Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Journal of Environmental Psychology 86 (2023) 101945

Contents lists available at ScienceDirect

Journal of Environmental Psychology


journal homepage: www.elsevier.com/locate/jep

Promoting new pro-environmental behaviors: The effect of combining


encouraging and discouraging messages☆,☆☆
Ann Kronrod a, *, Anat Tchetchik b, Amir Grinstein c, d, Lior Turgeman e, Vered Blass e
a
Department of Marketing, Entrepreneurship and Innovation, University of Massachusetts, Lowell, United States
b
Department of Geography and Environment, Bar Ilan University, Ramat-Gan, 5290002, Israel
c
Department of Marketing, Northeastern University, Boston, United States
d
Department of Marketing, Vrije Universiteit, Amsterdam, the Netherlands
e
Porter School of the Environment and Earth Sciences, Tel Aviv Universit, Israel

A R T I C L E I N F O A B S T R A C T

Handling editor:Sebastian Berger Promoting pro-environmental behavior faces multiple challenges. Promoting new pro-environmental behaviors is
even more challenging, due to additional barriers, such as perceived lack of information. Traditional pro-
Keywords: environmental communication often either encourages desired behaviors or discourages undesired behaviors.
New pro-environmental behavior We argue that separately, these two approaches are limited in their ability to elicit perceptions of informativeness
Education
and therefore they may not be effective enough in the context of new pro-environmental behaviors, because of
Environmental communication
the profound need in educating the public about these new behaviors. Addressing this challenge, we test across
Field experiment
six studies the effectiveness of a communication approach based on education psychology (specifically the
“behavior reorientation” approach), which combines the encouraging and the discouraging language in a single
integrated message. In three large field experiments and a field survey we find that, compared with communi­
cation that uses separately an encouraging or a discouraging message, a combined message that integrates both
approaches elicits higher engagement with new pro-environmental behaviors. Three follow-up online studies
demonstrate that the effect of the combined message occurs only in the context of new (rather than established)
pro-environmental behaviors, and show the mediating role of perceived informativeness, echoing the need for
education in such contexts.

1. Introduction to be new behaviors, are driving alternative energy vehicles (Tchetchik,


Zvi, Kaplan, & Blass, 2020), participating in recycling activity of elec­
A critical path towards positive environmental change is likely to tronic waste (e-waste) (Echegaray & Hansstein, 2017; Pérez-Belis,
require the adoption of new pro-environmental behaviors (e.g., Byrka & Bovea, & Simó, 2015), moving towards behaviors such as home-grown
Kaminska, 2022; Carrilo-Hermosilla, Del Río González, & Könnölä, hydroponic agriculture (Möhlman, 2015) or consuming lab-grown
2009). New pro-environmental behavior is defined as a novel behavior food (Jha, 2013; Marr, 2019; Monbiot, 2020).
aiming at significant and demonstrable progress towards the goal of Pro-environmental behavior is often difficult to attain (Dietz,
sustainable development (European Commission, 2013). The behavior Ostrom, & Stern, 2003; Echegaray & Hansstein, 2017; Morren & Grin­
may be novel to society or to the individual (Garcia & Calantone, 2002). stein, 2021; White & Simpson, 2013). Persuasion efforts thus far have
For example, entrepreneurial and business opportunities around new yielded an upsurge in the adoption of environmental values and atti­
solutions to mitigate climate change are on the rise (EIO, 2010; Maz­ tudes but are much more limited in enhancing actual behavior (John­
zanti, 2018; Ter Mors, Weenig, Ellemers, & Daamen, 2010). Examples of stone & Tan, 2015; Lane & Potter, 2007; White, Habib, & Hardisty,
environmental behaviors that are considered, at least in most countries, 2019). Further, the challenge of promoting new pro-environmental


This research was supported by the ISRAEL SCIENCE FOUNDATION (grant No. 789167/). This research was also supported by the Henry Crown Institute of
Business Research in Israel, Tel Aviv University.☆☆ The authors wish to thank Lee Grafter for her advice on statistical analysis. The authors also thank Dizengof
Center, the Kipod Center at Kfar Saba Municipality, and M.A.I - ISRAEL ELECTRONICS RECYCLING CORPORATION LTD for their collaboration on the field research.
* Corresponding author.
E-mail addresses: ann_kronrod@uml.edu (A. Kronrod), Anat.Tchetchik@biu.ac.il (A. Tchetchik), a.grinstein@neu.edu, a.grinstein@vu.nl (A. Grinstein), turgeman.
lior@gmail.com (L. Turgeman), vblass@tauex.tau.ac.il (V. Blass).

https://doi.org/10.1016/j.jenvp.2022.101945
Received 14 December 2021; Received in revised form 10 November 2022; Accepted 18 December 2022
Available online 29 December 2022
0272-4944/© 2022 Elsevier Ltd. All rights reserved.
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

behavior is especially pronounced, as it combines both barriers to 2. Conceptual development


