The COVID-19 crisis does not diminish environmental motivation: Evidence from two panel studies of decision making and self-reported pro-environmental behavior.

The literature shows that threats unrelated to environmental problems can shift attention away from these problems and affect pro-environmental behavior. It is not clear whether the COVID-19 crisis that started in 2019 had any uniform effect on pro-environmental behavior and decision making. In two preregistered panel studies conducted before and during the first COVID wave (n1 = 206, n2 = 164) and before and during the second COVID wave (n3 = n4 = 260), we found that the crisis had had no uniform effect on pro-environmental behaviors, environmental attitude, nor on the behavioral costs of general pro-environmental behavior. Analysis of one specific pro-environmental behavior, the choice of environmentally friendly delivery of products, revealed that the general preference for green delivery services and heightened preference for green delivery services among people with higher attitude levels remained unchanged by the COVID-19 crisis. Thus, if the COVID-19 crisis has had any effects on pro-environmental behaviors, these effects are probably fragmented, specific to certain population segments, and not visible in the short-term perspective.

The COVID-19 crisis that started at the end of 2019 has had a dramatic effect on the lives of people around the globe (IEA, 2020). Some hopeful voices have suggested that the crisis might be a turning point for humanity and would lead to people adopting more sustainable lifestyles (e.g., Kanda & Kivimaa, 2020; Muhammad et al., 2020; Sarkis et al., 2020), whereas others have warned that the crisis may shift attention away from environmental problems and weaken pro-environmental motivation and behavior (e.g., Rosenbloom & Markard, 2020). However, evidence concerning shifts in environmental motivation and behavior remains fragmented. While several studies reported strengthening of environmental motivation (e.g., O'Connor & Assaker, 2021; Schiller et al., 2021), other studies found no effect of the crisis on environmental motivation (e.g., Lucarelli et al., 2020; Rousseau & Deschacht, 2020; Čadová, 2020). Correspondingly, some studies reported increased (intention to) pro-environmental behaviors (e.g., recycling and waste handling, sustainable food consumption, and sustainable travel behavior; Kim et al., 2021; O'Connor & Assaker, 2021; Tchetchik et al., 2021), whereas other studies reported increased environmentally harmful behavior (e.g., an increase in the use of plastic products; Prata et al., 2020).

In this study, we investigate three of the possible avenues through which the crisis might have affected pro-environmental behavior and green decision making, namely: (i) by changing the level of environmental attitude (motivation) of individuals; (ii) by affecting situational constraints (herein termed behavioral costs) of pro-environmental behaviors; and (iii) by moderating the effect of environmental attitude on decision making. Given that literature on effects of the COVID-19 crisis on pro-environmental behavior and decision making is limited, our study provides an important insight into the dynamics of pro-environmental behavior and its potential drivers during the COVID-19 crisis. As such, this study contributes to the growing literature that helps us to understand, predict, and perhaps avert potential negative impacts on environmental motivation and behavior resulting from the COVID-19 crisis and similar crises that may appear in the future.

The COVID-19 crisis and preventive measures

The COVID-19 pandemic was declared by the World Health Organization on March 11, 2020 (“COVID-19 pandemic, 2022). By this time, many governments around the world had introduced measures preventing the spread of the disease (for details of these measures across countries, see, e.g., Halle et al., 2020). The government of the Czech Republic, where we collected our data, took a relatively proactive approach early on (for details, see “Průběh Pandemie Covidu-19 v Česku,” 2022), starting with a ban on direct flights from China (February 7), Italy and South Korea (March 2), simplification of the procedure for imposing a quarantine on individuals by general practitioners (March 8), restrictions on visits to medical facilities (March 9), a ban on public gatherings of more than 100 people (March 10), the closing of schools (primary, secondary and tertiary education, March 11), and declaration of a state of emergency (March 12). Thus, starting March 13, Czech citizens could not leave the country and foreigners could not enter. All shops (except for grocery stores, hygienic supply stores, pharmacies, and fuel stations) and restaurants (except for canteens at workplaces) were closed on March 14. Starting from March 15, the free movement of people was severely curtailed by allowing only journeys to and from work, family-care visits, and journeys to medical facilities. Wearing face masks in indoor public spaces and even outdoors became obligatory on March 18. Further restrictions were introduced between March 24–30 (restrictions on accommodation services, new hygienic measures for grocery stores etc.).

The easing of the restrictions in the Czech Republic took place between April 20 and May 25 according to a four-stage plan (for a concise summary, see, e.g., Government of the Czech Republic, 2020b), starting with the opening of outdoor markets, craft workshops, and outdoor sporting facilities (April 20), through the opening of small supermarkets, sporting indoor facilities, and libraries (April 27), the opening of all shopping malls, hair salons and barber's shops, museum and galleries, and small cultural facilities (May 11), until most of the remaining services (pubs and restaurants, accommodation facilities, zoos, and botanical gardens) were opened up on May, 25 but schools were not fully open until September (Government of the Czech Republic, 2020b).

The second COVID-19 wave started brewing in the Czech Republic during the summer of 2020 and gained momentum in the first weeks of September with the opening of schools and full return of the labor force. Restrictive measures were not introduced until October 14 when all schools and restaurants were closed and other restrictions were imposed (e.g., a ban on alcohol consumption in public space and a ban on gatherings of more than 6 people); additional measures were introduced on October 28 (night curfews), and November 18 (e.g., restrictions on the number of customers in stores; “Průběh Pandemie Covidu-19 v Česku,” 2022). Even though the situation continually worsened, and the Czech Republic had the second highest number of COVID-related deaths per capita worldwide already at the end of October (Cameron, 2020), complete closure of all facilities (except for groceries, drugstores, and pharmacies) was introduced only on December 27.

Behavioral impacts of the crisis

The COVID-19 crisis had a dramatic effect on the everyday behavior of people worldwide, also including behaviors with important environmental impacts. For instance, energy demand in the industrial and commercial sectors dropped by 20%, whereas energy consumption in the residential sector increased as did the use of ICT technologies (IEA, 2020). Likewise, we could observe a drop in the use of public transportation and many carbon-intensive transport modes (e.g., air travel, automobiles), a drop in shared micro-mobility, and an increase in active low-carbon transportation modes (e.g., walking, cycling; see IEA, 2020). Overall, these changes resulted in a 4% drop in energy consumption and an 8% drop in CO2 emissions worldwide (IEA, 2020). Other studies reported increased intention to recycling and waste handling (Tchetchik et al., 2021), sustainable dining out (Kim et al., 2021), and sustainable travel behavior (O'Connor & Assaker, 2021). Likewise, the average frequencies of climate mitigation actions either increased slightly (e.g., using active transportation modes) or remained the same (e.g., avoiding products with excessive packaging; IPSOS, 2020). However, many studies also reported an increase in environmentally harmful behavior such as increased waste disposal (Cheval et al., 2020) and increased use of disposable products (Prata et al., 2020).

