Introduction

Both social science and psychological research find increasing evidence that spending time in green spaces impacts positively on subjective well-being. For example, positive associations have been reported for seeing green (Kaplan Mintz et al., 2021), living close to public green spaces (Nutsford et al., 2013), having a private garden (Lehberger et al., 2021), and engaging with the natural environment (Biedenweg et al., 2017). Moreover, a recent study finds that the frequency of green space visits is associated with lower odds of using psychotropic medication (Turunen et al., 2023). However, there is little research on how using green spaces impacts on individual well-being in societal crisis situations.

In this article, we take the coronavirus pandemic as an opportunity to examine the role of green spaces for well-being during a specific crisis situation. Policies designed to contain the number of infections during the pandemic had far-reaching effects on all segments of the population, particularly in the spring of 2020 and winter of 2020–2021. The first so-called “lockdown” in Germany started on March 22, 2020, and lasted until May 2020 after which most restrictions were withdrawn step by step. The most severe lockdown measures were school closures, contact restrictions, cancellation of all cultural and sports events, business closures in the gastronomic and retail sectors, as well as short-time work and working from home for many employees (Hannover, 2022). During the second lockdown, starting on November 2, 2020, intensified on December 16, 2020, and lasting until May 2021, similar measures were implemented, although schools remained open. The drastic measures had a massive impact on everyday life in Germany. In particular, they reduced the range of public places available for leisure activities or social meetings. Public greenspaces remained one of the few places people could still access.

In the context of the extensive restrictions in Germany during the coronavirus pandemic, several studies find that overall life satisfaction—as a possible indicator of subjective well-being—declined during the initial lockdown (Ahlheim et al., 2020; Handschuh et al., 2021; Zacher & Rudolph, 2021). However, whether green space usage contributed to making the decline in individual life satisfaction less pronounced during the pandemic is an open question. Such an effect seems likely, because research indicates that green space usage affects life satisfaction positively through mechanisms of stress reduction (Kaplan, 1995; Ulrich, 2014), social gatherings (Baumeister & Leary, 1995; Weinstein et al., 2015), and physical activity (Claßen & Bunz, 2018). Indeed, studies focusing on the period of the coronavirus pandemic find a positive association between green space usage and subjective well-being (e.g., Kaplan Mintz et al., 2021; Mayen Huerta & Utomo, 2021). However, most of these studies rely on cross-sectional data (sometimes with retrospective elements), and their analyses do not permit strong inferences about causality. In contrast, we use longitudinal data and panel regression models that are better suited to draw conclusions about cause–effect relationships.

Our data were collected as part of a panel study in the cities of Hannover and Braunschweig in Lower Saxony, Germany, and they are based on random samples from the population registers of the two cities. A total of 561 adults were interviewed twice: in fall–winter both 2019–2020 and 2020–2021. The first round of interviews took place before the start of the pandemic; the second, at the beginning of the second major wave of Covid-19 infections (Appendix Fig. 5). This data structure allows us to gain insight into the impact of behavioral changes in the usage of public green spaces on the development of life satisfaction during the coronavirus pandemic.

The article addresses the following research question: Does a change in the use of public green spaces during the coronavirus pandemic affect individual life satisfaction? It also explores whether specific types of usage (such as doing sports or meeting friends) exert different effects on life satisfaction.

Our article is structured as follows: First, we discuss the concepts of life satisfaction and subjective well-being, and present possible explanations for their relationship to green space usage. Then we describe the data and variables, before explaining our methodological approach. Using difference-in-difference models, we examine the extent to which a change in green space usage during the coronavirus pandemic affects life satisfaction on the individual level. Finally, we discuss limitations of our study and perspectives for further research.

Theoretical Background and State of Research

General Life Satisfaction and Subjective Well-Being

According to Diener’s influential definition, subjective well-being is composed of positive affect, negative affect, and general life satisfaction, with the latter referring to a global cognitive evaluation in contrast to the emotional elements of subjective well-being (Diener, 1984; Diener et al., 1999). Theoretically and empirically, general life satisfaction is related to income and unemployment, social contacts and social integration, as well as health (e.g., Diener et al., 1999, 2018). Although general life satisfaction has been shown to be relatively stable on an intrapersonal level (Eid & Diener, 2004), it can nevertheless be affected by major life events such as marriage, divorce, or childbirth (Diener et al., 2013) as well as by severe changes on the societal level such as economic crises (Delle Fave, 2014). Because the coronavirus pandemic represents a societal crisis that had a strong impact on people’s everyday lives, it can be expected to have had an impact on general life satisfaction.