adoption of pro-environmental behavior (e.g., higher costs or efforts, the
perception of climate crisis/change as distant and future oriented and 2.1. Persuading the public to adopt new pro-environmental behaviors
not urgent; Bockarjova & Steg, 2014; Lacroix & Gifford, 2018; Schultz &
Kaiser, 2012; Van Der Wal, Van Horen, & Grinstein, 2018), and barriers The relationship between innovation and sustainability has received
to innovation adoption (e.g., perceived lack of knowledge or experience, increasing attention in the last decades (e.g., Beise & Rennings, 2003;
resistance to change due to status-quo bias, uncertainty and risk aver­ Varadarajan, 2017). New pro-environmental behaviors involve a change
sion; Gifford, 2011; Ozaki, 2011; Sood & Kumar, 2017). Given these in behavior, perceptions and attitudes related to the environment, such
multiple challenges, the unique context of new pro-environmental be­ as the adoption of new technologies that can mitigate environmental
haviors is likely to require extra-efforts to modify behaviors. Addressing degradation (e.g., consuming cultured meat or hydroponic home agri­
these issues, the goal of this research is to identify and test the effec­ culture) or adopt a new pattern of activity or a service that carries
tiveness of a novel communication strategy that may be especially environmental benefits (e.g., Uber Pool). Recent studies tried to shed
relevant for promoting the adoption of new pro-environmental light on the processes, drivers and barriers of adopting new
behaviors. pro-environmental behaviors (e.g., Byrka & Kaminska, 2022; Echegaray
Prior research on sustainability and health communication has often & Hansstein, 2017; Tchetchik et al., 2020, Tchetchik, Kaplan, & Blass,
investigated the effectiveness of one of two key messages: Either 2021). Overall, the findings suggest that the adoption of new
encouraging desired behaviors, or discouraging harmful behaviors pro-environmental behaviors is likely to be slow (Pérez-Belis et al.,
(Keller & Lehmann, 2008; Snyder et al., 2004). This research reveals 2015), and to strongly depend on multiple aspects of the situation, such
mixed findings regarding the effectiveness of each of these communi­ as messaging, perceptions, the regulatory environment, the specific
cation strategies, especially when tested separately (e.g., Passyn & behavior at hand etc. (e.g., Byrka & Kaminska, 2022; White & Simpson,
Sujan, 2006; Stivers, 2005, 2006). Further, the traditional encouraging 2013)
or discouraging communication may not fully satisfy the additional One of the main barriers for the adoption of new pro-environmental
education efforts needed with new pro-environmental behaviors, given behaviors is poor knowledge about the harmful environmental impacts
the perceived lack of knowledge, the uncertainty and the risk associated of existing behaviors or products, as well as ignorance about the benefits
with the innovation (Hoch & Ha, 1986; McGregor, 2005; Zhang, Liang, of the new environmental behaviors and the ways to adopt and imple­
& Wang, 2016). A more integrative and informative approach is ment them (e.g., Lane & Potter, 2007; Sierzchula, Bakker, Maat, & Van
essential for the success of modifying behaviors and adopting new ones Wee, 2014; Wang, Guo, Wang, Zhang, & Wang, 2018). Echoing these
(Gifford, 2011; Landry, Gifford, Milfont, Weeks, & Arnocky, 2018; findings, an analysis of online chatter about e-waste in the country
Osbaldiston & Schott, 2012). where the current research took place (where e-waste collection is a
Addressing these challenges, the current work tests an integrative relatively new phenomenon) was conducted in collaboration with a
approach to promoting new pro-environmental behaviors, which is professional market research company (Buzzilla) in May 2018. The
based on education psychology (specifically the “behavioral reor­ study revealed that out of 866 social network conversations about
ientation” approach). The approach combines in one message the e-waste, 36% focused on asking for advice, such as information about
discouraging of undesired behavior (e.g., “Don’t draw on the walls”) and e-waste collection locations and ways to learn about e-waste recycling,
the encouraging of desired behavior (e.g., “Draw on the paper”). The and 48% focused on sharing information with others. Thus, information
educative value of this approach is in that it is perceived as more (and lack of it being a barrier) appears to be a priority topic in online
informative and more integrative because it simultaneously indicates chatter about this type of new pro-environmental behavior.
what behaviors should be stopped and opens a path to an alternative How can we overcome the barrier of perceived lack of information
behavior. We argue that this approach is more effective than using using appropriate communication strategies? Literature on pro-social
separately encouraging or discouraging messages – especially for new persuasion often classifies messages to either encouraging desired be­
pro-environmental behaviors – because the combined message makes haviors or discouraging harmful ones (Keller & Lehmann, 2008; Snyder
people feel more informed. Information is an aspect that has been shown et al., 2004; White & Simpson, 2013). However, the findings are mixed
especially important in the context of encouraging new behaviors and is with regards to the effectiveness of these two communication strategies.
a key component of education (e.g., Gifford, 2011; Osbaldiston & Schott, For instance, in their meta-analysis of health communication, Keller and
2012; White et al., 2019), because it solves the uncertainty involved in Lehmann (2008) found no effect of discouraging messages on compli­
adopting the new behaviors. ance and a negative impact for encouraging messages. In an earlier
Three large field experiments concerning two different new pro- meta-analysis (Mullen et al., 1997), no substantial difference in effect
environmental behaviors (residential hydroponic agriculture and e- sizes was found between interventions that tried to encourage a desired
waste recycling) and a field survey support the benefits of a combined behavior (e.g., seat belt use) and those that tried to discourage behavior
message over only an encouraging or only a discouraging one. Three (e.g., alcohol reduction). In an environmental context, Bockarjova and
follow-up online experiments demonstrate that the effect holds only for Steg (2014) found that respondents were more likely to adopt an electric
new, but not for established pro-environmental contexts. The online vehicle when they perceived the negative consequences caused by
studies also show that the combined message is more effective because it conventional vehicles as more severe (a discouraging message), and also
makes people feel more informed, compared with any of the single when they expected electric vehicles to decrease these consequences (an
messages. encouraging message). Conveying only a discouraging message, which
The current research offers multiple contributions to literature on presupposes that harmful action is being, or is about to be, done
pro-environmental behavior. First, new pro-environmental behavior – a (Gamliel & Herstein, 2011) may induce perceptions of a dead-end that
key social, economic, and technological global trend, requires special lacks an alternative. Conversely, conveying only encouraging messages
attention because of its growing magnitude and importance and also due does not carry this presupposition (Austin, 1968), but presupposes that
to the multiple barriers to adoption. Second, we test the effectiveness of the individual needs advice, puts the advice giver at a higher position,
a new communication approach in driving adoption of new pro- and consequently may lead them to be perceived as dominating and
environmental behaviors. Third, we identify a mediator that echoes authoritative (Feng & MacGeorge, 2010; Schwartz, Luce, & Ariely,
the importance of feeling informed in the context of new pro- 2011). Noteworthy is that this earlier work studied the encouraging and
environmental behavior adoption as well as a boundary condition (the discouraging messages in separation.
effect is evident with new pro-environmental behaviors and not with Apart from the mixed findings reported above, we argue that mes­
established ones). sages that either encourage or discourage behaviors may not be optimal

