrification process and the underlying factors or conditions that contribute to it. The outcome of green gentrification can be understood as the complex interrelationship between social upgrading and real estate valorisation (Helbrecht 2018; Holm and Schulz 2017; Schulz 2017), with increases in real estate prices attracting higher-income residents, subsequently driving social upgrading of the neighbourhood and placing upward pressure on property prices (Anguelovski et al. 2018; Gould and Lewis 2018). This complex conjunctural outcome requires a combined measurement approach that incorporates both social and real estate indices (Holm and Schulz 2017). The factors or conditions contributing to the occurrence (and non-occurrence) of green gentrification can be classified into two interlinked categories: those directly associated with specific UGIs and those pertaining to the neighbourhood context. While these two categories are conceptually distinct, they are interconnected. It is unlikely that any single condition alone is sufficient to drive gentrification outcomes. However, specific combinations of these factors can create pathways that lead to such outcomes.

A ‘scattered’ approach, scattering greening across both time and space, is anticipated to mitigate green gentrification (Curran and Hamilton 2012). Park size is expected to play a highly impactful role. Larger parks are expected to increase the likelihood of green gentrification, while small parks are expected to have a mitigating effect (Gould and Lewis 2017). The relative change a greening initiative brings about in the overall supply of UGSs in a neighbourhood is postulated to influence the outcome. The presence of culturally relevant functions in a park is anticipated to enhance its appeal (Dai 2011), potentially contributing to gentrification. A park’s location and connectivity to transportation networks may also have an effect, with poorly connected parks potentially hindering gentrification outcomes. Neighbourhood centrality, especially in proximity to the city centre, is anticipated to trigger more frequent and intense green gentrification (Rigolon and Németh 2020). In contrast, non-central neighbourhoods might require other specific combinations of conditions for similar outcomes. The overall supply of green space, or the initial GSA, is likely to play a crucial role in determining how responsive a neighbourhood is to urban greening in terms of subsequent gentrification, particularly in the case of low initial GSA.

3. Methods

3.1. Study location

Drawing upon the embedded case study methodology proposed by Yin (2017), this research focuses on the city of Berlin as the primary case, with a particular emphasis on analysing the various UGIs implemented between 2007 and 2016 in Berlin as the secondary level of analysis (see Annex 1 [online supplementary information] for a list of the UGIs). The study period was chosen for practical and substantial reasons, which mirror those cited by other scholars who have studied gentrification in Berlin (Holm and Schulz 2017; Schulz 2017). In 2007, a new phase of real estate upgrading processes began in Berlin, setting off gentrification processes across the city (Schulz 2017). Moreover, most of the required data is only available at the spatial scale of Lebensweltlich Orientierten Planungsräume (LOR or real life-oriented planning areas) from 2007 onwards. LOR are the smallest spatial unit of data collection in Berlin, correspond to the neighbourhood level in terms of size and number of inhabitants (SenSW 2017) and, therefore, are appropriate to investigate neighbourhood gentrification patterns. The year 2016 was chosen as the cut-off point for green space construction to allow for a minimum of 5 years to evaluate the impact of greening on gentrification. Similar observational timeframes have been chosen by other researchers when measuring gentrification patterns (Schneider, Oana, Thomann 2021; Schulz 2017).

Berlin, the capital of Germany, serves as an ideal case study for examining the phenomenon of green gentrification. The city has experienced significant population growth, with a notable increase of 421,622 inhabitants between 2007 and 2021, reaching a total of 3,775,480 residents at the end of the last year (Amt für Statistik Berlin-Brandenburg 2021). This population growth has not been evenly distributed but concentrated in specific areas (SenSW 2017). Consequently, Berlin is experiencing immense pressure on the housing market (Helbrecht 2018), while at the same time undergoing gentrification, with an increasing number of socially disadvantaged groups being pushed out. In 2021, Berlin had a total of 142,343 rent-controlled housing units (8.5% of the total rental housing stock), a number that has continued to decline compared to previous years (Investitionsbank Berlin 2023). In response to this, Berlin initiated the creation of 65 new parks and greenways. However, rapidly growing neighbourhoods continued to see a decrease in the green space supply ratio, reflecting the total area of green spaces in a region against the number of inhabitants. To address this, the city of Berlin published a greening strategy for 2018 onwards, aiming to provide six square meters of accessible green space per inhabitant (SenSW 2017). Given these linked greening initiatives and their potential impact, Berlin is of particular interest for determining the conditions contributing to green gentrification, both historically and in response to future planned greening initiatives. The 65 green spaces created between 2007 and 2021 provide a robust foundation of cases for study.

The embedded cases, 33 in total (see Annex 1 [online supplementary material]), are newly created green spaces within one LOR and are categorised as UGIs for analytical purposes. When multiple parks are created within one LOR but more than 5 years apart, they are studied as individual UGI. Parks created within a single initiative are located within the same LOR and built within a maximum of 5 years of each other. It should be noted that, given Berlin’s complex history, the city lacks a singular centre characteristic of other urban environments. In this study, employing Alexanderplatz as a measure of centrality does not suggest it as the exclusive, definitive centre of Berlin. Instead, Alexanderplatz functions as a distinct point of reference within the city’s fabric, occupying a central location at Berlin’s midpoint. It is also important to emphasise that the 33 UGI examined in this study are situated in LOR with varying levels of socio-economic development, and thus, they are differently exposed to the risk of gentrification. This variation represents a limitation of the study and may restrict the generalisability of its findings.