COVID-induced behavioral change that took place in the Czech Republic during the first COVID-19 wave (spring 2020) can be illustrated by the 23% drop in overall household expenditures, including expenditures on clothing (−82%), household equipment (−66%), restaurants (−61%), gasoline and diesel (−32%), electronics (−21%); the only two categories of expenditures that increased were groceries (+27%) and medicines (+34%; for details, see Navrátil, 2020). In the same period, the proportion of online transactions more than doubled for many categories (restaurants, clothing, electronics, household equipment) and more than tripled for groceries and drugs (for details, see Navrátil, 2020). At the macroeconomic level, these trends resulted in a 10% decline in demand for services and a 9.3% decline in retail sales in March 2020 (Czech Statistical Office, 2020).

According to information from telematics sensors in the capital city of Prague, the total car traffic decreased by one third after the state of emergency was declared on March 12 (Ludvík, 2020). One of the contributing factors could have been the temporary removal of parking charges in Czech cities during the state of emergency (Government of the Czech Republic, 2020c). In addition, operation of public transportation was partially restricted shortly after the state of emergency was declared (e.g., by introducing summer schedules with longer intervals and a reduction in several bus and tram lines; Prague Transportation Service, n.d.), and the use of the public transportation in Prague dropped by 80% around the same time (Ludvík, 2020).

There is also some evidence, albeit not systematically documented, that the crisis increased the use of packaging materials. For instance, during the crisis, grocery stores introduced additional packaging for pastry products that would otherwise have been sold without packaging (Beránková & Švejdová, 2020). It is also very likely that increased online shopping during the crisis (Navrátil, 2020) accelerated the use of additional packaging.

Potential drivers of environmental behavioral change

Given that there is evidence that a number of pro-environmental behaviors changed due to the COVID crisis, the question arises whether these changes were due to the new situational constraints (and opportunities) that arose due to COVID-related measures, or whether these were driven by a change in the environmental motivation of people, or both. The distinction between situational and motivational aspects of behavior is common to many theories of environmental behavior, but we will specifically refer to the model of environmental behavior known as the Campbell paradigm (Kaiser et al., 2010; Kaiser & Wilson, 2019). In this model, environmental behavior is a function of behavioral costs (situational constraints and leverages) and environmental attitude (the motivational part is the property of persons).

The term behavioral costs, originally coined in the field of economic psychology (e.g., Robben & Verhallen, 1994), has been recently adopted also in environmental psychology (Kaiser et al., 2010) to denote the costs—including monetary and other costs (time, effort, convenience, social disapproval etc.)—associated with pro-environmental behaviors. Such behavioral costs are typically imposed externally, and they make pro-environmental behaviors less likely (Kaiser & Biel, 2000; Kaiser & Keller, 2001). Under the Campbell paradigm, behavioral costs are measured as implicit behavioral costs and are inferred from observed behavioral manifestations (for theoretical justification, see, e.g., Kaiser & Wilson, 2019; for technical details of the underlying measurement model, see, e.g., Andrich & Marais, 2019). However, other behavioral theories use perceived difficulty of behavior to capture situational constraints of behavior (e.g., Fishbein & Ajzen, 2010). There is some evidence that objective situational constrains are captured both in implicit behavioral costs (e.g., Kaiser & Lange, 2021) as well as in perceived behavioral costs (e.g., Fishbein & Ajzen, 2010).

We use the term environmental attitude interchangeably with the term of pro-environmental motivation in line with some other studies using the Campbellian framework (e.g., Kaiser et al., 2017). Environmental attitude or motivation encapsulates tendency of each person to engage in pro-environmental behavior and refrain from environmentally harmful behavior (Kaiser et al., 2010; Kaiser & Wilson, 2019).

Behavioral costs change due to the COVID-19 crisis

There is anecdotal evidence that the behavioral costs of some pro-environmental behaviors have changed due to the COVID-19 crisis and its related restrictions. For instance, people substituted public transportation with personal car transportation (Ludvík, 2020) either because of the risk of contracting COVID-19 (a specific type of behavioral cost) or because of the restrictions of public transport (longer waiting intervals as another type of behavioral costs; (Prague Transportation Service, n.d.), or because of the temporal removal of parking charges during the state of emergency (which increase behavioral cost of public transportation relative to personal car use; Government of the Czech Republic, 2020c). Likewise, increased use of additional packaging in grocery stores (Beránková & Švejdová, 2020) and an increase in online shopping (Navrátil, 2020) probably made it more difficult for consumers to avoid excessive packaging of products. On the other hand, some environmentally harmful behaviors became more difficult during the crisis. For instance, long-distance travel to foreign countries became restricted or impossible (Government of the Czech Republic, 2020a; “Průběh Pandemie Covidu-19 v Česku,” 2022), as did flying in general (IEA, 2020).

Change of environmental motivation due to the COVID-19 crisis

The COVID-19 crisis might also have changed people's environmental attitude (their motivation to engage in environmental protection). Previous studies have demonstrated that people have a limited ability to care about different problems, such as environmental protection (i.e., the limited pool of worry hypothesis; see, Linville & Fischer, 1991; Weber, 1997). The COVID-19 crisis might have changed people's goal structure by accentuating the existential threat of the disease and downgrading the perceived urgency of environmental problems (Rousseau & Deschacht, 2020). Also, there is some evidence that worries about the spread of diseases affect social values and attitudes (Thornhill & Fincher, 2014), including even apparently unrelated political preferences (Tybur et al., 2016) and interpersonal relationships (Park et al., 2012). It is plausible that by shaping values (e.g., by diminishing openness to change, see Thornhill & Fincher, 2014), worries about the spread of COVID-19 could have also affected environmental motivation.

Direct evidence of COVID-19 affecting people's environmental attitude is scarce. Only a few studies have found that the crisis has strengthened environmental attitudinal constructs (e.g., O'Connor & Assaker, 2021; Schiller et al., 2021). A study by Rousseau and Deschacht (2020) revealed a positive shift in awareness of nature-related topics but no change in awareness of environmental issues. Likewise, comparison of a large nationwide quasi-panel survey conducted in the Czech Republic revealed that the rating of global environmental problems in June 2020 was not statistically different from the rating from 2018, except that concern about genetically modified produce has risen slightly (Čadová, 2020).

Moderation of environmental motivation effect due to the COVID-19 crisis

Another possibility how the COVID-19 crisis might have impacted pro-environmental behaviors is by increasing some situational constraints that attenuated the effect of environmental attitude on behavior; this moderation pattern is known as the low-cost hypothesis (Diekmann & Preisendörfer, 2003). Such a moderation pattern would result in environmental attitude having lower weight in people's decision making.