2
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

to promote new pro-environmental behavior, because of perceived lack effectiveness of the campaign, because of the educational and infor­
of knowledge, which is embedded in providing just one of the messages mational value of these combined messages (Helmig & Thaler, 2010).
given their disintegrative, limited nature. We suggest that to promote For example, multiple messages in an anti-smoking context were effec­
new pro-environmental behaviors, communicators may need to apply a tive because they enabled communicators to highlight both the benefits
more informative and integrative approach in their communication, of quitting and the costs of continuing smoking (Strahan et al., 2002).
similar to the way market education is required when introducing new Finally, two-sided messages – combining pros and cons, or positive and
products (McGregor, 2005; Zhang et al., 2016). We next discuss a negative aspects of a product – increase perceptions of message/firm
communication strategy inspired by educational psychology to credibility (Kamins & Assael, 1987), create attitude certainty about a
encourage new pro-environmental behaviors. product or a task, and importantly for the current research – strengthen
the positive link between attitudes and actual behavior (Rucker, Petty, &
2.2. A combined message approach to promote new pro-environmental Briñol, 2008).
behavior While not dealing with education or with encouraging and discour­
aging approaches, this literature supports the educational benefits of
In light of the notion that promoting new pro-environmental be­ combined messaging. Yet, prior work has not directly tested the effect of
haviors requires education, we turn to education psychology, focusing combining encouraging and discouraging messages in general, as well as
on approaches aimed at achieving desired behavioral change. One in the specific context of pro-environmental behavior. We posit that
highly relevant approach in education psychology is based on the given the major barrier to adopting new pro-environmental behaviors -
Adlerian “behavior reorientation” notion (Dreikurs & Soltz, 1991; Kvols, perceived lack of knowledge - and the consequent perceived uncertainty
1998). This educational approach involves communication practices and risk associated with these behaviors (Nygrén, Kontio, Lyytimäki,
that increase child cooperativeness (Adler, 1998; Dreikurs & Soltz, Varho, & Tapio, 2015; Ozaki, 2011), combining an encouraging and a
1991; Oryan & Gastil, 2013). The behavioral reorientation approach has discouraging message is an optimal way to modify such behaviors,
also been implemented in adults, to modify unhealthy recurrent be­ because of its educational and informational value. Specifically, since
haviors or habits to form new patterns (e.g., sex orientation – Murphy, perceived informativeness generally increases the likelihood of inno­
1992; sex pathologies – Kelly, 1982; diet – Vossenaar, Solomons, vation adoption (White et al., 2019), and subjective knowledge (the
Valdés-Ramos, & Anderson, 2011). The guiding principle of this perception that one is more informed) is sufficient to improve decisions
approach is the person’s freedom of choice (Gastil, 1993), while (Hadar, Sood, & Fox, 2013), we suggest that the reason that our com­
acknowledging the consequences of each possible choice of behavior: bined messaging approach can be useful for new pro-environmental
While certain actions will bring about positive consequences, other be­ behaviors is that it increases perceived informativeness, which in turn
haviors may bring negative consequences. Having this information, it is enhances behaviors. This approach is supported by research showing
then the person’s choice which behavior to perform (Dreikurs & Soltz, that combined messages are especially effective for behaviors that are
1991). This literature explains that prohibition alone (i.e., discouraging new to the target audience. In one study (Niemiec, Jones, Lischka, &
message) directs the individual towards a dead-end, without informa­ Champine, 2021), in the context of encouraging more gardening in the
tion about the existence of an alternative option. On the other hand, community – claimed to be a relatively new behavior – a message
mere allowance (i.e., encouraging message) also does not leave the in­ combining efficacy and norm-based arguments was one of the most
dividual free choice, and in addition – it does inform the individual effective ones. Another study (Walker, Gutierrez, Torres, & Bertozzi,
about the undesirable behavior. Therefore, prohibition coupled with 2006) in the context of adolescent education on HIV and risky sex, finds
allowance (i.e., combined message) juxtaposes the two behaviors, that a message combining encouragement of using condoms and a
providing the individual with an alternative, making the individual feel warning regarding emergency contraceptives was behaviorally effective
more informed, and therefore this approach is more effective in behavior in the short term while also increasing knowledge of the available so­
formation. lutions to avoid risky behavior by adolescents in the long term.
Relying on this education psychology literature, we argue that to Overall therefore we hypothesize that.
educate people and direct them to a desired behavior, the communicator
H1. Communication that combines encouraging and discouraging
needs to address the undesired behavior and revert people from
message in one will be more effective in promoting new pro-
repeating it, and at the same time offer a desired behavior that can
environmental behavior, compared with communication that uses
replace the undesired one, so that people have an alternative path to
only encouraging or only discouraging messages.
choose. In addition, in the context of a new behavior and the required
educational efforts needed to overcome knowledge gaps and uncer­ H2. Perceived informativeness of the communication mediates this
tainty, the combined communication approach elicits perceptions of effect.
greater informativeness because it includes both a what-not-to-do and a
what-to-do option, and is therefore more integrative. Informativeness 3. Method
and integration are key dimensions of market education (Osbaldiston &
Schott, 2012; White et al., 2019). We therefore propose that a message 3.1. Overview of studies
combining both encouraging and discouraging behaviors is more
effective than presenting only one of these messages. We tested our theory across multiple pro-environmental contexts to
Research studying the effectiveness of multiple messages within the increase the validity of our findings. We tested H1 in three large field
same campaign or advertisement offers further support for the combined experiments and a field survey (Studies 1–3). H2 was tested in another
messaging approach. For example, Olsen, Slotegraaf, and Chandukala three online studies (Studies 4–6). Studies 1 and 2 used Facebook
(2014) found that the number of claims used to communicate a message advertising platform to promote residential hydroponic agriculture
and its valence can influence the degree to which communication about (Study 1) and recycling e-waste (Study 2). In both studies we tested user
new environmentally friendly products changes brand attitudes. While engagement with three message types representing our theory
they studied negative vs. positive framing (e.g., "no pesticides" vs. (encouraging, discouraging, combined). Study 3 tested actual e-waste
"biodegradable") of an environmental message (and not encouraging vs. recycling behavior after exposure to one of our three message types. A
discouraging messages), they found that a hybrid message (providing post-experiment survey conducted in the same residential area
both negative and positive framing) is effective in elevating attitudes augmented this field experiment.
towards new environmentally friendly products. Overall, it seems like Studies 4, 5 and 6 were online studies testing the mediating role of
multi-faceted messages within one pro-social campaign contribute to the perceived message informativeness. Study 4 compared a new pro-