Between 2006 and 2018, average life satisfaction in Germany increased almost continuously (Priem et al., 2020). However, the restrictive measures to reduce the number of Covid-19 infections impacted on employment, social contacts, and public life. This suggests an end or interruption to the positive development of life satisfaction. Using data from the National Education Panel Study (NEPS), Handschuh et al. (2021) find that average life satisfaction declined among adults during the first lockdown. This finding is confirmed by Ahlheim et al. (2020). In contrast, based on the German Socioeconomic Panel, Entringer and Kröger (2021) find that although average life satisfaction did not change significantly during the first lockdown, it did decrease during the second. Similarly, Wettstein et al. (2022), using the German Ageing Survey, find no significant difference in life satisfaction in June–July 2020 (at the end of the first wave of infection) compared with 2017 and 2014. Hence, most, though not all, studies for Germany find evidence that the coronavirus pandemic had a negative effect on life satisfaction.

Use of Green Spaces and Subjective Well-Being

What are the mechanisms by which spending time in green spaces can affect life satisfaction? In the following, we explicate the main mechanisms suggested in the relevant literature and discuss their relevance for the situation during the coronavirus pandemic.

Theoretical discussions on the effects of green spaces mostly address the positive effect on subjective well-being in general terms without explicit reference to life satisfaction. The first argument refers to psychological stress that is seen as an important determinant of subjective well-being (Diener et al., 1999; Sonnentag & Fritz, 2015). Following attention restoration theory (ART) (Kaplan, 1995), spending time in green spaces is expected to reduce psychological stress via natural stimuli that lead to recovery from directed attention. Attracting intrinsic attention through sensory stimuli enables the neurocognitive mechanism that has become exhausted through processes of directed attention to recover. Complementing this, Ulrich (2014) suggests a psycho-evolutionary argument with stress reduction theory (SRT). This assumes that psychophysiological stress can be reduced by natural stimuli. Accordingly, by indicating the absence of threats, attributes of natural environments elicit innate adaptive responses in humans. For example, spatial expansiveness, visual perceptions of water, or patterns and structures of green environments can evoke positive emotions that block negative thoughts and emotions, thereby alleviating stress responses (Ulrich, 2014). In sum, whether based on either ART or SRT, psychological research suggests that individuals who spend much time in areas close to nature experience greater benefits to subjective well-being and health than if they were to spend time in built-up environments with a lower recreational quality.

This mechanism is highly relevant to our research question: The first wave of infection during the coronavirus pandemic was accompanied by several potential stressors such as the risk of infection with a mostly unknown virus for oneself and those close to one, restrictions such as school closures, contact restrictions including bans on visits to retirement homes, a sudden increase in hours working from home, and the loss of opportunities for leisure. Thus, possibilities for stress reduction gained increasing importance with the onset of the pandemic. Public green spaces were an important resource in this regard, possibly contributing to stress reduction and buffering a decrease in life satisfaction during the pandemic. Vice versa, the theories suggest that reduced green space usage exacerbates negative effects of the pandemic on life satisfaction.

In addition to their possible direct stress-reducing effect, green spaces can also provide space for maintaining social contacts and thereby promoting social integration and the sense of belonging. The latter are basic human needs (Baumeister & Leary, 1995), and their satisfaction has a positive impact on subjective well-being (Cramm et al., 2013). Spending time in green spaces can enhance feelings of belonging and trust, both through meeting with friends or acquaintances and through sharing space with other urban residents. Moreover, social cohesion in a neighborhood can be strengthened by the joint use of public spaces (Weinstein et al., 2015). Because social cohesion is associated positively with mental health (Rios et al., 2012) and mental health is a predictor of subjective well-being (Joshanloo & Nosratabadi, 2009), green space usage can contribute to both higher subjective well-being and higher life satisfaction by intensifying the sense of belonging (Markevych et al., 2017).