3.2. Analytical approach

Fuzzy set QCA (fsQCA) is used to conduct the empirical investigation. QCA is a research method that assesses how combinations of different conditions contribute to an outcome, emphasising the configurational patterns of factors rather than individual variable effects. fsQCA extends traditional QCA by allowing for the consideration of degrees of membership or partial inclusion in sets, accommodating situations where elements may exhibit varying levels of adherence to defined conditions (Schneider and Wagemann 2012). Fuzzy sets avoid binary categorisation and are used to capture the subtleties and complexities of the relationship between urban greening and gentrification, accommodating shades of membership and non-membership as certain factors may be neither fully in nor fully out of a set (Ragin 2014). QCA as a set-theoretic approach explicitly acknowledges the presence of conjunctural causality, equifinality, and causal asymmetry. Conjunctural causality refers to the idea that outcomes result from the combination or joint presence of multiple conditions, emphasising the interplay and synergy among these conditions rather than isolated effects. Equifinality suggests that different combinations of conditions can lead to the same outcome, indicating that there are multiple pathways or configurations that can produce a similar result. Causal asymmetry implies that the presence and the absence of the outcomes may require different explanations (Schneider, Oana, Thomann 2021; Ragin 2014; Schneider and Wagemann 2012). QCA appears to be an appropriate methodological approach to study green gentrification, as this is a multifaceted phenomenon which may require the co-occurrence of multiple conditions, such as park size and neighbourhood centrality. Furthermore, different combinations of conditions may lead to the same outcome. Additionally, green gentrification may exhibit causal asymmetry, wherein the factors contributing to its occurrence or non-occurrence are not symmetrical. By incorporating complexity-theoretic assumptions into the analytical framework, fsQCA offers valuable insights into elucidating the intricate relationship between urban greening and gentrification. More specifically, it enables the identification of the necessary or sufficient conditions (and their combinations) proposed in the conceptual framework that contributes to the occurrence of green gentrification.

Threshold values play a defining role in set theoretic methods. Full set-membership (corresponding to a score of 0.95) indicates that a condition is expressed at a high level, meaning the case is fully within the set (Schneider, Oana, Thomann 2021). Conversely, full set non-membership (0.05) denotes that a condition is entirely absent, with the case entirely out of the set (Schneider, Oana, Thomann 2021). The membership threshold/crossover point, set at a score of 0.5, marks the point at which a condition is neither fully expressed nor entirely absent; the case is in a state of indifference regarding set membership (Schneider, Oana, Thomann 2021). These values are carefully translated into raw data scores for the studied conditions and outcomes to demarcate set membership and non-membership. To mitigate the possibility of obtaining unreliable results, the direct calibration method, as outlined by Schneider, Oana, Thomann (2021), was employed. Following the establishment of the calibration system, each case was assigned a score derived from the raw data. Notably, the data pertaining to all fuzzy sets were intentionally collected and meticulously prepared to accurately capture nuanced variations in numerical variables, in accordance with the prerequisites of the direct calibration method (Schneider, Oana, Thomann 2021).

Gentrification, as the studied outcome, was identified and modelled at a neighbourhood level using the Gentrification Index (GI), based on Holm and Schulz’s GentriMap model (Holm and Schulz 2017), a statistical model that was specifically developed to measure both small scale gentrification dynamics and compare city-wide patterns of gentrification. Holm and Schulz designed the model and proposed the indicators adopted in this study (see Annex 3 [online supplementary information]) for the context of Berlin. The GI combines two key sub-indexes: the real-estate index (REI) and the social index (SI). The REI measures the real-estate value increase associated with gentrification, while the SI captures the social upgrading associated with the process. To calculate the GI, a shift-share analysis is employed to determine the deviation of each LOR from the city-wide trend in social and real estate developments during the same timeframe (Holm and Schulz 2017). A positive output value (OUT) indicates above-average changes, reflecting social upgrading and real estate valorisation. Afterwards, the individual indexes are standardised through a z-transformation to allow for their combination (Holm and Schulz 2017). Finally, the two standardised indexes are combined (as an average) to indicate the gentrification intensity (see Annex 3 for the relevant formulae and computational steps [online supplementary information]). The GI (GentriMap model) is specifically designed to be as generalisable and transferable as possible to different urban contexts, and to allow the results of different studies to be compared (Holm and Schulz 2017). Consequently, it does not take into account factors such as rent control when operationalising gentrification.

Necessity and sufficiency analysis were then performed for the occurrence and non-occurrence of the outcome. In this context, ‘sufficiency’ refers to the identification of combinations of conditions that lead to gentrification or non-gentrification. Conversely, ‘necessity’ pertains to conditions that are always present or absent concerning the outcome. Consistent with standard conventions, the threshold for determining necessity was typically established at 0.9, allowing for a certain degree of deviation from absolute necessity, as outlined by Schneider, Oana, Thomann (2021) and Ragin (2014). To assess trivialness, the analysis incorporated fit parameters such as Coverage and Relevance of Necessity (RoN), with a threshold value of 0.5 deemed as critical according to Schneider, Oana, Thomann (2021). While there is no consensus in the literature regarding a specific minimum value for consistency sufficiency, it is widely acknowledged that a threshold of 0.75 should not be exceeded, as noted by Schneider, Oana, Thomann (2021). Consequently, the lower boundary for this parameter was established accordingly. An enhanced standard analysis (ESA) was conducted to mitigate logical contradictions.

3.3. Data and operationalisation

The operationalisation of the variables and indicators is presented in Annex 2, data sources and computations in Annex 3, calibration in Annex 4, the raw data matrix in Annex 5, and the fuzzy set and crisp set scores in Annex 6 (online supplementary material). Table 1 provides an overview of the key variables considered in the study.

Table 1. Variables.