3
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

environmental behavior – recycling e-waste – to an established behavior session. See Web Appendix 1 for a full description of common research
– recycling paper. Study 5 tested the full theoretical model, using approaches in both studies, as well as budget and advertising consid­
geothermal home energy versus solar home energy as new and estab­ erations dictated by the local companies. While online campaign
lished behaviors, respectively. Finally, Study 6 used consumption of engagement is not direct evidence of actual behavior, literature suggests
cultured meat as a new behavior and explored the mediating role of a strong link between campaign engagement and further behavior (e.g.,
perceived informativeness by comparing our combined message to a Buller et al., 2022; Wagner, Fernandez, Jordan, & Saggese, 2019).
message that actually provided additional written information. All be­
haviors tested in this work were pretested on the extent to which they 3.3. Study 1: field study on facebook – residential hydroponic agriculture
were new to participants (see Web Appendix 7 for more details on the
pretests). Study 1 tested the impact of our three messages promoting urban
Statement of Research ethics: All procedures performed in studies residential hydroponic agriculture. Hydroponic agriculture is defined as
involving human participants were in accordance with the ethical a new pro-environmental behavior because it is a new activity that re­
standards of the authors’ institutional and/or national research com­ duces environmental harms across the entire life-cycle (EIO, 2012). In
mittee and with the 1964 Helsinki Declaration and its later amendments recent years, as a result of sustainability concerns and population
or comparable ethical standards. In addition, informed consent was growth, advanced residential growing methods using hydroponic tech­
obtained from all participants of Studies 4, 5 and 6. nology are developing into a stronger alternative to soil-based gardening
methods or conventional agriculture. For urban households interested in
growing their own vegetables but facing lack available space, advanced
3.2. Studies 1 and 2 – common procedures
residential agriculture is a plausible solution. Home-grown hydroponic
agriculture is considered a new behavior for the same reasons as e-waste
We chose two pro-environmental behaviors that were new to the
recycling – lack of public services and infrastructure – as well as lack of
studied target audience: residential hydroponic agriculture and col­
information. While there is a variety of vegetable growing hydroponic
lecting e-waste for recycling. We collaborated with local companies
systems that replace traditional gardening (Möhlman, 2015), public
operating in these areas and aimed to promote their activities through a
education and information about these systems is quite limited. We
Facebook ad campaign.1
collaborated with Green in the City (GC), a farm located on the roof of a
Both studies utilized an A/B testing approach, where we compared
central shopping mall in the largest metropolitan city in the focal
user reactions to advertising that used one of three messages: a com­
country. GC is one of the few farms that operates in the studied country
bined message, encouraging only message, and discouraging only mes­
and currently the largest one in terms of volume production. Aiming to
sage. This approach represented a 3-cell between subjects design. For
enhance urban sustainability, the farm offers workshops and educa­
our sample we chose three cities located within the same metropolis
tional activities to the public. The current experiment was used to attract
(with a maximal distance of 18 km from its center) in a developed
participants to these activities.
country (we name them here CITY1, CITY2 and CITY3). The three cities
resemble each other across a plethora of socioeconomic demographic
3.3.1. Procedure
variables including age groups, income, household size, and socio-
We designed a sponsored slideshow advertisement campaign (Gra­
demographic ranking.2 We prepared advertising messages using Face­
ham, Milner, Saul, & Pfaff, 2008; Lohse, 2013) to run on Facebook for
book Business platform. Each ad contained a Learn More button and a
three months from January 25th to April 25th, 2017. We limited the
clickable link leading to a specific landing page on the company’s
audience’s age range to 25–45, because this is the primary target market
website. Adult Facebook users in these three cities were randomly
of GC. The campaign’s optimization goal was based on Reach4 – the
assigned to receive one of the three messages, as shown in Table 1.
number of unique people who see the content of the ad. Since we opt for
Econometric analyses for both studies controlled for age group and
maximizing Reach, we limited the frequency – the number of times each
gender, relying on Facebook’s metrics for these data. Campaign results
user is exposed to the ad – so that each user was exposed to the ad up to
for Studies 1 and 2 were retrieved from Facebook’s metrics, and from the
six times, on average, for the entire campaign duration. This limitation
landing website of each campaign via Google analytics web service
intended to minimize negative feelings towards the ads due to excessive
(Google, 2019). Employing Facebook’s metrics, we measured users’
advertising frequency (Hitchon & Thorson, 1995). The campaign’s bid
direct response to our advertising, including link clicks, unique clicks,
strategy was lowest cost, and the charge was per impression – the
people taking action, and post engagements.3 The metrics on Google
number of times the ad appeared on the screen. The stimuli were chosen
Analytics included: users, visits, session duration and page views per
in accordance with past GC campaigns and depicted visuals of growing
hydroponic vegetables on a windowsill or on a rooftop (see Fig. 1 with
Table 1 English translations). Depending on city of residence (CITY1, CITY2 or
Conditions and cities in Studies 1 and 2.
CITY3), targeted Facebook users were exposed to one of our three
Study 1: Study 2: messages. The Facebook users then could engage with the ad via mul­
CITY1: discouraging combined tiple actions, such as clicking the ad, sharing, commenting, or clicking
CITY2: combined encouraging link to GC website. We designed a different landing page on GC’s
CITY3: encouraging discouraging website for each of the conditions, to capture post-Facebook engage­
ment behavior. Additional technical details about the campaign are
provided in Web Appendix 2.

1
Facebook advertising platform currently serves the highest number of active
users worldwide (Statista, 2020). Previous research successfully used Facebook
advertising for field experimentation (e.g., Gordon, Zettelmeyer, Bhargava, &
4
Chapsky, 2019; Mochon, Johnson, Schwartz, & Ariely, 2017; Yoon et al., 2018). Reach strategy allows advertisers to bid at auction for a certain number of
2
According to the country’s Central Bureau of Statistics; see WA Table 1 in impressions. This objective is designed to show ads to the maximum number of
Web Appendix 1. users in the defined target audience, while controlling for the frequency of ad
3
We tracked additional measures such as post comments or page engage­ exposure per user.
ment. Due to space limitations we report the full list of metrics in Web Ap­
pendix 1.

4
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

Fig. 1. Facebook ads in the three conditions (Study


1).
Note. To maintain the same message length across
conditions and avoid a possible confound of the
message appearing informative merely because it is
longer, our messages always included two phrases.
Thus, the encouraging message included two
encouraging phrases, the discouraging message
included two discouraging phrases, and the combined
message included one encouraging and one discour­
aging phrase.