Even though the low-cost hypothesis has been supported by some empirical studies (e.g., Farjam et al., 2019) it remains controversial as other studies failed to corroborate the moderating effect of behavioral costs on the relationship between general environmental attitude and specific pro-environmental behaviors (e.g., Byrka et al., 2017; Kaiser & Schultz, 2009; Taube et al., 2018). In any case, it is still plausible that the changing structure of behavioral costs, such as restricted availability of green products and services during the COVID-19 crisis made environmental motivation a less important factor of decision-making compared to the pre-crisis situation.

Research objectives

The objective of this paper was to investigate whether the COVID-19 crisis has affected self-reported general pro-environmental behavior and decision making related to the choice of environmentally friendly delivery service, as an example of specific pro-environmental behavior. More specifically, we expected to find that the COVID-19 crisis had increased the behavioral costs of general pro-environmental behavior and specific pro-environmental behavior (choice of environmentally friendly product delivery), which should be manifested in an increase in average implicit behavioral costs of the general pro-environmental behavior and perceived costs of green delivery, respectively. We also expected to find that the COVID-19 crisis had decreased the average environmental attitude of people as revealed in the large number of self-reports of pro-environmental behaviors. Finally, we also expected to find that the crisis had attenuated the effect of environmental attitude on decision making regarding specific pro-environmental behavior (intention to choose environmentally friendly delivery of products).

Study 1

In Study 1, we aimed to investigate the effect of the first COVID-19 wave on general pro-environmental behavior and decision making regarding specific pro-environmental behavior using a panel data from a survey conducted before the COVID-19 wave, and a follow up survey conducted during the COVID-19 wave on participants of the initial survey.

Method

Materials, analytical scripts, data, and registrations of all studies can be found here: https://osf.io/tn28x/.

Participants

A convenience sample of 206 Czech adults recruited from a participant pool of an experimental laboratory (with more than 10,000 participants) was invited to the laboratory and completed an electronic questionnaire in Czech language using desktop computers (Sample 1) between February 19 and March 6, 2020. The sample size was determined by the available resources. The sample clearly differed from the Czech general population in terms of demographics: 70.4% of females, M age = 25.6, SD age = 9.6; 7.8%, 52.4%, and 39.8%, respectively had primary, secondary and tertiary education.

All participants who took part in Sample 1 were invited to participate in Sample 2 of the study that took place more than one month later during the COVID-19 crisis. One hundred and sixty-four participants took part in Sample 2 of the study (80.0% of those who took part in Sample 1) and filled in the electronic questionnaire online between April 10 and May 11, 2020. The demographic characteristics of participants taking part in Sample 2 remained similar as in Sample 1: M age = 26.0, SD age = 9.1; 72.3% were females; 6.7%, 52.4% and 40.9% of participants respectively had primary, secondary and tertiary education.

Materials

General Pro-environmental Behavior. The general pro-environmental behavior of participants was assessed with the GEB scale (e.g., Byrka et al., 2017; Kaiser, 2020), respectively its Czech version (e.g., Urban et al., 2019, 2021); this scale consists of 50 self-reports of pro-environmental behavior. Example of an item: “I use a clothes dryer.” Participants indicated whether they performed given behaviors (18 items, response options yes, no, I don't know/not applicable) or how frequently they perform each behavior (32 items, response options never, seldom, sometimes, often, very often, I don't know/not applicable).

Prior to analysis, we recorded the response option “I don't know/not applicable” as a missing value and we dichotomized polytomous items by merging the three lowest answer categories (never, seldom, sometimes) and by merging the highest two categories (often, very often); such dichotomization (not to be confused with the controversial median-split technique) has been found to reduce measurement bias in polytomous GEB items (e.g., Kaiser & Wilson, 2004). Also, dichotomization helps to simplify statistical models used to analyze pro-environmental behavior.

The scale had sufficient person separation reliability in both samples, rel  = .71, rel 2 = .72. After being linked through invariant items, the attitude scores were similar for the two samples, M 1 = 0.00, SD 1 = 0.74, M 2 = 0.01, SD 2 = 0.75.

Implicit Behavioral Costs of Pro-environmental Behaviors and Environmental Motivation. Implicit behavioral costs of pro-environmental behaviors and participants’ levels of environmental attitude were derived from self-reports of pro-environmental behaviors using the Rasch measurement model (for details of the model, see, e.g., Bond & Fox, 2012). This approach, grounded theoretically in the attitude theory of the Campbell paradigm (for a review of the theory, see, e.g., Kaiser et al., 2010; Kaiser & Wilson, 2019), seeks to separate statistically latent propensities of individuals to engage in pro-environmental behavior (attitude) from latent properties of pro-environmental behaviors (their behavioral costs). Such an approach has turned out to be productive both for the measurement of behavioral costs of pro-environmental behaviors (e.g., Kaiser & Keller, 2001; Kaiser & Wilson, 2000) and for attitude measurement (e.g., Henn et al., 2019; Kaiser et al., 2018).

To be able to compare the behavioral costs of pro-environmental behaviors and average attitude levels between Sample 1 and Sample 2, we capitalized on the fact that estimates from the Rasch model, even if they come from different datasets, can be linked by common items (for technical details, see, e.g., Andrich & Marais, 2019; Bond & Fox, 2012). Specifically, we a priori determined that behavioral costs of nine behaviors included in the GEB scale were unlikely to be affected by the COVID-19 crisis (see the preregistration and Appendix for details). A priori determined invariant items featured pro-environmental behaviors that could not change due to COVID-19 restrictions (such as how people usually behaved in winter months) or where the change was unlikely between Sample 1 and Sample 2 due to the very short time between them (such as ownership of an energy-efficient car or buying renewable energy from an energy supplier). For the remaining 41 items, we assumed that they might have been affected by the COVID-19 crisis and treated them as non-invariant. As we mentioned earlier, there is already some evidence that many pro-environmental behaviors featuring shopping, commuting, and various types of consumption were likely affected by the COVID-19 crisis.

Perceived Change of Behavioral Costs of Online Shopping. The perceived change of behavioral costs of online shopping between, before, and during the COVID-19 crisis was measured by four items that asked participants to compare the shopping experience in terms of price, delivery time, and availability of products (specific behavioral costs) and overall (generalized behavioral cost). Participants indicated perceived change in each specific aspect for shopping experience using five-point Likert-like scales with labeled points specific to each aspect (for instance, price of delivery of products during the crisis, as compared to before the crisis, was assessed on a scale with the following labels: 1 = much smaller, 2 = rather smaller, 3 = same, 4 = rather higher, 5 = much higher). Two of the scales (availability and overall difficulty) were reverse-coded before analysis so that higher scores indicated higher behavioral costs. An overall mean score was computed from the four items. Values smaller than three can be interpreted as indicating that behavioral costs were perceived as being lower during the COVID-19 crisis, whereas mean scores above three indicated an increase in perceived behavioral costs.