This mechanism is also relevant during the coronavirus pandemic. The primary aim of the restrictive coronavirus policy was to reduce social contacts, because the fewer contacts people have, the lower the risk of virus transmission. This is especially true for indoor meetings, because transmission occurs primarily via aerosols that accumulate during prolonged stays indoors, especially while people are speaking (Zhang et al., 2020). The legal restrictions as well as the awareness of one’s own risk of infection led to strong contact reductions during lockdowns. At the same time, meetings with friends and acquaintances were shifted to the outside—that is, there was an increase in the importance of using public green spaces for social life. Hence, it is also plausible that green space usage had a positive effect on life satisfaction during the coronavirus pandemic through the mechanism of social integration.

In addition, green space usage is associated with physical activity, which can range from simple walking to doing sports. It is undisputed that physical activity has a positive effect on subjective well-being (An et al., 2020; Claßen & Bunz, 2018). Studies disagree, however, on the specific underlying mechanisms. On the one hand, it is argued that biochemical processes lead to the release of the mood-enhancing hormone serotonin. On the other hand, physiological explanations argue that the increased body temperature caused by movement leads to relaxation and improved mood in the subsequent recovery phase (Fox, 1999). Regardless of the biological mechanism, regular physical activity can thus have a positive effect on subjective well-being and life satisfaction.

German coronavirus policy included the suspension of activities in sports clubs. In particular, team and indoor sports were not compatible with the government’s goals of reducing the number of infections. Individual sports in public spaces, such as running in the park, were still allowed and thus gained increased relevance for the positive effects of physical activity on subjective well-being (Brailovskaia et al., 2021).

All three mechanisms suggest that green space usage might affect the development of general life satisfaction during the coronavirus pandemic. This leads us to the following hypotheses: Individuals who increase their use of green space during the coronavirus pandemic will experience a weaker reduction in their general life satisfaction compared to those who do not change their use of green space (Hypothesis 1a). In contrast, those who reduce their use of green space will experience a stronger reduction of life satisfaction compared to those who do not change their use of green space (Hypothesis 1b).

The mechanisms described above also suggest that the impact might depend on how green space is used. For example, the effect on life satisfaction of using public green space for doing sports might differ from that of meeting with friends. Because we do not have any theoretical assumptions about the effect sizes of different types of green space usage, we simply explore whether effects differ.

Previous Studies on Green Spaces and Subjective Well-Being in the Coronavirus Pandemic

Several recent empirical studies have indicated positive associations between green spaces and subjective well-being during the coronavirus pandemic: Cross-sectional regression analyses find positive correlations between public green space usage and well-being during the pandemic for Israel (Kaplan Mintz et al., 2021) and Mexico City (Mayen Huerta & Utomo, 2021). The latter study also shows that continuing or starting to use urban green spaces during the coronavirus pandemic is positively related to sustained or increased subjective well-being on the individual level. Another study, conducted in Portugal and Spain, reports positive associations between mental health and the maintenance or increase of exposure to nature (Ribeiro et al., 2021). For Germany, Lehberger et al. (2021) show that the availability of a private garden is positively related to subjective well-being during the coronavirus pandemic. Additionally, the study finds a positive effect on life satisfaction of increasing the time spent outside in nature for leisure. However, such an effect is not found for increasing the time spent outside doing sports. A UK panel study looking at subjective health and subjective well-being during the first lockdown (March/April 2020) and the time after the lockdown (June/July 2020) finds that the subjectively perceived availability of green space as well as access to a private garden had positive effects on health and well-being (Poortinga et al., 2021).

These five studies vary in many details. Not only is green space usage or exposure to nature operationalized in different ways, but the measurement of subjective well-being also varies. Only the study by Lehberger et al. (2021) includes a measure of general life satisfaction; all other studies refer to other measurements. At the same time, all studies use some type of multiple regression model controlling for variables that might bias the effect between the main independent and the dependent variable. Some research (Lehberger et al., 2021; Mayen Huerta & Utomo, 2021; Ribeiro et al., 2021) included a longitudinal perspective and examined how an increase or maintenance of green space usage (compared to a decrease) affects subjective well-being. These studies are somewhat better able to assess the causal effect of green space use (or exposure to nature) on subjective well-being than those that apply a purely cross-sectional design. However, there are still limitations, because these three studies mostly ignore the distinction between an increase or a maintenance of green space usage, and they measure only change in the independent, but not the dependent variable. They also do not employ panel regression models, although these are better suited to control for the unoberserved heterogeneity that might distort the results. Therefore, in our own study, we used panel regression models (specifically, difference-in-difference models) to examine in detail the association between changes in green space usage and changes in life satisfaction during the coronavirus pandemic.