day of the week). We therefore first conducted a Wooldridge test for


Table 2
autocorrelation in panel data (Drukker, 2003) finding that the null hy­
Facebook metrics and proportion test results (Study 1).
pothesis of no serial correlation was not rejected. We then conducted a
City/Message Reach Link Unique Post People Poisson regression analysis that controlled for possible time trend and
phrasing clicksa clicksb engagementc taking
seasonality, taking into account that our dependent variables were count
actiond
variables. Notably, there was a significant proportion of observations
CITY3 88,160 0.39% 0.75% 6.86% 6.13% (days) where the dependent variables equaled zero. In particular, the
encouraging
CITY1 59,592 0.29% 0.59% 7.65% 6.57%
average share of zeros in both link clicks and unique clicks was 69%,
discouraging with this share being maximal for the discouraging treatment (79%) and
CITY2 48,131 0.63%* 1.32%* 27.83%* 21.83%*** roughly the same for the encouraging and combined treatment (64% and
combined 65% respectively). The share of zeros in both people taking action and
Note. The table shows a consistent advantage of the combined condition over the post engagement was 17% with this share being minimal (6%) for the
encouraging or discouraging conditions, on all relevant metrics. combined treatment and maximal (24%) for the discouraging treatment.
Significance levels for difference from the other two cities: *p < .05, **p < .01, See WA Table 4b for zeros distribution by study and by treatment. Given
***p < .001. the large share of zeros, we employed the zero-inflated negative bino­
a
Link clicks is the number of clicks on links within the ad that led to advertiser- mial regression, (ZINB command in Stata 15.1.6 Table 3 portrays the
specified destinations, on or off Facebook. This metric counts link clicks on the results of this model for the post engagement metric. We focused on this
ad’s text, media or call-to-action, that link to destinations or experiences spec­
metric given its inclusive nature (capturing link clicks, likes, comments,
ified by the advertiser. It excludes clicks on content or links in the comments
section of a post.
b
Unique clicks counts the number of people who performed a link click. The Table 3
metric counts people, not actions. ZINB regression model coefficients (Study 1).
c
Post engagement is the total number of actions that people take involving an
ad on Facebook. Post engagement includes all actions that people take involving Post engagement (SE)
ad while it is running. It includes actions such as reacting to, commenting on or CITY1 (discouraging) − 0.762*** (0.022)
sharing the ad, claiming an offer, viewing a photo or video, or clicking on a link. CITY3 (encouraging) − 0.399*** (0.015)
d Constant − 1.563*** (0.133)
People taking action is defined as the total number of actions people took that
are attributed to the ad. Actions include (but not limited to) clicks or conver­ PseudoR2 26%
sations, content viewed, comments on posts, page likes, post shares, and Observations 3905
messaging replies. If the same person does multiple actions, they are included Note. The table shows a significant difference in post engagement
once within this number. between the combined condition (CITY 2) and the other two condi­
tions (CITY 1 & CITY 3).
2
3.3.2. Results Due to space limitation, the table does not report a list of control
Facebook Metrics Analysis. Table 2 presents four Facebook metrics variables. These include: the day in the week, the day of campaign,
obtained for each city.5 See notes below Table 2 for detailed descriptions age, and gender. In Table 5b we demonstrate how the coefficient es­
timates change with and without the inclusion of each set of control
of the metrics. Since Reach (the number of unique people who were
variables (socio-demographics and weekdays/day of campaign).
exposed to the campaign in each message) was different in each city, the 3
Randomization tests on control variables (age, gender and day in the
table presents the percentage out of the total Reach in each city. To test week) revealed non-significant differences between the 3 cities on
the significance of the results, we employed Stata 15.1 proportion test- these variables (all p’s > 0.4).
prtesti- (StataCorp, 2019). Proportion tests revealed that the message 4
The other remaining metrics (link clicks, unique link clicks and
that performed best in all metrics was the combined one. For example, people taking actions) were employed in Table 5a as a robustness
21.83% of the people who were exposed to the combined message took check.
an action, compared with 6.13% and 6.57% for the encouraging and Significance level for difference from the combined campaign: *p <
discouraging messages, respectively. .05, **p < .01, ***p < .001.
Our data had the structure of a panel with time series dimension,
which might be subject to bias due to serial correlation, as well as the
existence of time trend and seasonality (e.g., fixed effects for the specific

6
ZINB fits a zero-inflated negative binomial (ZINB) model to over-dispersed
count data with excess zero counts. The ZINB model assumes that the excess
5
See WA Table 4a in Web Appendix 2 for the full report with additional zero counts come from a logit model and the remaining counts come from a
Facebook metrics. negative binomial model.

5
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

and shares (see Table 2 for details)). In addition, based on the occur­ Table 5a
rences of zeros in the data (see WA Table 4b) post engagement has the Robustness check for the remaining three metrics (Study 1) - ZINB regression
lowest share of zeros (particularly in study 1), which allows for a richer model results.
dataset-based estimation. Finally, post engagement is a common metric Link clicks Unique clicks People taking
in previous works studying the effectiveness of Facebook campaigns action
(Eriksson, Sjöberg, Rosenbröijer, & Fagerstrøm, 2019; Menon et al., CITY1 (discouraging) − 0.251*** − 0.247*** − 0.759***
2019).7 All four regressions treated CITY2 with the combined message rowhead (0.038) (0.015) (0.023)
as the omitted one, and used clustered robust standard errors that allow CITY3 (encouraging) − 0.036*** − 0.0133*A − 0.441***
rowhead (0.009) (0.008) (0.012)
for intragroup correlation, with city code as the clustering variable.
Constant rowhead − 6.892*** − 6.845*** − 3.339***
Google Analytics Results. Users who clicked through to GC’s website (0.350) (0.426) (0.051)
were transferred to one of the three landing pages that we created for
Note. The table shows significant difference in post engagement between the
each of the messages. Table 4 shows Google Analytics results for Face­
combined condition (CITY 2) and the other two conditions (CITY 1 & CITY 3).
book user behavior after visiting the landing pages.8 Proportion tests
A When corrected for alpha-inflated significance level, this variable becomes
using Stata 15.1 software showed the advantage of the combined mes­ insignificant (Bonferroni corrected, α levels divided by 3).
sage on all metrics. For example, after seeing the combined message on Significance level for difference from the other two cities: *p < .05, **p < .01,
Facebook, the average duration of browsing the GC’s website was 3.55 ***p < .001.
times higher than after seeing the discouraging message and 6 times
higher than after seeing the encouraging message.
We conducted two robustness checks. First, we ran the ZINB
regression shown in Table 3 on the remaining three metrics. Second, we Table 5b
ran the ZINB regression shown in Table 3, for post engagement, twice: Regression coefficients of the discouraging and encouraging messages without
once, with the socio-demographics as control variables, and then again, the two sets of control variables. Dependent variable: Post engagement (Study
with the time related variables (day of the week and day of the 1).
campaign). Table 5a portrays the results of the first robustness check. As Study 1 GC Study 1 GC
can be seen, for link clicks and unique clicks, the coefficient estimates of discouraging encouraging
the treatment variables are different in size, compared to the coefficients Full mode − 0.762*** − 0.399***
in Table 3 (post engagement), however, the order of magnitude is Without socio-demographics − 0.851*** − 0.453***
maintained, such that the encouraging message performs better than the Without weekdays and day of the − 0.854*** − 0.845***
campaign
discouraging one, and both coefficients are negative compared to the
combined message. The coefficients of people taking action are very ***p < .001.
similar in size to those in Table 3.
Table 5b demonstrates that the coefficient estimates of the encour­ 3.4. Study 2: facebook field study – E-waste recycling
aging and discouraging messages maintain their significance and order
of importance when the regression model from Table 3 is regressed E-waste is defined as any equipment or appliance designed to operate
without the socio-demographics, or without the time related variables by an electrical current or an electromagnetic field that reached its
(day of the week and day of the campaign). useful end of life and is not used or not functional anymore. About 45
million metric tons of e-waste were generated globally in 2016 (Baldé,
Forti, Gray, Kuehr, & Stegmann, 2017) and E-waste collection for
recycling is the fastest growing waste stream in the world (Abdelbasir,
Hassan, Kamel, & El-Nasr, 2018). It is considered a new behavior
Table 4 (Cheng, Hung, Tsai, & Chou, 2020; Saphores, Oladele, & Shapiro, 2012).
Google analytics results for Facebook user behavior on GC’s website (Study 1). Some key reasons for this are that in many counties the public is still not
Message Users Visits Avg. Avg. Session The total aware of various options to return e-waste and the infrastructure for
phrasing per per page session duration number e-waste collection is not developed yet. Even where there are public
reach reacha views duration per visit of pages services available, the behavior has not reached the level of a recurring
per in in viewed
habit that citizens perform frequently, as much as recycling paper or
session seconds seconds per
reach plastics. In the studied country, 130,000 tons of e-waste were generated
annually (Shamai, 2017), but it was still considered a novel environ­
Encouraging 0.05% 0.05% 1.02 7.39 0.06 0.053%
Discouraging 0.08% 0.09% 1.06 13.55 0.15 0.091% mental context (Steiner, 2015). We collaborated with the leading com­
Combined 0.24% 0.25% 1.22** 44.04*** 0.48*** 0.303%* pany for recycling e-waste and batteries in the studied country, which
*** *** was also the franchisee for e-waste collection in the sampled cities
Note. The table shows consistent advantage of the combined condition on all studied (called here COMPANY).
relevant metrics.
*p < .05, **p < .01, ***p < .001. 3.4.1. Procedure
a
Visits refers to the number of times visitors entered the website. As in Study 1, a three-month sponsored campaign was designed to
run on Facebook from April 1 to June 30, 2017. The campaign aimed to
promote e-waste recycling into designated bins located at different
places in the city. Per COMPANY’s requirement, Study 2 campaign was
7
designed as a Link-Click campaign – where advertisers pay per number of
See full table with all Facebook covariates in Web Appendix 2, Table 5. See
clicks on their ad.9 Thus, we could not control for frequency of exposure
also Web Appendix 2 Table 6 for results of differences in marginal effects.
8
Since we did not have access to GC Google Analytics data, we received these
and therefore the optimization goal was link clicks, the bid strategy was
data from both organizations as aggregate data (e.g., the grand total of visits or lowest cost, and the charge was for link clicks. We also did not specify
number of pages viewed, and the average of page views per session). Therefore,
we were not able to calculate the medians of these performance indicators over
9
time, but rather we used the (normalized per reach) total volumes of these These decisions were based on considerations of Facebook advertising costs
indicators or the average of averages (over time). (https://revealbot.com/blog/facebook-advertising-costs/).