Intention to Choose Environmentally Friendly Product Delivery. The intention to choose environmentally friendly product delivery was assessed with a choice experiment (for an introduction, see, e.g., Holmes et al., 2017). A choice experiment is a method used in economics and other social sciences to study preferences based on the repeated choices people make between two or more hypothetical options. Each option in a hypothetical option set is characterized by a number of attributes. Statistical models are then used to reveal implicit weights that people attach to each attribute (such as CO2 emissions attributable to delivery, delivery price, and delivery time) in their decision making. Choice experiments allow for a simultaneous study of multiple factors of decision making under conditions similar to everyday situations (such as when people choose a delivery option when shopping online).

In our choice experiment (for details, see preregistrations), participants went through 12 trials of the choice experiment. In each trial, participants saw a pair of delivery options and chose a delivery option that they preferred for a product bought in a Czech Internet shop (12 products typically bought on the Internet were embedded in trials in random order). Each delivery option was characterized by the following attributes (and their levels): transportation mode (a cargo bike, an electric car, or a conventional lorry), expected delivery time (within 12 h, 24 h, 48 h, or 3 days), average CO2 emissions associated with the delivery (CO2 emissions were nested in delivery modes; 0 g for a cargo bike, 20 g, 40 g or 60 g for the electric car, and 150 g, 200 g, or 250 g for the conventional lorry), and price of delivery (CZK 100 or USD 4.35, CZK 140 or USD 6.09, CZK 170 or USD 7.39, CZK 200 or USD 8.70, and CZK 230 or USD 10.00). Delivery modes used in the choice experiment were delivery modes used in large cities in the Czech Republic, levels of CO2 emissions reflected estimated CO2 emissions for these delivery modes (Edwards et al., 2010), and price levels reflected the usual price range for similar deliveries in the Czech Republic.

Given the combination of attributes and their levels and the fact that each option could appear on the right or the left side, there were 17,280 possible choice sets (excluding choice sets featuring the same two delivery modes). In each sample, each participant received 12 randomly chosen choice sets, each featuring two delivery options, and indicated their preferred delivery option. Participants were instructed to make each choice as they would in reality and disregard any choices they had made previously. In addition, participants in Sample 2 were specifically asked to make their choice as they would “at the present moment”.

Design

This study had a panel design but the datasets from Sample 1 and Sample 2 were not matched on participant level because we originally did not plan to continue to study participants from Sample 1 and we did not include participant identifiers.

Procedure

Participants recruited from the laboratory's participant pool were invited in groups of a maximum of 15 participants to the laboratory to participate in a study ostensibly focusing on lifestyles (Sample 1). Participants were seated in cubicles and initiated the study. After providing their informed consent, participants completed 12 rounds of the choice experiment and answered 50 items of the GEB scale. Next, participants answered 40 items of a specific measure of attitude to sustainable logistics (not analyzed in this study) before proceeding to demographic questions, a question on the type of living area, and two self-reported ratings of data quality. After completing the study, participants collected a reward (CZK 150, equivalent of USD 6.52) for participation in the study from the envelope on the table and left the laboratory.

About seven weeks later, all participants who took part in Sample 1 were invited to a follow-up study ostensibly focusing on potential lifestyle changes that might have taken place “over the last seven weeks” (Sample 2). After accessing the study online and providing their informed consent, participants completed 12 rounds of the choice experiment. Next, participants answered 50 items of the GEB scale and four items capturing the perceived change of behavioral costs of online shopping. Finally, participants proceeded to demographic questions and a question on the type of living area. After completing the study, participants were redirected to the laboratory's web-page to receive their reward for participation in the study (CZK 50, equivalent of USD 2.17) via banking transfer.

Analyses

We originally planned to analyze the data from the choice experiment in Sample 1 using a frequentist statistical framework (see preregistration for Sample 1 for details). However, mixed logit models estimated on our choice data resulted in a singular fit and unreliable parameter estimates. For this reason, we switched to a Bayesian statistical framework (for an overview of the Bayesian framework, see, e.g., Gelman et al., 2014; Kruschke, 2015) and conducted these and other analyses in the Bayesian framework in Stan (Goodrich et al., 2020; Stan Development Team, 2022) using non-informative priors (see also the preregistration for Sample 2 for details). We report mean-based point estimates of parameters. Uncertainties of parameter estimates are expressed as Bayesian 90% credible intervals (even though conceptually different from confidence intervals, CIs, credible intervals, CrIs, are used in Bayesian statistics to quantify the uncertainty associated with parameter estimates in a somewhat similar fashion as confidence intervals are used in frequentist statistics; Morey et al., 2016). Note that 90% credible intervals are preferred in the Bayesian framework because they are more stable than 95% or 99% credible intervals (Gabry & Goodrich, 2018). If a certain value falls below (or above) the 90% credible interval, one can conclude that there is a 95% or higher probability that the parameter is larger (or smaller) than that value. We also report the posterior probability that the parameter is higher than zero; this statistic can be interpreted as the probability that the parameter value is higher (for very large probabilities) or lower (for very small probabilities) than zero. For instance, P(β > 0) = .99 means that we can be rather certain that parameter β is larger than zero.

We used exploratory Rasch models (e.g., de Boeck & Wilson, 2011) to analyze the change in people's attitude levels and implicit behavioral costs of pro-environmental behaviors based on behavioral self-reports from the GEB scale. More specifically, we build on the fact that exploratory Rasch models can be expressed as mixed regression models (e.g., de Boeck et al., 2011; de Boeck & Wilson, 2011). Using a priori designated invariant items to link Rasch-based measurements from Sample 1 and Sample 2, we tested for uniform change in the difficulty of the remaining (non-invariant) GEB items and for the change in the average attitude level of the sample between Sample 1 and Sample 2.

We used a mixed logit model to analyze the data from the choice experiment. We benefited from the Bayesian framework by being able to compare directly estimated effects of independent variables between Sample 1 and Sample 2; such comparison is done by considering the posterior distribution of the difference between the posterior values from the two models. Note that such an approach is advantageous from a technical point of view because it does not introduce additional collinearity in the model that would have otherwise occurred when the interaction of time-indexing indicator and predictor variables are included in the model.

Results

General pro-environmental behavior

Probability of engagement varied substantially across pro-environmental behaviors in both samples (see Fig. 1 ; see Appendix for details). Visual comparison of engagement probability from Sample 1 and Sample 2 (see Fig. 1; see Appendix for details) reveals no uniform pattern in the change of engagement probability, except perhaps for a methodological artifact due to regression to mean whereby less frequent behaviors became more likely during the lockdown and vice versa.