Variables UGI	Description
Scatter approach	Scatter level of new green space across time and space.
Park size (1)	Average size of the newly created green space.
Park size (2)	Relative increase in the total supply of green space.
Park community and recreational function	Community and recreational (leisure, sports) functions provided by the green space.
Park accessibility and connectivity	Park level of connectivity to the public transportation system.
Variables neighbourhood context	Description
Neighbourhood centrality	Distance of the neighbourhood from the city centre (i.e. Alexanderplatz).
Initial green space availability (GSA)	Availability of public green space prior to the start of the greening initiative in the respective LOR.
Variables gentrification	Description
Gentrification intensity	Intensity of the gentrification process.
Real estate index (REI)	Measure of real estate value increase that is associated wdith gentrification.
Social index (SI)	Measure of social upgrading associated with gentrification.

4. Results

4.1. Necessary conditions for gentrification

When examining individual conditions to determine a superset relationship with gentrification outcomes, only one condition can be deemed ‘potentially necessary’: ‘Not central neighbourhood’ (∼CENTR1) (refer to Table 2). In the case of this condition, the consistency necessity parameter of fit surpasses the established threshold of 0.9, indicating that non-centrality (∼CENTR1) serves as an empirically consistent superset of the outcome (GI1). The substantial consistency implies that the majority of the investigated neighbourhoods demonstrating initial gentrification (membership score <0.5) are situated away from the city centre (membership score <0.5). Both the Coverage and RoN parameters of fit exceed the critical point of 0.5, thereby alleviating any immediate concerns regarding empirical triviality.

Table 2. Conditions and parameters of fit.

Condition	Consistency necessity parameter of fit (Cons.Nec)	Coverage necessity parameter of fit (Cov.Nec)	Relevance of necessity parameter of fit (RoN)
Scattered greening (SCAT1)	0.164	0.504	0.919
Small UGS (UGSS_NW1)	0.802	0.530	0.473
Significant UGS supply increase (UGSS_PC1)	0.561	0.590	0.755
Recreational function (FUNC1)	0.461	0.442	0.656
Well-connected by public transportation (CONN1)	0.763	0.536	0.524
Central neighbourhood (CENTR1)	0.449	0.629	0.844
Adequate initial GSA (GSA1)	0.772	0.576	0.588
No scattered greening (∼SCAT1)	0.836	0.458	0.248
No small UGS (∼UGSS_NW1)	0.484	0.759	0.908
No significant UGS supply increase (∼UGSS_PC1)	0.692	0.576	0.652
No recreational function (∼FUNC1)	0.539	0.487	0.647
Not well-connected by public transportation (∼CONN1)	0.544	0.748	0.886
Non-central neighbourhood (∼CENTR1)	0.915	0.637	0.578
Inadequate initial GSA (∼GSA1)	0.417	0.514	0.773

Note: Bold rows indicate a consistency score above the relevant threshold (Cons.Nec ≥ 0.9).

The proposed necessity claim, suggesting that the non-centrality of the greened neighbourhood is indispensable for the subsequent occurrence of gentrification (∼CENTR1 ← GI1), can be visually examined through an XY plot (refer to Figure 1). The plot illustrates that the majority of cases fall below the diagonal line, indicating their consistency with the superset relationship. Additionally, a considerable number of cases below the diagonal are positioned in the upper-right quadrant. These cases exhibit a membership score above the crossover point (0.5) for both the condition and outcome sets, implying their higher inclusion in both the non-centrality (∼CENTR1) and gentrification (GI1) categories. Such cases can be considered typical and thus provide robust evidence supporting the implied superset relationship.

Figure 1. Non-centrality as a necessary condition of gentrification.

However, the plot also reveals that although the majority of cases align with the necessity claim, there are several instances that lie above the diagonal line, deviating from the superset relationship (UGI 3, 11, 12, 13, 32). Additionally, two of these deviant cases (cases 3 and 11) can be classified as deviant cases consistency in kind (DCK) since they are situated in the upper-left quadrant. This implies that they are more excluded than included in the condition set (∼CENTR1), yet they exhibit a greater inclusion than exclusion in the outcome set (GI1). Put simply, cases 3 and 11, provide substantial empirical evidence contradicting the necessity claim by demonstrating that gentrification can indeed occur in central neighbourhoods.

4.2. Necessary conditions for non-gentrification

Upon conducting the necessary analysis for the negated outcome (∼GI1) it is evident that no individual condition, whether in its present or absent form, exhibits consistent superset status. None of the conditions attain a consistency necessity parameter of fit surpassing the applicable threshold of 0.9 (see Annex 8 [online supplementary information]). Given that all conditions fall below the threshold for consistency necessity, there is no requirement for assessing empirical trivialness. Consequently, it can be concluded that no individual condition must necessarily be present or absent to prevent the occurrence of gentrification following a UGI.

During the investigation of necessary combinations of conditions for the negation of the outcome (∼GI1), a single empirically consistent superset disjunction has been identified: ‘CONN1+∼GSA1’. Individually, the conditions ‘well connected by public transportation’ (CONN1) and ‘inadequate initial GSA’ (∼GSA1) do not constitute consistent supersets of the outcome. However, when combined through the logical OR (+) operator, these conditions form a consistent superset of ‘non-gentrification’ (∼GI1). This SUIN combination demonstrates adequate values for all three parameters of fit (refer to Table 3 and Table 4). This configuration implies that in cases where gentrification does not occur following urban greening, it can be anticipated that the newly established park will be adequately connected by public transportation (CONN1), or the initial GSA prior to the UGI was deemed inadequate (∼GSA1).