6
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

age range in this study, per COMPANY’s strategy. We collaborated with advertising reveal the advantage of our combined message over
COMPANY’s graphic designer and created three banner-format ads, encouraging or discouraging only messages in attracting user engage­
corresponding with our three messages. The ads provided a visual of an ment on Facebook, as well as post-exposure behavior on the landing
e-waste pile in a residential setting, and contained either an encour­ pages of the organizations promoting new pro-environmental behaviors.
aging, a discouraging or a combined message. As in Study 1, each ad was These results remain largely stable across various metrics and analysis
linked to a different landing page on COMPANY’s website. All three approaches. Importantly, the combined message performed especially
landing pages looked the same and provided information about the well on activities related to receiving more information. This result is
specific locations within the city where e-waste can be brought, as well important from a theoretical standpoint, as our suggestion is that
as general information on e-waste. See ad screenshots and additional perceived informativeness is the driving force behind the increased
methodological details in Web Appendix 3. persuasiveness of the combined message, while the message itself does
not actually contain more information.
3.4.2. Results
Facebook metrics analysis. Table 6 presents the results of proportion 3.5. Study 3: field study testing actual E-waste recycling behavior among
tests on Facebook metrics for the campaign as a percentage of the total city residents
reach in each city.10 The table shows that for the metrics of people
taking action and post engagement, the combined message performs The purpose of this field study was to augment our Facebook
better, whereas for clicks (all), unique clicks (all) the discouraging engagement results with a test of actual new pro-environmental
message performs better. These results partially confirm H1. behavior - e-waste collection for recycling. In addition, although we
Further, as in Study 1, we performed a ZIP model regression on these controlled for various aspects in the three cities examined in Studies 1
metrics. As in Study 1, all regressions treated CITY1 (the combined and 2, and switched the cities around between the studies, and were still
message) as the omitted one. Table 7 provides results for the same four able to replicate our results, we still wanted to test the effect of our
metrics as in Study 1.11 The table shows that the combined (omitted) messages within the same city. on. The study involved posting signs with
category performs better than the other messages on all four metrics. one of each of our three message types on e-waste collection bins at
These results are similar to those reported in Study 1, though the R2s several locations in an urban area, and weighing the contents of
were much higher here. collected waste in subsequent months. We also ran a survey among the
Google analytics metrics. Table 8 presents results of Google Ana­ residents of the same areas to validate whether the objective behavior
lytics metrics for Facebook users who clicked through the link in the ads corresponds with residents’ perceptions. We collaborated with a city in
and were redirected to COMPANY’s website landing pages. As in Study the central district of the studied country. We chose this city for several
1, the table shows that the combined message was superior to the other reasons. First, this city was not part of Studies 1 and 2 sample, and
two messages on all metrics. For example, among the total number of therefore we could be sure that our Facebook campaigns did not inter­
people who were exposed to the combined message, the visits per reach vene with the current experiment. Second, it was contracted with
were 1.3%, compared to 0.7% and 0.8% in the encouraging and COMPANY which is recognized by the government as an authorized
discouraging messages, respectively. recycling service provider and the leading in the country. We therefore
As in Study 1, we conducted two robustness checks. First, we ran the had access to the collection bins for posting and for later receiving
ZINB regression shown in Table 7 on the remaining three metrics. Sec­ weights of e-waste collected from the bins. Third, at the time of the
ond, we ran the ZINB model shown in Table 7, for post engagement as a experiment, this was the only city that utilized e-waste collection bins,
dependent variable, twice: once, with the socio-demographics as control which were large enough to carry prominent posters with our messages.
variables, and then again, with the time related variable (day of the Fourth, the city’s municipality was highly committed to sustainability
week and day of the campaign). Table 8a portrays the results of the first goals and was willing to collaborate with us. The e-waste collection bins
robustness check. As can be seen, the coefficient estimates of the treat­ were located across the city on 12 streets, which we divided into three
ment maintain their ordinality, with the encouraging message per­ clusters that were geographically far enough from each other to repre­
forming better than the discouraging one, and both coefficients are sent three separate areas (Area 1, Area 2, and Area 3), based on prox­
negative compared to the combined message. imity and neighborhood affiliation (e.g., minimizing situations where a
Table 8b demonstrates that the coefficient estimates of the encour­ person in Area 1 is also exposed to the campaign in Area 3). The
aging and discouraging messages maintain their significance and order campaign was conducted in coordination with the municipality and the
of importance when socio-demographic variables are omitted from the partner COMPANY starting February 2019. To be able to measure
model. However, the coefficients become insignificant when omitting changes over time in the quantities collected, we received data from the
the time-related variables from the model. In order to study these results COMPANY from January 2018 to May 2020.
deeper, we performed a moderation analysis for the effect of day-of-the- When we received the collection weights data, we were informed
week on the message. Results were highly significant for the interaction that during 2019 some of the e-waste collection bins were vandalized
term (PV < 0.0001). These results imply that while the effect of the and the collected e-waste was stolen. To assure more equal conditions
messages is overall significant, it was especially significant on particular we therefore included in the analyses weights from bins located inside
days of the week – specifically on weekdays rather than the weekends. local schools, because these were protected behind a fence. The schools
This may reflect general reduction in Facebook usage on weekend.12 were located in the three different areas identified for the experiment.
All schools were public schools participating in a recycling program for
3.4.3. Discussion – studies 1 and 2 schools (“the e-waste spaceship”) independently ran by the COMPANY.
Results of two large scale field experiments utilizing Facebook As these were all elementary schools serving 1st-8th graders, the
structure and size of the schools are comparable. Our campaign involved
three different variations, so any differences between the three areas can
10
See WA Table 7 in Web Appendix 3 for the full report with additional be attributed to the differences between our messages. Thus, our sample
Facebook metrics.
11
See also WA Table 8 which reports the results of ZIP regressions on the full
list of metrics, and WA Table 9 which reports the marginal effects for this
campaign.
12
https://blog.hootsuite.com/best-time-to-post-on-facebook-twitter-i
nstagram/.