Fig. 1

Average observed probabilities of engagement in pro-environmental behaviors before and during the COVID-19 vaves. Note. Panel A shows average observed probabilities of engagement in pro-environmental behaviors before and during the first COVID-19 wave (Samples 1 and 2), Panel B shows average probabilities of engagement in pro-environmental behaviors during the second COVID-19 wave (Samples 3 and 4). The X-axis represents the average observed frequency of engagement in pro-environmental behavior before the COVID-19 waves; the Y-axis represents the average frequency of engagement in pro-environmental behavior during the COVID-19 waves. The diagonal line represents equal engagement before and during the COVID-19 waves. Behaviors appearing above the diagonal became, on average, more likely during the COVID-19 wave, whereas those below the diagonal became less likely.

Results from the mixed logistic model with the responses to the GEB items as the dependent variable, and random effect for participants, fixed effects for items, and a dummy-coded indicator of time as independent variables, suggested there was no credible general change in pro-environmental behavior due to COVID-19 restrictions, β = 0.003, 90% CrI [−0.061, 0.067], OR = 1.003, P(β > 0) = .537.

Environmental motivation and implicit behavioral costs of pro-environmental behaviors

An exploratory Rasch model estimated as a mixed logistic regression model (de Boeck et al., 2011; de Boeck & Wilson, 2011) with a priori determined non-invariant items to link measurements from before and during the COVID-19 crisis revealed that average attitude level of participants remained essentially unaffected by the crisis, β = 0.043, 90% CrI [−0.158, 0.247], OR = 1.044, P(β > 0) = .629. Also unaffected was the average implicit behavioral cost of pro-environmental behaviors, β = −0.047, 90% CrI [−0.226, 0.135], OR = 0.954, P(β > 0) = .340.

Online purchase of products

Perceived Change in Behavioral Costs of Online Shopping. Perceived behavioral costs of delivery of products bought on the Internet have not increased in terms of price, M = 3.09, SD = 0.85, 95% CI [2.96, 3.22], t(163) = 1.38, p = .17, d = 0.11, but they have increased in terms of delivery time, M = 3.49, SD = 0.89, 95% CI [3.35, 3.62], t(163) = 7.02, p < .001, d = 0.55, availability of products, M = 3.32, SD = 0.89, 95% CI [3.19, 3.46], t(163) = 4.67, p < .001, d = 0.36, and overall difficulty, M = 3.27, SD = 0.85, 95% CI [3.14, 3.41], t(163) = 4.12, p < .001, d = 0.32; the overall mean of perceived behavioral cost also suggests an increase during the COVID-19 lockdown, M = 3.29, SD = 0.37, 95% CI [3.23, 3.35], t(163) = 10.29, p < .001, d = 0.80.

Choice of Delivery. Results from the mixed logit model (see Fig. 2 for summary of effects, see also Appendix for details) estimated on the data from the choice experiment in Bayesian framework revealed that time and price had both expected negative effects on the choice of the delivery both before and during the COVID-19 crisis. In both waves of data collection, we found highly credible evidence that environmentally friendlier delivery options were preferred and that people with higher environmental attitude levels preferred green delivery options comparatively more.

Fig. 2

Estimated effects from the choice experiment model and their 90% Bayesian credible intervals (mixed logit models). Note. Panels A, B, C, D display estimates based on Samples 1, 2, 3, 4, respectively. Dependent variable is the choice of a delivery alternative. Time is the (expected) delivery time; price is the price of the delivery; on left is a dummy indicator of the selected option being displayed on the left side of the choice experiment card; green is a dummy indicator of a delivery option from each set that is environmentally friendlier; round × on left is an interactive term of the trial and the delivery option being displayed on the left side of the choice set; round × green is an interactive term of the trial and the green delivery option; green × attitude is an interactive term of the green delivery option and attitude level of a person. Dots and error bars denote medians and 90% credible intervals of posterior distributions, respectively. All variables were standardized prior to analysis. Broken vertical lines denote no effect (beta = 0). Attitude scores were estimated independently in each sample.

Using the posterior distribution of parameter estimates from models estimated on Sample 1 and Sample 2 with attitude estimates linked by invariant items, we estimated the differences between effects estimated in the two models. This analysis revealed that the preference for the green delivery options remained virtually unchanged between Sample 1 and Sample 2, M(β1 - β2) = 0.067, 90% CrI [−0.103, 0.234], P((β1 - β2) > 0) = .751. There seemed to be a small drop in the relative preference for green products among people with higher levels of environmental attitude, M(β1 - β2) = 0.134, 90% CrI [−0.020, 0.290]; this change was relatively credible, P((β1 - β2) > 0) = .922, even if did not reach the preregistered inference criteria. In addition, there was also a relatively credible increase in the effect of delivery time on delivery preference, M(β1 - β2) = 0.082, 90% CrI [−0.010, 0.179], P((β1 - β2) > 0) = .928, which also did not reach the preregistered inference criteria. On the other hand, the effect of delivery price remained virtually unchanged between the two waves, P((β1 - β2) = −.048, 90% CrI [−0.158, 0.064], P((β1 - β2) > 0) = .238. Note that these results remain essentially the same when we estimated attitudes scores independently using Sample 1 and Sample 2 (see Fig. 2 for details).

Discussion

In Study 1 we found essentially no effect of the first COVID-19 wave on general pro-environmental behavior, environmental attitude, and implicit behavioral costs of pro-environmental behaviors. Even though the crisis increased perceived behavioral costs of online shopping, it had no effect on preference for green delivery service. We found a relatively week evidence that the crisis attenuated weight of environmental attitude in decision making. However, this study was limited in that we were not able to link data from Sample 1 and Sample 2 on the individual level.

Study 2

In Study 2 we aimed to corroborate results of Study 1 using measurements conducted before and during the second COVID-19 wave that were now linked on an individual basis.

Sample size determination

The sample size was determined primarily based on the resources available. We wanted to achieve a sample size of at least N = 200 (participants taking part in both study waves) similar to Sample 1. Hence, we aimed to sample 300 participants for Sample 3 with the hope of having at least 200 participants in Sample 4. We conducted two power analyses (see preregistration for details) to check that such a sample size (i.e., N = 200) would be sufficient to capture effect sizes of interest. These power analyses revealed that given the sample size, our study would have a power of .850 to detect a drop in the preference for green products among people with high attitude levels (Δβ = 0.14, logistic coefficients, 90% CrI) and a power of .924 to detect a uniform change in attitude levels of 0.3 logits (a power of .731 do detect a change of 0.2 logits).