4.3. Sufficient conditions for gentrification

When investigating the occurrence of gentrification, there are eight logically plausible configurations of conditions that demonstrate acceptable parameter fits as sufficient conditions (consistency ≥ 0.75, PRI > 0.5) and are supported by sufficient empirical evidence (refer to Table 5). These specific combinations of conditions, represented in the truth table rows, can be classified as potentially sufficient for the occurrence of gentrification, indicated by an OUT of 1 (OUT = 1).

Table 3. Potential SUIN conditions and parameters of fit for non-gentrification outcomes.

SUIN configuration	Consistency necessity parameter of fit (Cons.Nec)	Relevance of necessity parameter of fit (RoN)	Coverage necessity parameter of fit (Cov.Nec)
CONN1 + ∼GSA1	0.918	0.507	0.662

Note: CONN1 = Well-connected by public transportation; ∼GSA1 = Inadequate initial GSA; The symbol + means logical OR.

Table 4. No significant UGS supply increase or Inadequate initial GSA and parameters of fit for non-gentrification outcomes.

SUIN configuration	Consistency necessity parameter of fit (Cons.Nec)	Relevance of necessity parameter of fit (RoN)	Coverage necessity parameter of fit (Cov.Nec)
∼UGSS_PC1 + ∼GSA1	0.918	0.420	0.625

Note: ∼UGSS_PC1 = No significant UGS supply increase; ∼GSA1 = Inadequate initial GSA; The symbol + means logical OR.

Table 5. Potentially sufficient configurations (OUT = 1) for gentrification (GI1).

	Scattered greening (SCAT1)	Small UGS (UGSS_ NW1)	Significant UGS supply increase (UGSS_PC1)	Recreational function (FUNC1)	Well- connected by public transportation (CONN1)	Central neighborhood (CENTR1)	Adequate initial GSA (GSA1)	Output value (OUT)	Number of cases in configuration (n)	Sufficiency inclusion score (incl)	Proportional reduction in inconsistency (PRI)	Deviant cases consistency (DCC)
2	0	0	0	0	0	0	1	1	2	0.916	0.755	5, 31
18	0	0	1	0	0	0	1	1	1	0.947	0.550	17
30	0	0	1	1	1	0	1	1	2	0.892	0.743	1, 32
34	0	1	0	0	0	0	1	1	2	0.963	0.898	16, 29
49	0	1	1	0	0	0	0	1	1	0.905	0.644	19
102	1	1	0	0	1	0	1	1	1	0.896	0.828	7
106	1	1	0	1	0	0	1	1	1	0.864	0.535	6

Note: 1 signifies the presence of a condition while 0 indicates its absence.

Upon examining Table 5, it is evident that the majority of rows indicate the absence of ‘scattered greening’ (SCAT1 = 0) and ‘central neighbourhood’ (CENTR1 = 0), while simultaneously featuring the presence of ‘sufficient initial GSA’ (GSA1 = 1). This observation implies that a non-scattered approach to greening, the non-centrality of neighbourhoods, and the existence of an initial GSA that fulfils government requirements are recurring elements within configurations that are deemed sufficient for green gentrification.

The process of logical minimisation results in multiple solution formulas for both the most parsimonious and intermediate solutions. Model ambiguity exists to a similar extent in both the intermediate and parsimonious solutions, which limits their interpretability in terms of causality. However, despite this limitation, the parsimonious solution, owing to its simplicity, can still be subjected to analysis and interpretation. Furthermore, by conducting a brief analysis of the enhanced conservative solution, which, although more complex overall, does not exhibit model ambiguity, the reliability of the inferences made can be enhanced. The enhanced conservative solution can be succinctly summarised in Boolean terms (see Annex 9 [online supplementary material]).

Several observations can be made regarding the differences and similarities among the various sufficiency terms. The most significant observation is that all six terms concur on the absence of neighbourhood centrality (∼CENTR1). Interestingly, it is notable that the conditions of ‘recreational function’ (FUNC1) and 'well-connected by public transportation’ (CONN1) are absent in the majority of basic expressions (rows with OUT = 1) for the outcome of 'gentrification’ (GI1). The recurrence of these conditions is surprising since their absence is generally assumed to diminish the attractiveness of a park, thereby impeding green gentrification. These unexpected elements within the sufficiency configuration further highlight the intricacy of the underlying mechanisms of gentrification. This suggests that the path leading to the outcome is determined by the combination of these conditions rather than their individual presence or absence. The enhanced parsimonious solution, along with its pertinent parameters of fit, can be summarised as depicted in Annex 10 (online supplementary information).

The initial sufficiency term (T1) has a consistency score of 0.878 and PRI parameters of fit of 0.714, both falling within the acceptable range, indicating that this configuration can be deemed sufficient for the occurrence of gentrification. However, the unique coverage, which measures the extent to which the individual solution path covers the outcome (Schneider, Oana, Thomann 2021), is relatively low at 0.041. Nevertheless, when comparing this path with the other solution terms, it is apparent that its unique coverage is the highest, making it empirically the most significant in relation to the outcome. The solution term can be interpreted as the pathway to green gentrification involving a combined approach of scattered greening (SCAT1), a negligible proportional increase in the overall supply of Urban Green Space (UGS) (∼UGSS_PC1), non-centrality (∼CENTR1), and adequate initial GSA1. The combination of scattered greening and an insignificant increase in the supply of Urban Green Space (UGS) (SCAT1*∼UGSS_PC1) contradicts initial theoretical expectations. At first glance, this combination is assumed to impede gentrification rather than promote it. However, this assumption holds true only within the context of green gentrification, which assumes that the occurrence of gentrification is necessarily a consequence of urban greening. It is plausible, though, that the gentrification explained by this solution term follows a different pattern and is not directly caused by Urban Green Infrastructure (UGI). The resulting outcome may be better explained by an alternative factor, not considered in this analysis, and potentially unrelated to greening.