Data and Method

Our data were collected as part of a panel study conducted at the Georg August University of Göttingen in fall/winter of 2019–2020 and 2020–2021 using standardized online surveys in the cities of Braunschweig and Hannover. The first step was to contact 5,988 randomly selected individuals between 18 and 70 years of age by mail (sampled from population registers). Following two reminders after three and nine weeks, a total of 1,193 persons took part in the first panel wave of the survey (response rate: 19.9%). Of these, 561 individuals (47%) also participated in the second wave in 2020.Footnote 1 In the survey, we collected information on sociodemographic characteristics, green space use, health, relocation behavior, and (in the second panel wave) subjective assessments regarding the coronavirus pandemic. To be able to analyze as many cases as possible, we applied multiple imputations for missing values (Table 1 in Appendix 1) using 10 generated data sets (Rubin, 2010). Unlike other methods of dealing with missing data such as listwise deletion or single imputation, multiple imputation produces unbiased estimates, preserves the associations of the dataset, and incorporates the uncertainty associated with predicting the values (Young et al., 2011). Furthermore, multiple imputation does not require missing data to be completely random. Instead, missing values only need to be distributed randomly after their dependence on other variables is taken into account. To meet this assumption, we included all available information in the imputation model.

A particular advantage of our panel study lies in the timing of the panel waves. Whereas the first panel wave (from August 31, 2019 to January 14, 2020) took place clearly before the start of the pandemic, the second wave—from October 13, 2020 to January 17, 2021—falls within a period of dynamic Covid-19 infection development. At the end of the first panel wave, only few reports about the coronavirus virus spreading in China had been published in German media. In contrast, the second survey period coincides with the beginning of the second major wave of infection. On November 2, 2020, drastic restrictions on public life came into effect, so that sporting and cultural events were canceled, gastronomic establishments were closed, and meetings in public were permitted with only a maximum of two households and 10 persons. Unlike during the first wave of infections, schools and kindergartens remained open.

On average, respondents were 47.3 years old at the time of the first survey, 50.3% were female, and they lived within an average of 432 m (with a standard deviation of 250 m) from the nearest green space with a minimum size of one hectare. Braunschweig and Hannover are the largest cities of Lower Saxony with a high proportion of green spaces. Social disparities with regard to the availability of green spaces at the place of residence are therefore not very pronounced. Nonetheless, due to city-specific characteristics and higher educated people being overrepresented in our sample, our results do not generalize to all cities in Germany (Kohler et al., 2019).

Variables

To measure general life satisfaction, we used a question similar to that used in other surveys (e.g., World Value Survey, Socioeconomic Panel). “In general terms, how satisfied are you with your life at present?” Respondents could indicate the extent of their satisfaction on a Likert scale from 0 to 10. Jovanović and Lazić (2020) emphasize the validity and reliability of this operationalization. Our dependent variable “change in life satisfaction” was calculated as the difference in scores between the two survey waves, and thus represents the development of life satisfaction between the survey dates.

The key independent variable is change in green space usage during the coronavirus pandemic. It refers to parks, woods, and other green spaces in the neighborhood and it was asked retrospectively, “Did you use these green spaces more or less than before in recent months, due to the coronavirus crisis?” Respondents could choose “more often than usual,” “the same as usual,” or “less often than usual” on a 3-point scale. In the same way, the change in frequency was surveyed for different types of usage: “As a result of the Corona crisis, have you used parks, woods, or other public green spaces for these activities more or less in recent months than before? (1) … to relax, (2) … to meet with acquaintances, (3) … to do sports.”

In addition, we applied two other operationalizations of the change in green space usage. First, in both survey waves, participants were asked about the frequency of green space visits on an ordinal 5-point scale: “How often do you use parks, woods, and other green spaces in your immediate neighborhood?” Response options were “never,” “once a month at most,” “several times a month,” “several times a week,” and “daily.” In the same way, participants were asked how often they used green spaces to relax, to meet acquaintances, or to do sports. The difference in responses between the survey waves allowed us to determine whether respondents used green spaces more often, less often, or in the same frequency compared to the previous year.