A convenience sample of 380 Czech adults recruited from a participant pool accessed the study online (participants taking part in a previous study were not invited) and 310 participants completed the first wave of the study between September 2 and October 3, 2020 (henceforth referred as Sample 3). All participants who completed the first wave of this study were invited, seven weeks later, to participate in the second wave of Study 2; 260 participants (dropout 14.71%) completed the second wave of the study (henceforth referred to as Sample 4) between November 20 and December 30, 2020. The final sample (N = 260 participants who participated in both waves) was variable but not representative, 65.8% were females, M age = 24.8, SD age = 8.2; 5.0%, 55.8%, and 39.2%, respectively had primary, secondary, and tertiary education.

We used the same measures of general pro-environmental behavior, implicit behavioral costs of pro-environmental behavior, and environmental attitude as in Study1. GEB scale had a sufficient person separation reliability in both waves, rel 3 = .75, rel 4 = .73.

Perceived Change of Behavioral Costs of Online Shopping. We used the same measure of perceived change of behavioral costs of online shopping as in Study 1 except that in the current study, participants indicated perceived change between September and the current situation (i.e., when Sample 4 was collected, November 20-December 30). For linking of attitude measures, we a priori designated a subset of six items from Study 1 (we excluded three items that referred to household heating behavior because of the large outdoor temperature differences between the two time points when Samples 3 and 4 were taken; see Appendix for details of items).

Intention to Choose Environmentally Friendly Product Delivery. We used a similar choice experiment as in Study 1 to measure the intention to choose environmentally friendly delivery of products. The only difference was that we used slightly different attribute levels in the current study (the same in Samples 3 and 4): (i) we added a courier taxi; (ii) we increased the range of delivery time values and presented them as discrete values rather than intervals (1 h, 12 h, 48 h, 3 days, 7 days); (iii) we increased the range of CO2 emissions that were nested in transportation modes (0 g for a cargo bike; 0 g, 20 g, 40 g or 60 g for the electric car; 200 g, 350 g, or 500 g for the conventional van; and 800 g, 1000 g, 1200 g for the taxi courier); (iv) we also increased the range of delivery costs (CZK 80 or USD 3.49, CZK 100 or USD 4.35, CZK 150 or USD 6.52, CZK 200 or USD 8.70, CZK 250 or USD 10.87, and CZK 290 or USD 12.61). We instructed participants to make their choice as they would “at the present moment” and disregard any choices they had made during the preceding trials of the choice experiment.

This study had a panel design: Samples 3 and 4 were matched using individual identifiers.

Participants from a laboratory participant pool were invited to a study ostensibly focusing on lifestyles, decision making, and everyday activities. After accessing the study online and providing informed consent, participants completed 12 rounds of the choice experiment. Next, participants answered 50 items of the GEB scale and four items capturing the perceived change of behavioral costs of online shopping. Participants then evaluated six perceived attributes (reliability, speed, environmental friendliness, contribution to traffic flow, quality, and cost) of four delivery modes (bicycle, conventional van, electric car, and taxi courier) each on semantic scales (not analyzed in this study). Finally, participants proceeded to demographic questions and a question on the type of living area. After completing the study, participants were redirected to the laboratory's web page to receive their reward for participation (CZK 50, equivalent of USD 2.17) via banking transfer. Seven weeks later, all participants who completed the first wave of the study (Sample 3) were invited to a follow-up study (Sample 4) ostensibly focusing on potential lifestyle changes that might have taken place over the last seven weeks. The structure of the study was similar to Sample 3 study. After accessing the study online and providing their informed consent, participants completed 12 rounds of the choice experiment. Next, participants answered 50 items of the GEB scale and four items capturing the perceived change of behavioral costs of online shopping. Next, participants evaluated six perceived attributes (reliability, speed, environmental friendliness, contribution to traffic flow, quality, and cost) of four delivery modes (bicycle, conventional van, electric car, and taxi courier) each on semantic scales (not analyzed in this study). After completing the study, participants were redirected to the laboratory's web page to receive their reward for participation (CZK 50, equivalent of USD 2.17) via banking transfer.

Analysis

We used the same analytical approaches as in Study 1. The only difference was that we included participant identifiers that paired observations from Sample 3 and 4. These identifiers were included as random intercepts (exploratory Rasch model) and random effects (choice model; see preregistration and analytical scripts for technical details).

Similar to Study 1, the probability of engagement varied substantially across pro-environmental behaviors in both waves of data collection (see Fig. 1; see Appendix for details). Visual comparison of engagement probability from Samples 3 and 4 (see Fig. 1; see Appendix for details) revealed no uniform pattern in the change of engagement probability. Similarly to Study 1, the mixed logistic model with the responses to the GEB items as the dependent variable revealed no credible uniform change in pro-environmental behaviors due to COVID-19 restrictions by preregistered criteria, β = 0.055, 90% CrI [−0.002, 0.112], OR = 1.056, even though a drop in general pro-environmental behavior was relatively plausible, P(β > 0) = .945.

Environmental attitude and implicit behavioral costs of pro-environmental behaviors

Similar to Study 1, the exploratory Rasch model did not reveal any uniform change attributable to COVID-19-related restrictions in people's attitude levels, β = −0.001, 90% CrI [−0.102, 0.099], OR = 0.999, P(β > 0) = .492, or in the implicit behavioral costs of pro-environmental behaviors, β = 0.029, 90% CrI [−0.068, 0.127], OR = 1.029, P(β > 0) = .675.

Perceived Change in Behavioral Costs of Online Shopping. Similar to Study 1, we saw an increase in perceived behavioral costs of online shopping concerning availability of products, M = 3.13, SD = 0.85, t(259) = 2.55, p = .011 (one-sided t-test), d = 0.16, delivery time, M = 3.42, SD = 0.82, t(259) = 8.34, p < .001, d = 0.52, and overall difficulty, M = 3.40, SD = 0.75, t(259) = 8.56, p < . 001, d = 0.53, but not in terms of perceived monetary cost, which dropped slightly, M = 2.91, SD = 0.67, t(259) = −2.12, p = .034, d = 0.13. However, the total behavioral costs were perceived as being higher during the COVID-related restrictions, M = 3.22, SD = 0.35, t(259) = 9.92, p < .001, d = 0.62.

Choice of Delivery. Results from the mixed logit model estimated on Samples 3 and 4 corroborated the negative effects of delivery price and time on the choice of delivery option, the relative preference for green delivery options, as well as the relatively higher preference for green delivery options in people with higher attitude levels (see Fig. 2 for a summary of effects, see also the Appendix for details). We did not observe any change in preference for the green delivery option or in how delivery time and environmental attitude affected people's choice of delivery options, .198 > Ps(βs > 0) > .763. We only detected two changes that occurred between Samples 3 and 4. First, people preferred options displayed on the left side slightly more in Sample 4 than in Sample 3, M(β1 - β2) = −0.076, 90% CrI [−0.135, −0.016], P((β1 - β2) > 0) = .016. Second, price weighted more in people's decision making in Sample 4 than in Sample 3, M(β1 - β2) = 0.075, 90% CrI [−0.017, 0.171]; this effect did not pass preregistered inference criteria but was relatively credible, P((β1 - β2) > 0) = .903.