All the solutions are in agreement regarding the second causal path (T2), which combines a non-scattered greening approach (∼SCAT1), a small average park size (UGSS_PC1), the absence of recreational functions (∼FUNC1), non-connectivity (∼CONN1), and non-centrality (∼CENTR1). This solution term exhibits an acceptable inclusion score of 0.827. However, the PRI value for this term is critically low at 0.462. A low PRI value suggests that this solution term does not effectively explain the occurrence of gentrification. Instead, it may serve as a better explanation for the non-occurrence of the outcome. In summary, all the solutions exhibit high and closely comparable coverage and PRI scores. The solution coverage for M1 and M2 is approximately 60%, meaning they explain around 60% of the gentrification outcomes (GI1). M3 and M4 have slightly higher solution coverage scores at 61%, indicating their ability to explain a slightly larger portion of gentrification outcomes.

4.4. Sufficient conditions for non-gentrification

Based on their inclusion consistency sufficiency and PRI parameters of fit (see Table 6), there are seven empirically tested combinations of conditions that can be categorised as potentially sufficient for the non-occurrence of gentrification.

Table 6. Potentially sufficient configurations (OUT = 1) for non-gentrification (∼GI1).

	Scattered greening (SCAT1)	Small UGS (UGSS_ NW1)	Significant UGS supply increase (UGSS_PC1)	Recreational function (FUNC1)	Well- connected by public transportation (CONN1)	Central neighborhood (CENTR1)	Adequate initial GSA (GSA1)	Output value (OUT)	Number of cases in configuration (n)	Sufficiency inclusion score (incl)	Proportional reduction in inconsistency (PRI)	Deviant cases consistency (DCC)
38	0	1	0	0	1	0	1	1	2	0.816	0.595	20, 30
39	0	1	0	0	1	1	0	1	1	0.867	0.555	13
40	0	1	0	0	1	1	1	1	2	0.848	0.683	3, 9
45	0	1	0	1	1	0	0	1	2	1	1	2, 24
47	0	1	0	1	1	1	0	1	1	1	1	10
55	0	1	1	0	1	1	0	1	1	1	1	22
126	1	1	1	1	1	0	1	1	1	1	1	21
128	1	1	1	1	1	1	1	1	1	0.858	0.627	11

Note: 1 signifies the presence of a condition while 0 indicates its absence.

All the listed configurations agree that the presence of small green spaces (UGSS_NW1 = 1) and transportation connectivity (CONN1 = 1) are essential components of the sufficient configurations for the outcome of ‘non-gentrification’ (∼GI1). To further clarify these observations and draw potential causal inferences, the enhanced solution formulas are now considered (see Annex 11 [online supplementary material]).

The first solution term (T1) combines the adoption of a non-scattered approach (∼SCAT1) with an insignificant increase in the overall supply of Urban Green Space (∼UGSS_PC1), good transportation connectivity (CONN1), and insufficient prior GSA1. One surprising aspect is the presence of ‘transportation connectivity’ (CONN1) in this configuration, as its absence was initially expected to result in non-gentrification or prevent the occurrence of gentrification. However, other elements of the configuration align with revised directional expectations. An inadequate initial GSA is believed to restrict a neighbourhood’s perceived attractiveness, particularly to members of the sustainability class. If the overall supply of UGSs is not significantly increased, the greening initiative does not trigger social upgrading and real estate valorisation. The studied empirical data demonstrates a consistency sufficiency of 0.896, indicating a general alignment with the subset relationship, although some deviations exist. The PRI value of 0.767 is well above the critical point, providing strong supporting evidence for the proposed sufficiency path. Furthermore, the proposed mechanisms explain a relatively low but comparatively significant portion of the set of non-gentrifying neighbourhoods, as indicated by the unique coverage of 0.117.

The second solution term (T2) introduces the aspect of a small average size of the newly created green space (UGSS_NW1) and the absence of recreational functions (∼FUNC1), in addition to the previously mentioned conditions of an insignificant increase in the total urban green space (∼UGSS_PC1) and good transportation connectivity (CONN1) as paths to non-gentrification. The overall configuration aligns with previously hypothesised relationships. With this combination of conditions, it is unlikely that the greening initiative can significantly improve the perceived attractiveness of the neighbourhoods, resulting in the absence of gentrification. However, the parameters of fit for this solution term are closer to the respective critical levels. The inclusion score of 0.754 is approaching the lower bound of 0.75, indicating a relatively high level of deviation from perfect subset relationships between the studied cases. Similarly, the PRI value of 0.560 is nearing the critical level of 0.5. Finally, the low unique coverage of 0.081 suggests that the explanatory power of the path to gentrification expressed by this solution term is limited. Therefore, fewer inferences will be drawn from this configuration of conditions.

The third agreed solution term (T3) implies that when a scattered approach is used to create small average-sized Urban Green Space (UGSS_NW1) without explicitly incorporating recreational functions (∼FUNC1), but with good transportation connectivity (CONN1) and located in a central neighbourhood (CENTR1), gentrification does not occur. All parameters of fit for this solution term are at an acceptable level, allowing us to examine the conceptual meaningfulness of this claim. The absence of a recreational park function and a small average park size were expected components in this configuration. However, the presence of transportation connectivity and neighbourhood centrality is surprising, as they were initially assumed to promote, rather than prevent, gentrification. Given their consistent presence in all sufficient configurations for non-gentrification, it is likely that there is an alternative explanation for the consistent inclusion of ‘transportation connectivity’ (CONN1). One possible explanation is that neighbourhoods with well-connected parks tend to be generally better integrated into the transportation network. It is possible that these well-connected parks and neighbourhoods have already undergone prior gentrification. Alternatively, a neighbourhood’s good connectivity may make it more attractive, independent of greening initiatives, and therefore the addition of UGSs only contributes a small marginal increase in perceived attractiveness. However, these explanations are hypothetical and would require systematic testing against prior gentrification patterns, taking into consideration other factors that may contribute to changes in a neighbourhood’s attractiveness.