Second, we constructed a measurement from the following question that was also asked in both survey waves: “On average, how many minutes do you spend per day in parks, woods, or other public green spaces on weekends?” Because the variable refers to weekends, it largely excludes differences due to employment status. Again, a difference variable was constructed to capture whether green space usage increased, decreased, or remained constant between waves. Only differences of at least 20 min per day were counted as a change. We are not aware of any theoretical arguments for thresholds above which an effect on life satisfaction can be expected. Therefore, we chose a threshold that resulted in similar distributions of reduction, increase, and stagnation of green space usage between different operationalizations (Fig. 1).

All three questions refer to typical behavior, allowing us to measure lasting changes in behavior. However, the three operationalizations of the independent variable differ in several ways: First, the retrospective question about changes in green space use explicitly refers to the response to the “Corona crisis”, asking about the “last few months”. Thus, in contrast to the other operationalizations, a (vague) time period is explained and it is also implied that the Corona pandemic caused the behavioral change. Whereas the questions for the difference variables refer to typical behavior, they aim at the status quo of the frequency or duration of green space use and thus correspond well with the status quo measurement of life satisfaction. The question on the frequency of use also indicates a longer period of time due to the answer categories “once a month at the most” and “several times a month” (Kelle et al., 2021). The question about the duration of use in minutes, on the other hand, does not make this reference, so it remains unclear to which period respondents should refer. However, given the positioning of the question in the questionnaire, a longer reference period is suggested here as well.

The first operationalization, the retrospective recording of changes in green space use, is somewhat problematic from a methodological point of view. Personal assessments of past events are subject to cognitive biases (recall bias), which are more pronounced during emotionally loaded periods and events (Bell et al., 2019; Blane, 1996). The onset of the coronavirus pandemic represented an emotional phase for many people with feelings of fear and uncertainty, so statements about this time may be inaccurate.

The two other operationalizations are not plagued by this problem. However, calculating the difference between the frequencies of green space usage is blurred because it refers to ordinal variables. The difference in the (metrically measured) minutes spent in green spaces appears methodoldogically more sound.

Methodological Approach

The purpose of this study was to determine whether a causal relationship existed between individual changes in green space usage and changes in life satisfaction during the coronavirus pandemic. Much of the research to date has been based on cross-sectional designs, making causal inference possible only under strong theoretical assumptions. Panel data are better suited, because individual-level changes can be observed over time, and these provide a better basis for causal conclusions (Gangl, 2010). To examine the causal relationship between a change in individual green space usage and life satisfaction during the pandemic, we used difference-in-difference models. Such models allowed us to estimate causal effects by comparing temporal changes in a variable of interest for a treatment (T) and a control (C) group. In line with our hypotheses, we formed a control group consisting of all respondents whose frequency of green space usage did not change during the pandemic and two treatment groups consisting of respondents who used green spaces either more frequently or less frequently during the pandemic compared to the time before. Causal estimations were possible, because the temporal comparison within the groups precluded possible distortions by unobserved time-constant characteristics, whereas the between-group comparison excluded unobserved time-varying confounders that affected both groups equally. Our difference-in-difference estimator was

$$DiD= \left({\beta }_{green, {t}_{1}}^{T}-{\beta }_{green,{t}_{o}}^{T}\right)-\left({\beta }_{green, {t}_{1}}^{C}-{\beta }_{green,{t}_{o}}^{C}\right)$$
(1)

where \({\beta }_{green}\) is the effect of green space usage on life satisfaction, t denotes the time, and T and C are the treatment and control groups.

However, this approach was based on the theoretical assumption that the interviewees are subject to the same probability of being in one of the treatment and control groups. It was also assumed that the life satisfaction of the two treatment groups would have developed identically to that of the control group if they had not received the treatment. In order to come as close as possible to this theoretical assumption, we used an inverse probability of treatment weighting (Chesnaye et al., 2022). In keeping with a causal theory perspective, weighting was based on variables expected to influence the development of life satisfaction and changes in green space usage during the coronavirus pandemic (Brookhart et al., 2006; Cinelli et al., 2022; Wyss et al., 2013). A maximum weight of 5 was chosen to avoid overly strong influences from individual cases.