Similarly to Study 1, this panel study found no change in environmental attitude levels and no change in the average implicit behavioral cost of pro-environmental behaviors due to COVID-related restrictions. We were also able to corroborate the findings of Study 1 that people generally preferred green delivery options and that people with higher levels of environmental attitude preferred them more, whereas delivery price and time had negative effects on the choice of delivery options. The weight of these factors in decision making did not change, despite the fact that people perceived most behavioral costs of online shopping as being higher during the COVID-related restrictions. We only found that delivery price weighted more in people's decision making regarding delivery of products during the COVID-related restrictions.

Longer-term change in decision making

One of the limitations of Studies 1 and 2 was that the studies assessed attitudinal and behavioral change over relatively short stretches of time (seven weeks in each of the studies). To compensate for this deficiency, we also analyzed attitudinal and behavioral change between Sample 1 and Sample 4 (time gap of 9 months).

Samples 1 and 4 included different individuals but the two samples had similar characteristics as they came from the same participant pool. Thus, we can assume that the effect of attitude on decision making in the two samples would be similar if we assessed them at the same time (i.e., homogeneity of the effect). We can then use the two samples to estimate how COVID-related restrictions affected the weight of environmental attitude in decision making regarding the choice of green delivery of products. Likewise, assuming that behavioral costs of pro-environmental behaviors would be approximately similar in the two samples if they were taken at the same time and using a set of a priori designated invariant items in the GEB scale, we can also test for a uniform effect of COVID-related restrictions on behavioral costs of pro-environmental behaviors. Note that none of these comparisons requires that attitude levels or intensity of pro-environmental behaviors would be similar in the two samples if they were taken at the same time.

We used the same analytical approach as in Study 1 to compare the role of environmental attitude in decision making and behavioral costs between two samples of participants that were not matched on an individual basis.

Implicit behavioral costs of pro-environmental behavior

The exploratory Rasch model revealed that average attitude levels were similar in Samples 1 and 4, β = 0.111, 90% CrI [−0.134, 0.367], OR = 1.118, P(β > 0) = .768. Likewise, we did not observe a uniform change in implicit behavioral costs of pro-environmental behaviors, β = −0.053, 90% CrI [−0.174, 0.065], OR = 0.948, P(β > 0) = .235.

Choice of delivery

Results from the mixed logit model corroborated the findings from Studies 1 and 2, namely the negative effect of delivery time and price on the choice of delivery option, and the relative preference for the green delivery option, as well as a higher preference for green delivery options among people with higher attitude levels (see graphical summary in Fig. 2, for details of parameters estimates, see Appendix). Importantly, a comparison of effects estimated on Sample 1 and Sample 4 revealed no difference in preference for green delivery options, M(β1 - β2) = −0.105, 90% CrI [−0.287, 0.072], P((β1 - β2) > 0) = .172, no change in how people with higher attitude levels preferred green delivery options, M(β1 - β2) = 0.043, 90% CrI [−0.104, 0.188], P((β1 - β2) > 0) = .685, and no change in the weight of price in decision making, M(β1 - β2) = 0.054, 90% CrI [−0.046, 0.156], P(β1 - β2) > 0) = .817. On the other hand, we also found that the delivery time weighted in more in people's decision making in Sample 4 than in Sample 1, M(β1 - β2) = 0.351, 90% CrI [0.257, 0.444], P((β1 - β2) > 0) > .999.

Our analysis of attitudinal and behavioral change that took place between after the outbreak of the COVID-19 crisis until the full swing of the second wave of the crisis (nine months period) corroborated the remarkable stability of general pro-environmental behavior, environmental attitude, and implicit behavioral costs of pro-environmental behaviors found in Studies 1 and 2. We also found that the general preference for green delivery services and the weight of environmental attitude in decision making remained remarkably stable, similarly to the findings of Studies 1 and 2, whereas delivery time weighted more in people's decision making during the COVID-19 crisis than before.

General discussion

The goal of this study was to examine the effect of the COVID-19 crisis on self-reported general pro-environmental behavior and decision making concerning one specific pro-environmental behavior (choice of environmentally friendly delivery service). We analyzed three possible routes through which the COVID-19 crisis might have affected pro-environmental behavior and decision making, namely by (i) changing the behavioral costs of pro-environmental behaviors; (ii) changing the environmental attitude of people (their environmental motivation); and by (iii) moderating the effect of environmental attitude on behavior and decision making.

Across two preregistered panel studies conducted before and during the first and second COVID-19 waves in the Czech Republic, we found no uniform change in engagement in general pro-environmental behavior. We also did not find a uniform change in people's environmental attitudes or in the behavioral costs of pro-environmental behaviors. Finally, in both of our studies and even across the period of nine months that passed between our first measurement (before the COVID-19 outbreak) and our last measurement (during the second COVID-19 wave), we found that the preference for green delivery options and the weight of environmental attitude in decision making remained remarkably stable.

Relative stability of pro-environmental behavior, environmental motivation, and behavioral costs

Finding no uniform change in general pro-environmental behavior, environmental attitude, and the behavioral costs of pro-environmental behaviors may seem surprising considering the tremendous impact COVID-19 had on the everyday lives of people (e.g., Czech Statistical Office, 2020; IEA, 2020; Navrátil, 2020). However, this apparent inconsistency is due to the fact that in the present study, we were searching for uniform changes in behaviors, attitude, and behavioral costs that may be hidden behind the myriads of singular behavioral changes that we could witness during the COVID-19 crisis. In other words, rather than looking at behavioral changes in singular pro-environmental behaviors separately (except for green delivery decision making) we were interested in whether we can find a common change in attitude or in behavioral costs that would be manifested in a large number of singular pro-environmental behaviors; our answer to that question is no. In that respect, our results are consistent with other studies that found no change in environmental attitude (e.g., Lucarelli et al., 2020; Rousseau & Deschacht, 2020; some other studies reported a small increase in environmental attitude, see; O'Connor & Assaker, 2021; Schiller et al., 2021).