The presence of multiple solution terms indicates that there are different paths to non-gentrification, and the adoption of a scattered greening approach (SCAT1) alone is not the only determining factor. The configurations of conditions in T4, T5, and T6 demonstrate that different combinations of factors can lead to non-gentrification outcomes. In T4, the combination of a scattered greening approach, small green spaces, a significant increase in the overall supply of UGSs, and transportation connectivity is suggested as a path to non-gentrification. This supports the idea of the ‘just green enough’ approach, where the scattering of small green spaces can prevent gentrification. However, in T5, the presence of recreational functions and park connectivity are introduced as additional factors alongside the scattered greening approach and a significant increase in UGS supply. This suggests that factors beyond just the scattered approach and UGS supply can influence the occurrence of gentrification. Similarly, in T6, the relative increase in the supply of UGSs is considered irrelevant, and the focus shifts to the presence of an adequate initial GSA. This highlights the importance of the initial GSA as a determinant of gentrification outcomes, regardless of the relative increase in UGS supply. These different configurations demonstrate that there are multiple paths to non-gentrification, and the combination of various conditions and factors determines the likelihood of gentrification occurring. The absence or presence of certain conditions in different solution terms suggests that the relationship between specific factors and non-gentrification is context-dependent.

The final minimisation (T7) proposes that newly created UGSs, characterised by a small average size (UGSS_NW1), leads to a significant increase in the overall supply of UGSs in the neighbourhood (UGSS_PC1). Furthermore, when UGSs are well connected to the public transportation system (CONN1) and located in a neighbourhood with an adequate initial GSA1, they do not result in gentrification, regardless of the presence or absence of scattered greening or specific park functions. It is important to note that this configuration exhibits the highest unique coverage (0.225) among the various sufficient paths, indicating its empirical significance. The pathway suggests that even in neighbourhoods with a sufficient initial GSA, gentrification does not occur when these specific conditions are combined.

All four parsimonious solution formulae exhibit similarly high consistency sufficiency and PRI values. M1 and M2 demonstrate comparable solution coverage scores of approximately 58% for the set of ‘non-gentrification’ (∼GI1). M2 has a slightly lower coverage score of 0.578, while M4 has a slightly higher coverage score of 0.606. Therefore, it remains uncertain which solution formula provides the most accurate explanation for the non-occurrence of gentrification following a UGI. However, the analysis highlights the robust causal influence of ‘transportation connectivity’ (CONN1) in preventing gentrification.

5. Discussion

5.1. Combinations of conditions

In the examination of whether necessary combinations of conditions exist for the occurrence of gentrification (GI1), no SUIN (i.e. sufficient but unnecessary part of a condition that is itself insufficient but necessary) combinations were identified as a superset of the outcome. Several configurations surpassed the consistency necessity threshold (0.9) (see Annex 7 [online supplementary material]), as anticipated, considering that combining a sufficient number of conditions through logical OR (+) can yield adequate consistency (Schneider, Oana, Thomann 2021). However, none of these configuration sets demonstrated coverage and RoN parameters of fit values within acceptable ranges (>0.5). Consequently, no configuration achieved the criteria of being both sufficiently consistent and empirically non-trivial, thus precluding the identification of any necessary combination. Consequently, it can be concluded that none of the studied conditions in combination are essential for the occurrence of gentrification subsequent to greening initiatives. This suggests that green gentrification can manifest in response to various types of greening initiatives and in diverse neighbourhood contexts.

The logical plausibility of good public transportation connectivity of a park, typically associated with enhanced perceived attractiveness, as a necessary factor for the prevention of green gentrification, is questionable. Through a different perspective, this finding could indicate that nearby public transport stations reduce the need for users to live near green spaces by allowing for easy access via public transport rather than on foot. Nonetheless, this hypothesis requires further investigation through qualitative primary data studies that examine park users’ opinions more closely.

Furthermore, it can be assumed that neighbourhoods with inadequate GSA would experience a greater marginal increase in perceived attractiveness when undergoing greening efforts. Consequently, insufficient initial GSA is anticipated to facilitate gentrification, rather than impede it. However, the findings reveal that non-gentrifying neighbourhoods overwhelmingly possess insufficient initial GSA or transportation connectivity. It is possible, that the availability of public green space, such as parks, is just one type of green infrastructure in a neighbourhood that contributes to the overall sense of ‘greenness’. Other elements, such as trees or private gardens, despite being inaccessible to the public, may still impact the perception of the level of greenness in a neighbourhood. A plausible explanation would then be, that the neighbourhoods in question have a limited supply of UGS but an abundant availability of other green elements. As a result, a UGI might only slightly increase the perceived greenness, and thus attractiveness, of the neighbourhood. This, however, is a highly hypothetical claim that requires empirical investigation of the availability of other green infrastructure in neighbourhoods.