Taking into account results of previous research, we weighted by employment status (Poortinga et al., 2021), subjective health (Jovanović et al., 2021), age (Wettstein et al., 2022), level of education, marital status (Haller & Hadler, 2006), income (Boes & Winkelmann, 2010), availability of green space at the place of residence (Kley & Dovbishchuk, 2021), gender, number of children, births between the survey waves (Angeles, 2010), and possible coronavirus-related job changes (such as working from home and short-time work). Because a greater proportion of daily life took place in one’s own home during periods of lockdown, we further weighted for satisfaction with home amenities and size. Given the high stress levels in some essential occupations during the pandemic (Santamaría et al., 2021), we further included whether the respondent worked in an essential occupation as a binary variable. For the inverse probability of treatment weighting, we used the information from the first survey wave, except for variables that were surveyed only in the second wave such as changes in one’s job.

Employment status was categorized as retirement, domestic work, dependent employment, self-employment, and unemployment, whereas household-equivalent net income was measured metrically. Level of education was included in the weighting procedure with three categories: lower secondary school or no certificate, intermediate secondary school, and higher education. Age was included metrically as well as squared. Binary variables informed whether there were children in the household, distinguishing the ages 0 to 6 years, 7 to 13 years, and 14 to 17 years. In addition, subjective health was included on the basis of an 11-point Likert scale. Indicators for working from home and short-time work were used as binary variables. The information came from the survey question “Has the coronavirus pandemic changed your daily work life? If yes: Which of the following changes in your working life apply to you? (1) I am working reduced hours because of the coronavirus pandemic. (2) I am working from home for the first time because of the coronavirus pandemic. (3) I am working from home more often because of the coronavirus pandemic.” Satisfaction with home amenities and size was measured on an 11-point Likert scale. Finally, the availability of green space at the place of residence was operationalized by the distance to the nearest green space of a minimum size of one hectare. This was determined as the shortest distance via footpaths to the nearest green space, calculated from geodata analyses. For univariate statistics on the weighting variables, see Table 2 in Appendix 1.

Using these variables, we first determined the probability of belonging to the treatment group using logistic regression

$$P\left(T=1\mid X\right)=\frac{1}{1+{e}^{-{\left(green\right)}^{\prime}}}$$
(2)

where

We then calculated the inverse probability of treatment weights by

$$w=\frac{1}{P(T=1|X)}$$
(3)

if a person belonged to the treatment group and

$$w=\frac{1}{1-P(T=1|X)}$$
(4)

if a person belonged to the control group. Using these weights, the effect of green space usage on life satisfaction, \({\beta }_{green}\), was estimated using weighted least squares regression for the treatment and control group and both time points, and this was used to calculate the difference-in-difference estimator (see Eq. 1).

Weighting was carried out separately for each model focusing on increase and decrease in green space usage. The same was repeated for the different types of usage, so that “more frequent/less frequent use of green spaces for meeting acquaintances/relaxing/doing sports due to the coronavirus pandemic” represented the treatments in the difference-in-difference models. Bivariate correlations showed that the weighting results in the control variables are associated only slightly (r < .1) with the treatment variables (see Figs. 6, 7, 8, 9, 10, and 11 in Appendix 3).

Results

We begin with a brief look at the descriptive statistics before presenting and discussing findings on the difference-in-difference models. Table 3 in Appendix 1 shows the distributions and means of all dependent and independent variables included in the models. Consistent with some of the previous research (Ahlheim et al., 2020; Handschuh et al., 2021), we find that respondents’ overall life satisfaction decreased significantly (p < .001) between 2019 and 2020 from 7.5 (SD = 1.7) to 7.2 (SD = 1.8) points (on a scale of 0 to 10). In both survey waves, life satisfaction is correlated with the level of green space usage (frequency variable: 0.13 in the first panel wave vs. 0.07 in the second wave; minutes variable: 0.09 vs. 0.11; all p < .001). These positive associations are also evident in linear multivariate cross-sectional models of both survey waves (Tables 4 and 5 in Appendix 1) and confirm results from previous studies (Kaplan Mintz et al., 2021; Mayen Huerta & Utomo, 2021).