7
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

Table 6
Facebook metrics and proportion test results (Study 2).
Message phrasing Reach Link clicks Unique clicks Post engagement People taking action

CITY2 (encouraging) 77,205 4.2% 3.1% 0.3% 0.3%


CITY3 (discouraging) 109,476 5.7%** 4.2%** 0.4% 0.4%
CITY1 (combined) 105,459 5.4% 4.0% 0.6%*** 0.5%***

Note. The table shows a significant advantage in post engagements for the combined message condition over the other two conditions.
Significance levels for difference from the other two cities: *p < .05, **p < .01, ***p < .001.

Table 7 Table 8b
ZINB regression model coefficients (Study 2). Regression coefficients of the discouraging and encouraging messages without
the two sets of control variables. Dependent variable: Post engagement (Study
Post engagement (SE)
2).
CITY2 (encouraging) − 0.145 (0.057)
CITY3 (discouraging) − 0.414*** (0.016) Study 2 e-waste Study 2 e-waste
Constant − 1.611*** (0.135) discouraging encouraging
Observations 3905 Full model − 0.414*** − 0.145***
PseudoR2 44.6% Without socio-demographics − 0.421*** − 0.165***
1 Without weekdays and day of 0.0101 − 0.0151
Due to space limitation, the table does not report a list of control
the campaign
variables. These include: the day in the week, the day of campaign,
1
age, and gender. In Table 8b we demonstrate how the coefficient es­ The coefficients are insignificant, p > .8.
timates change when including and excluding each set of control ***p < .001.
variables (socio-demographics and weekdays/day of campaign).
2
Randomization tests on control variables (age, gender and day in the included six schools: one in Area 1, three in Area 2, and two in Area 3.13
week) revealed non-significant differences between the 3 cities on
As mentioned earlier, each area was randomly assigned to one of our
these variables (all p’s > 0.9).
three messages (encouraging – Area 3, discouraging – Area 2, and
Significance level for difference from the combined message
campaign: *p < .05, **p < .01, ***p < .001. combined – Area 1). The 3 types of posters were attached to the bins

Table 8
Google analytics results (Study 2).
Users per Visits per Avg. page views per Avg. session duration in Session duration per visit in The total number of pages viewed
reach reach1 session seconds seconds per reach

CITY 2 0.9% 0.7% 1.5 50.2 4.5% 1.4%


(encouraging)
CITY 3 1.1% 0.8% 1.44 48.3 3.4% 1.3%
(discouraging)
CITY 1 (combined) 2.2%* 1.3%* 1.8** 115.7** 5.5%* 2.0%*

Note. The table shows a significant advantage in actual visits for the combined message condition over the other two conditions.
Significance level for difference from the combined message campaign: *p < .05, **p < .01.

according to their designation. The signs were posted on the bins on


Table 8a
February 1, 2019 and left on the bins throughout the end of the school
Robustness check for the remaining three metrics (Study 2): ZINB regression
year in June. Given that the signs required some time to get noticed and
model results.
begin to have an impact and given the fact that some of the e-waste in
Link clicks Unique clicks People taking action the bins accumulated before the campaign, we used February 2019 as a
Encouraging − 0.246*** (0.058) − 0.176*** (0.064) − 0.180*** (0.058) reference point month to measure the impact of our messages on recy­
Discouraging − 0.510*** (0.051) − 0.385*** (0.058) − 0.397*** (0.053) cling compared to the subsequent months of March–June 2019 (till the
Constant − 1.288*** (0.135) − 1.659*** (0.152) − 1.611*** (0.135)
end of the school year at the end of June).14 We next report the field
Note. Due to space limitation, the table does not report a list of control variables. experiment’s results and then a field survey that complemented it.
These include: the day in the week, the day of campaign, age, and gender.
When corrected for alpha-inflated significance level (Bonferroni corrected, α 3.5.1. Results – field experiment
0.001 The data included the weight of the e-waste collected in the bins
levels divided by 3), all variables remain significant with p < = 0.0003
3
Significance levels for difference from the combined message campaign: *p <
(mostly small electronic equipment and batteries) from each collection
.05, **p < .01, ***p < .001. trip (i.e., the same school location bin could have been emptied multiple
times during the campaign). Hence, we compared the three areas on the
difference between the average weight of e-waste per collection trip in
February to March–June 2019. Table 9 summarizes the collection

13
We cannot exclude the possibility for certain within-school biases (for
example, Area A included only one school), but it is important to note that not
only students were able to use the collection bins, therefore behavioral differ­
ences may reflect campaign message influences, as predicted.
14
In addition, for some areas (i.e., Area 2) we had no data before February
2019.