An alternative interpretation is that a neighbourhood’s perceived attractiveness is so profoundly constrained by an insufficient initial GSA that a single greening initiative cannot sufficiently alter this perception within the studied timeframe to trigger gentrification. This appears particularly evident when the overall supply of green space does not significantly increase. Notably, the combination of conditions (∼UGSS_PC1+∼GSA1), while not confirmed as a necessary configuration due to the critically low RoN value of 0.420, demonstrates a high consistency value, tentatively supporting the aforementioned hypothesis.

Due to the similar solution coverage, but varying combinations of unique coverage, it is challenging to determine which combination of solution paths provides the most accurate explanation for the occurrence of gentrification. However, an important inference can be drawn from all the configurations. It appears that any form of greening (scattered or non-scattered, regardless of size or impact on the total supply of UGSs, with or without a recreational function, connected or disconnected by public transportation) can lead to subsequent gentrification if it is situated in the appropriate neighbourhood context. Specifically, the neighbourhood must be non-central and possess adequate initial GSA. This notion is empirically supported by the truth table and the conservative solution.

Perhaps the presence of adequate initial GSA reflects findings about the overall development of a neighbourhood; neighbourhoods that are generally more developed and have other desirable characteristics, such as an already adequate initial GSA in this case, are more likely to produce gentrification outcomes. This may also be related to the tipping point hypothesis introduced earlier in the paper, where a minimum level of existing green space is required for gentrification to occur. What remains puzzling is that this is only the case when an adequate initial GSA is combined with the condition of non-centrality.

A possible explanation for the inclusion of neighbourhood centrality as a condition in the sufficiency pathway to non-gentrification could be that centrally located neighbourhoods have a more limited response to greening initiatives. Centrality can be seen as a strong determinant of a neighbourhood’s existing attractiveness. Therefore, when a greening project is implemented, the increase in attractiveness it generates may be relatively small compared to the initial level of attractiveness. Additionally, centrally located neighbourhoods may have already undergone gentrification prior to the introduction of UGIs, independent of these initiatives. In such cases, the addition of a new park or greenway may not lead to significant social upgrading or real estate price increases, resulting in the absence of observable gentrification. Given these factors, the presence of neighbourhood centrality in the sufficient configurations for non-gentrification may indicate different dynamics and responses to urban greening in centrally located neighbourhoods, making them less likely to experience gentrification, even when green spaces are introduced.

5.2. Multiple determinants of gentrification

To conclude, no consistent, empirically relevant superset(s) of gentrification were identified, aligning with the conflicting observations of other researchers regarding the multiple determinants of gentrification. These findings indicate that green gentrification can manifest in various neighbourhood contexts and in response to diverse parks and greening strategies. Surprisingly, the presence of sufficient initial GSA emerges as a central component in unexpected pathways to green gentrification. Our hypothesis posited that areas with inadequate GSA would exhibit greater sensitivity to urban greening and thus be more prone to gentrification, particularly when accompanied by large-scale UGSs and substantial overall supply increments. However, the analysis output does not substantiate this hypothesis. Equally unexpected is the seemingly inconsequential influence of recreational functions and transportation connectivity, which were initially anticipated to foster gentrification. These findings indicate that not all instances of gentrification following greening can be adequately explained by the conditions examined, further reinforcing the notion that greening alone is unlikely to be the sole driver of gentrification.

The absence of a single necessary condition for the non-occurrence of gentrification was not surprising. However, it was unexpected that in cases where gentrification does not take place, the newly established park demonstrates good connectivity to public transportation or is located in a neighbourhood with insufficient initial GSA. Once again, the behaviour of the GSA condition deviated from expectations, as an inadequate prior GSA was anticipated to explain the occurrence rather than the non-occurrence of green gentrification. Similarly, the condition of connectivity challenges the assumption that well-connected parks serve as a stimulus, rather than an impediment, to gentrification. The pathways leading to non-gentrification defy directional expectations concerning connectivity and initial GSA. Additionally, it was expected that centrality would strongly contribute to the stimulation of gentrification, irrespective of other conditions. The recurrence of small UGSs in the causal paths for the negated outcome of ‘non-gentrification’ aligns with expectations based on the ‘just green enough’ approach.

Among the 33 UGIs implemented in Berlin between 2008 and 2016, approximately half of the neighbourhoods that underwent greening exhibited signs of initial gentrification. This observation provides empirical support for the notion that newly greened areas often experience gentrification, as suggested by previous studies (Anguelovski et al. 2018; Anguelovski and Connolly 2022; Curran and Hamilton 2012; Draus et al. 2019). However, it also confirms the findings of other researchers (e.g. Rigolon and Németh 2020) that greening does not invariably lead to gentrification. Thus, we conclude that the mere occurrence of urban greening is insufficient to guarantee subsequent gentrification.

Contrary to initial expectations, neighbourhood centrality neither played a necessary nor a sufficient role in determining the outcome of green gentrification. Thus, the findings do not support Rigolon and Németh’s claim that centrality serves as a predictor of green gentrification (Rigolon and Németh 2020). On the contrary, non-centrality consistently emerged as a superset associated with gentrification across the examined cases. Although causal inferences cannot be made from this observed superset relationship due to the significant skewness of the dataset, it suggests that the majority of UGIs are implemented in non-central neighbourhoods. This could be reflective of Berlin’s unique approach to greening or indicate a broader trend, where UGIs target non-central areas due to the availability of vacant land. In the instances where greening took place in a centrally located neighbourhood, it was not always accompanied by subsequent gentrification. This suggests the possibility that centrality plays such a dominant role in determining a neighbourhood’s overall desirability that the incremental attractiveness resulting from greening is relatively modest, thus failing to stimulate gentrification.