For a large proportion of respondents, public green space usage changed over the two panel waves (Fig. 1). On average, public green spaces were visited more frequently and for longer periods during the coronavirus pandemic than before. The results from different operationalizations correlate significantly with each other (r values between 0.18 and 0.37).

Fig. 1
figure 1

Distribution of change in green space usage by operationalization

Full size image

In particular, green space usage for relaxation increased according to the retrospective variable. A total of 43.2% of respondents reported using green spaces for relaxation more often than before the beginning of the pandemic, whereas only 9.5% did so less often. On average, green space usage for meeting acquaintances and doing sports did not change significantly. The retrospective variables for change in green space usage asking for the types of usage correlate significantly with each other at r > .37. In the alternative operationalization using the difference in frequency of usage, the same trends appear, but they are less pronounced. Accordingly, 32.7% used green spaces more frequently to relax than in the previous year, whereas 17.6% used these places less frequently to relax.

Figure 2 shows the bivariate relationship of quantitative change in green space usage to change in life satisfaction between panel waves. Results are reported separately for the different operationalizations. Based on the retrospective variable, we find that those who reduced green space usage show a relatively large decrease in life satisfaction (approximately − 0.8 points), whereas those who increased usage show a somewhat smaller decrease (approximately − 0.4) that differs significantly from users with no change in behavior (approximately − 0.1). Using the difference in minutes spent in green spaces, we again observe the largest decrease in satisfaction for those who reduced the time, whereas the smallest decrease in life satisfaction is recorded for the group with increased time spent in green spaces. Turning to the difference in frequency of usage, there is evidence of an opposite relationship: Those with reduced frequency of green space usage show the smallest change in life satisfaction, whereas there is little difference in average trends between stable and more frequent users. Thus, the pattern of results is not consistent across operationalizations.

Fig. 2
figure 2

Average change in life satisfaction according to change in usage of public green spaces—as reported by different operationalizations

Full size image

The weighted difference-in-difference models (Fig. 3 and Table 6 in the Appendix) show hardly any significant effects. With regard to increased usage of public green spaces, we find a significant positive effect only for the difference-in-minutes variable,Footnote 2 whereas the effects for the difference-in-frequency variable and the retrospective indicator are either close to zero or positive and not significant. With regard to reduced green space usage, the only significant negative effect is for the retrospective measurement. The other models show no significant relationships.

Fig. 3
figure 3

Difference-in-difference estimators: Influence of changing green space usage on the development of life satisfaction. Coefficients are reported separately for different operationalizations of the independent variable

Full size image

It is also conceivable that, in addition to changes in green space usage in the period studied, the level of green space usage also has an effect on the development of life satisfaction. Accordingly, a high level of usage over both survey time points would have to associated with a less pronounced reduction in satisfaction. However, we find no empirical evidence for this (Fig. 13 in Appendix 4).

The difference-in-difference models for the different types of usage (Fig. 4 and Table 7 in Appendix 1) show that the retrospectively reported reduced usage of green spaces for relaxing, meeting acquaintances, and doing sports each have significant negative effects on life satisfaction. The coefficient is strongest for reduced usage for relaxing (r = − .81, p < .001), whereas it is slightly less strong for meeting acquaintances (r = − .70, p < .001) and doing sports (r = − .60, p < .05). However, these results are not robust when an alternative operationalization is applied. Using the difference-in-frequency variable, the difference-in-difference estimators remain nonsignificant for all types of use. Moreover, regardless of the operationalization, we find no effects of intensified green space usage on life satisfaction for any type of usage.

Fig. 4
figure 4

Difference-in-difference estimators: Influence of changes in green space usage (GSU) per type of usage on change in life satisfaction. Coefficients are reported for different operationalizations of the independent variable

Full size image

In sum, we do not find robust support for Hypotheses 1a and 1b formulated above. According to our exploratory analyses, the size of effects of green space usage on life satisfaction does not differ between specific types of usage.