Our finding of no uniform change in environmental attitude, general pro-environmental engagement, and the behavioral costs of pro-environmental behavior does not mean that singular pro-environmental behaviors were not affected by the COVID-19 crisis. There is ample evidence that the effect of the COVID-19 crisis on singular pro-environmental behaviors varied with a positive effect on some (e.g., recycling, sustainable dining out, sustainable travel behavior; Kim et al., 2021; O'Connor & Assaker, 2021; Tchetchik et al., 2021), no effect on others (e.g., avoiding products with excessive packaging; Čadová, 2020; IPSOS, 2020), and negative effect on still others (e.g., increased use of disposable products; Prata et al., 2020). In fact, some of these changes in singular pro-environmental behaviors can be observed also in our study (viz. the comparison of average observed engagement probabilities of singular pro-environmental behaviors before and during the COVID-19 waves). Our finding of a non-uniform effect of the COVID-19 crisis on pro-environmental behavior is somewhat similar to observations made during the 2008 economic crisis that also had a mixed effect on pro-environmental behaviors (i.e., positive in terms of energy and water savings but negative in terms of lower willingness to pay for climate mitigation; Ivlevs, 2019).

Likewise, there is some evidence that the behavioral costs of some of the pro-environmental behaviors have increased; for instance the introduction of longer waiting intervals and the reduction of tram lines in Prague during the first COVID-19 wave (Prague Transportation Service, n.d.) made travel on public transportation more difficult. Likewise, the obligatory introduction of additional packaging to pastry (Beránková & Švejdová, 2020) made individual waste reduction a more difficult task. A similar increase in perceived difficulty of online shopping was also observed in our study. On the other hand, there were some pro-environmental behaviors for which behavioral costs probably dropped as a result of COVID-related restrictions such as restrictions on air travel. Again, in our study, we did not find evidence that there was a uniform effect of the COVID-19 crisis on the behavioral costs of pro-environmental behaviors.

Choice of environmentally friendly delivery

Our study found the expected increase in perceived behavioral costs of online shopping during both COVID-19 waves; participants saw online shopping as more difficult during as compared to before a COVID-19 wave due to lower availability of products, longer delivery time, and worse shopping experience overall. Despite behavioral costs of online shopping having increased, the preference of participants for environmentally friendly delivery options remained essentially the same not only during both COVID-19 waves but also when we compared decision making prior to the first COVID-19 wave and during the second COVID-19 wave (about a nine-month gap). Finding such a stable preference for green delivery only suggests that the difficulty of online shopping did not affect how people wanted their products to be delivered. Note that this pattern does not violate the assumptions of the attitude model of the Campbell paradigm as the behavioral cost of online shopping (i.e., cost of purchasing the product online as opposed to purchasing it elsewhere) may be unrelated to the behavioral cost of opting for environmentally-friendly delivery (i.e., the behavioral cost of green delivery vs. conventional delivery).

We found some evidence in Study 1 that the difference in preferences for green delivery services that existed between people with high and low environmental attitude became smaller during the first COVID-19 wave. However, we did not replicate this finding in Study 2 or across the first and last measurement (i.e., the 9 months gap). Thus we conclude that the weight of environmental attitude in decision making remained rather stable; in other words, the increased behavioral costs of online shopping most likely did not moderate the effect of environmental attitude on this environmental decision making.

Theoretically expected negative effects of delivery time and delivery price became more important as factors of decision making over time, probably reflecting the fact that COVID-related restrictions increased delivery time and delivery prices. As a result of this, customers started paying more attention to these factors.

Limitation and avenues for future research

A limitation of Study 1 was that we could not match observations from the two measurements on an individual basis. However, given a relatively small dropout rate in Study 1 (20%), the average effects that we compare in Study 1 were estimated on a very similar sample of individuals and thus these comparisons remain, to a large degree, valid. Also, the consistency of key results with Study 2 that had participants matched on an individual basis is somewhat reassuring.

Another limitation of our studies was that they were conducted in an evolving situation that only allowed for limited planning and put some limits on how rigorous our study could be. For instance, we conducted measurements before and during the COVID-19 waves in each of the studies. When we started our study before the COVID-19 outbreak (Sample 1), some people might have been already aware of COVID-related risks and been taking some protective measures (the rapid spread of COVID-19 infection in China and beyond had already received some media attention in the second half of December 2019). Likewise, at the time of the first measurement in Study 2 (Sample 3), there was already a trend of an increasing number of COVID-19 cases in the Czech Republic, even though restrictions were not introduced until about a month later.

Another limitation is that we used non-representative convenience samples of participants in both studies. These had a high proportion of females, young people, and highly educated people. Our studies have corroborated some theoretically expected effects on decision making (i.e., the negative effects of delivery price and time on delivery choice, and higher tendency among high-attitude people to choose environmentally friendly delivery), which attests to some external validity of our studies. Combined with evidence from other studies, our work provides evidence that the COVID-19 crisis did not undermine environmental motivation and general pro-environmental behavior. However, it is still possible that the COVID-19 crisis might have triggered different behavioral reactions in some specific segments of the population and our results would not generalize to these segments.

Another potential limitation of our work is that the two panel studies had a time gap of only seven weeks due to how the COVID-19 situation unfolded and due to a lack of information about future development. However, even such relatively short periods were sufficient for a substantial change in many everyday behaviors (Czech Statistical Office, 2020; Navrátil, 2020). In the light of these changes, it is remarkable how stable general pro-environmental behavior and environmental decision making remained not only across the relatively short span of our two studies but also between the first and last measurements, which were separated by nine months.

Finally, the causal claims of COVID-19 crisis effects presented in the current study were based on observational evidence from panel studies. Such studies cannot control confounding factors that are intertwined with the presumed causal factor. As such, these causal claims should be read with some caution. However, isolation and corroboration of causal effects based on observational data requires that the causal model is well specified (e.g., Pearl & Mackenzie, 2018). Unfortunately, we are only starting to understand the behavioral dynamics resulting from the COVID-19 crisis. Given that randomized experiments may not be a feasible option in this context, future studies may use empirical data from this and similar observational studies in combination with refined causal models to better understand the causal effects of the COVID-19 crisis on pro-environmental behavior and motivation (see Pearl & Mackenzie, 2018 for discussion of such models).

Conclusions

The COVID-19 crisis had no uniform effect on general pro-environmental behavior and no effect on environmental decision making regarding specific pro-environmental behavior (choice of environmentally friendly delivery of products). The crisis affected different pro-environmental behaviors differently but did not change underlying pro-environmental motivation and its role in decision making.

Funding

This work was supported by the (grant No. TL02000437). We wish to thank Cliff McLenehan for his language support and two anonymous reviewers for their insightful comments. Correspondence concerning this article should be addressed to Jan Urban, Environment Centre, Charles University, José Martího 407/2, 160 00 - Prague 6, Czech Republic.

Author statement

JU: Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Writing – original draft. MBK: Funding acquisition, Project administration, Conceptualization, Investigation, Writing – review & editing.