6. Conclusions

Urban greening plays a crucial role in fostering sustainable (re)development. Nevertheless, recent research has revealed a frequent association between the creation of green spaces and subsequent gentrification. This study aims to advance the existing body of knowledge on green gentrification by investigating the conditions that contribute to both the occurrence and non-occurrence of such outcomes. To achieve this objective, we employed an embedded case study approach and utilised fuzzy-set fsQCA. Through this method, we were able to identify the specific conditions, or combinations thereof, pertaining to green space creation and neighbourhood context that are necessary and sufficient for the manifestation or absence of gentrification.

It should be noted that additional factors play a decisive role in determining the overall susceptibility to gentrification. For subsequent investigations, it would be beneficial to introduce a preliminary step in the analysis that assesses the gentrification eligibility of the neighbourhoods, mirroring methodologies employed by fellow researchers in the field. This proactive measure would enhance the contextual understanding of the neighbourhoods under scrutiny and provide a nuanced foundation for comprehending the intricacies influencing the potential for gentrification. It is also important to emphasise that preventing urban gentrification requires a multifaceted approach that balances economic development with social equity. First and foremost, the implementation of robust affordable housing policies is essential. Municipalities should focus on creating and preserving affordable housing units, employing strategies such as inclusionary zoning, rent control, and community land trusts. Additionally, incentivising the development of mixed-income housing can foster diverse communities, mitigating the risk of displacement.

Individually, a scattered approach to greening or small park size alone is inadequate in preventing gentrification. However, when both conditions co-exist, gentrification does not occur. This holds true even when there is a significant increase in the overall supply of green space. Hence, the findings lend support to the ‘just green enough’ approach, as proposed by Wolch, Byrne, and Newell (2014). Nevertheless, this approach is not indispensable for averting gentrification. Alternative strategies to prevent green gentrification are likely to exist. Moreover, park size and the (non)scattered nature of greening do not appear to reliably predict the occurrence of gentrification.

Furthermore, the diverse combinations of conditions that prove sufficient for gentrification imply that a neighbourhood’s initial GSA is a significant determinant of the outcome. Specifically, it appears that different forms of greening (scattered or non-scattered, irrespective of size, resulting in a substantial or minor increase in the total supply of UGSs, with or without a recreational function, connected poorly or well by public transportation) can trigger the occurrence of gentrification. However, for gentrification to take place, the neighbourhood context must be suitable, namely, the neighbourhood should be non-central and possess adequate initial GSA. The latter condition may indicate that the tipping point hypothesis holds, whereby a certain level of green space (above the level deemed ‘adequate’) is required for a neighbourhood to be perceived as attractive and thus be eligible for gentrification outcomes. To draw further conclusions from these findings, future studies could look at the overall level of greenness, including trees or private gardens, in addition to the supply of public green space in the form of parks.

It is important to highlight that in this study, green gentrification is defined as gentrification occurring subsequent to urban greening, not necessarily caused by greening itself. Future research could encompass non-greened neighbourhoods, examining the presence of a newly established park as an additional condition to verify whether measured gentrification outcomes indeed stem from greening. This would enable causal inferences regarding the role of urban greening in generating gentrification outcomes. Moreover, expanding the number of cases studied would allow for the inclusion of additional conditions such as the previous level of gentrification. In future investigations, it might be beneficial to operationalise a neighbourhood’s level of integration within a city’s transportation system, in addition to, or instead of, proximity to a train stop, to provide further insights into the unexpected findings regarding transportation connectivity. Furthermore, exploring different cities or countries would yield more diverse sets of cases, thereby generating more reliable and generalisable findings.

It is possible that, contrary to the conception of most gentrification research, which aims to uncover patterns in conditions contributing to gentrification, the question is not what these patterns are but if there are any universal patterns. It is possible that, just like the perception of what constitutes an attractive neighbourhood is dependent on socio-contextual factors, the conditions leading to gentrification (and non-gentrification) are also context-dependent and thus vary between countries. For this, a cross-country and cross-cultural study is necessary to compare the conditions that lead to gentrification and non-gentrification. A large scale fsQCA investigating various gentrification-eligible neighbourhoods around the world could be conducted. Moreover, claims pertaining to the influence of scattered greening and neighbourhood centrality (both exhibiting skewness in this analysis) in shaping (non-)gentrification outcomes could be tested with a larger and more diverse set of cases.

The findings of this research offer valuable insights and critical recommendations for urban planners and policymakers in Berlin as well as other cities worldwide. The study underscores the importance of considering the broader neighbourhood context and adopting a thoughtful approach when planning UGIs. These considerations are especially relevant given the growing popularity of urban (re)greening strategies and the significant issue of gentrification. It is crucial for urban practitioners to carefully assess the potential risks of further exacerbating or accelerating existing gentrification processes when designing and implementing park creation projects. Building upon these findings, planners can conduct further research to develop predictor variables for green gentrification. One important predictor variable, as indicated by our results, is the initial availability of green space or the ratio of UGSs per resident. Additionally, planners can utilise the identified pathways leading to non-gentrification outcomes as guidance for site selection and greening approaches, aiming to minimise the likelihood of gentrification. Furthermore, to mitigate the ongoing risk of increased real estate values and consequent displacement of economically vulnerable groups, planners must ensure the provision of affordable housing in newly greened neighbourhoods (Immergluck and Balan 2018; Rigolon and Németh 2020).

Acknowledgements

The authors express their gratitude to the RWI - Leibniz Institute for Economic Research (www.leibniz-gemeinschaft.de/en/institutes/leibniz-institutes-all-lists/rwi-leibniz-institute-for-economic-research) for supplying the relevant data.

Disclosure statement

The authors report there are no competing interests to declare.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/09640568.2024.2343059.