Summary and Discussion

Our panel study confirms previous findings (Ahlheim et al., 2020; Handschuh et al., 2021) that life satisfaction decreased during the coronavirus pandemic. At the same time, and contrary to the results reported by Geng et al. (2021), we find that green space usage increased overall during the pandemic. Similar to previous studies (Kaplan Mintz et al., 2021; Mayen Huerta & Utomo, 2021; Poortinga et al., 2021; Ribeiro et al., 2021), we find—in a cross-sectional perspective—that life satisfaction correlates positively with the level of green space usage. Typically this relationship is interpreted causally; it is assumend that more green space usage was able to attenuate the reduction of life satisfaction during the coronavirus pandemic. This may be due to the stress-reducing effect of spending time in green spaces (Kaplan, 1995), of physical activity (Fox, 1999), or of experiencing social cohesion (Weinstein et al., 2015).

However, studying causal effects within a cross-sectional research design can lead to biased results due to unobserved heterogeneity. To examine the causal relationship more appropriately, we applied panel data models (specifically, difference-in-difference models) to study whether people who increased or decreased their green space usage during the coronavirus pandemic reported corresponding changes in general life satisfaction. Given the widespread changes in the amount of green space usage, the coronavirus pandemic presents an opportunity to apply such a methodological framework. Contrary to our expectation, however, the results of our analyses do not provide robust support for our hypotheses that a change in green space usage would be associated with a corresponding change in life satisfaction. The results of two of the three operationalizations of individual change in green space usage provide only partial hints of such effects: Based on a question asking retrospectively about how much respondents changed their green space usage, results indicate a significant reduction in life satisfaction if green space usage was reduced (compared to no change). However, a reported increase of green space has no significant effect on life satisfaction. Conversely, based on the difference measure employing minutes of green space usage reported in the two panel waves, we find only a significant positive effect of an increased duration of green space usage on life satisfaction, but no corresponding effect of a decreased duration.

There could be several reasons for the lack of clear evidence for the expected associations between a change in green space use during the coronavirus pandemic and general life satisfaction. Methodologically, we believe that the third of our operationalizations is the one least suited, because using the difference between values of two ordinal variables is problematic (see “Variables” section). The lack of empirical support for our hypotheses based on this operationalization might result from this problem.

However, there might also be theoretical reasons for the overall weak support for our hypotheses. One reason might be that the causal relationships leading people to change their amount of green space use during the pandemic is more complex than we theorize. Our hypothesis might be too simplistic, because it assumes that there is no recursive causal relationship between green space use and life satisfaction. But it is conceivable that people with higher life satisfaction are more likely to maintain or even increase their level of green space usage during the pandemic.Footnote 3 Conversely, it is also conceivable that some people increased their use of green spaces because their life satisfaction had decreased as a result of the pandemic. However, the regression models assume only one direction of causality—from green space use to life satisfaction—and, thus, are not able to capture the possible complexity of causal relations.

Unobserved influences may also be a cause of our inconsistent findings. Although having pre-coronavirus data and data gathered during the coronavirus pandemic along with using panel data models are strengths, this does not rule out methodological problems. Particularly, the risk of unobserved influences is high, because the health risk from the virus and the drastic containment measures brought so many changes to people’s lives that it is difficult to weight for all relevant factors in the inverse probability of treatment weighting applied in our difference–difference models. We know, for example, that social relationships play a prominent role in subjective well-being during crises (Delle Fave, 2014). However, beyond the information on partnership status and children that we applied in the inverse probability weighting, we did not collect any other data on this topic in our survey.

More general limitations of our study also need to be mentioned: Having only two survey waves, we could not use our data to analyze more long-term influences of green space usage on life satisfaction. In future studies, it would be worth examining longer time periods to learn more about the development of the effect of behavioral changes on life satisfaction. Furthermore, our data originate from two large cities in Lower Saxony, so that results are not easy to generalize. Moreover, city comparisons between Braunschweig and Hannover were not possible due to low case numbers. The extent to which city-specific characteristics such as the distribution of green spaces play a role in life satisfaction could be a subject for future research.

To the best of our knowledge, we present the first empirical study using panel data models to examine the association between changes in green space usage and trends in overall life satisfaction during the coronavirus pandemic. The application of difference-in-difference models with weighted control and treatment groups enabled us to isolate the effect of changes in green space use on the development of life satisfaction. According to our data and analyses, there is only weak evidence that individually altered green space usage affected the development of life satisfaction in the coronavirus pandemic. Nevertheless, we find circumstantial evidence for associations, so our article also highlights the need for further research using high-quality data.