8
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

Table 9 Table 10
Results of E-waste collection weights in the 3 areas (Study 3). Statistics for field survey responses – percent of respondents who answered
Area 1 Area 2 Area 3
“yes”.
(combined) (discouraging) (encouraging) Question Area 1 Area 2 Area 3 Chi2
(combined) (discouraging) (encouraging) (p-
Average weight in 33 kg 57.5 kg 87.33 kg
value)
February 2019
Average weight in 130 kg 53 kg 55 kg Remembered 24.5% 2.1% 14.9% 10.2
March-June 2019 recycling bins (.006)
Difference in % Increase of Decrease of 8% Decrease of 37% Remembered 18.9% 2.1% 10.6% 7.2
294% messages on (.027)
Number of collection 1 4 3 the bins
trips in February Remembered the 18.9% 0.0% 8.5% 10.37
2019 specific (.006)
Number of collection 2 1 2 message in
trips in March-June their area
2019 Reported 41.5% 12.8% 34.0% 10.4
recycling (.006)
Note. The table shows the weights of e-waste collected from the bins in the
following the
different schools. campaign

Note. The table shows advantage for post-campaign behavior in the combined
weights results as the sum of collected weights divided by number of
message area, compared with the other two areas.
trips.
We conducted a number of robustness checks to demonstrate that the
significantly more respondents in the combined message area (Area 1)
effect we found was not an artifact of another variable. First, to rule out
remembered the bins, the signs and the message, and were significantly
the potential effect of the schools’ independent education campaign
more likely to recycle following a campaign (see Table 10).
(December 2018–February 2019) we compared the February data to
weights from an earlier period of January 2019. This data was available
3.5.4. Discussion
only for the combined message Area 1. The data showed a decreasing
Our third field experiment – with an actual new pro-environmental
trend (that may be explained by the tapering off of the educational
behavior – suggests that, across conditions, time periods, and robust­
campaign): from 250 kg in January to 33 kg in February. To rule out
ness checks, only Area 1 – associated with the combined message
potential seasonal effects we compared random but similar-in-length
approach – showed an increase in e-waste collection during the period of
time periods in Area 3 – the only area for which we had data long
our campaign. In addition, results of the complimentary field survey
before and long after the campaign. Specifically, comparing February
showed that residents in the area of the combined message were more
2018 (a year before the campaign) to March–June 2018 reveals a
likely to remember the campaign, and were more likely to recycle e-
decrease from 170 kg to 140 kg (− 16.5%) and comparing February 2020
waste following the campaign (based on self-reporting).
(a year after the campaign) to March–June 2020 reveals a decrease from
We now turn to online controlled studies that shed light on the
85 kg to 24 kg (− 71.7%). We had no data for Area 2 that could be used
process underlying the superior effect of a combined message on new
for any of these comparisons.
pro-environmental behaviors.
3.5.2. Follow-up field survey
To complement the field study, we also conducted a survey in the city 3.6. Study 4: testing the mediator of perceived informativeness
among 465 passers-by in three outdoors spots within the city, corre­
sponding to the three areas of the campaign. Surveyors located them­ Study 4 is an online study comparing perceived informativeness of
selves in these spots with i-pads and ran the survey among passers-by the combined message for a new pro-environmental behavior (e-waste
during the first week of Marr, 2019, in return for a coffee and pastry recycling) versus an established one (paper recycling).
voucher. The survey included questions about respondent’s recent
e-waste recycling activities, as well as awareness to our campaign. 3.6.1. Procedure
Specifically, the survey asked the following four binary (yes/no) Four hundred and fifteen participants were recruited through a panel
questions. company (Mage = 42.4, 50.8% female) and randomly assigned into one
of four conditions in a 2 (message: combined/control) X 2 (recycling:
• Have you noticed an e-waste recycling bin(s) located in your area of new (e-waste)/established (paper)) between subjects design. A power
residence? analysis using G*Power 3.1.9.7, specifying a small-medium effect of f =
• Have you noticed any signs posted on the e-waste recycling bin(s) 0.25 and numerator df = 1, revealed that a total sample size of 128
located in your area of residence? participants would provide 80% power, at α = 0.05. Hence, the sample
• Do you recall a message posted on the collection bin(s) in your area size was appropriate for this study.
of residence? Participants first imagined that they were reading an announcement
• The last time you brought e-waste for recycling – was it following published by their local municipality. They then read a combined or a
exposure to a campaign? control message, persuading them to recycle paper (not considered new)
or e-waste (considered new). The control message contained only in­
3.5.3. Results formation and no call for action, and read: “Recycling e-waste(paper) is
We used data from the 147 respondents who indicated their address, good for our planet. It helps eliminate contamination of our land and
so we could associate them with a particular area and therefore a water.” The combined message included the same information, and in
particular message in our field experiment. A power analysis using addition: “Don’t throw e-waste(paper) in the regular garbage bin.
G*Power 3.1.9.7 (Faul, Erdfelder, Lang, & Buchner, 2007) specifying a Discard unneeded e-waste(paper) in special recycling bins.” After
small-medium effect of W = 0.3 and df = 2 revealed that a total sample reading the message participants filled out two items pertaining to
size of 108 participants would provide 80% power, at α = 0.05. Hence, behavior intentions (e.g., “How likely are you to bring your e-waste
the sample size was appropriate for this study. (paper) to a nearby recycling bin?”; 1-not likely at all; 7-very likely, r =
We found that, consistently with the results of the field experiment, 0.623). Next, participants filled out two items pertaining to message

9
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

perceived informativeness (e.g., “To what extent did the message pre­ read a message encouraging them to implement a solar/geothermal
sent new information about pro-environmental behavior?”; 1-definitely energy solution in their home. All messages started with “Solar energy is
not, 7-definitely yes, r = 0.748). Finally, participants filled out de­ beneficial to our environment.” Then, the discouraging message read,
mographic items. “Don’t close your eyes to our world. Don’t leave your home without a
solar energy solution.” The encouraging message read, “Open your eyes
3.6.2. Results to our world. Install a solar energy solution in your home.” The com­
Two-way ANOVA demonstrated a significant interaction of message bined message read, “Don’t close your eyes to our world. Install a solar
type and recycling context on behavioral intentions (F(1, 396) = 5.02, p energy solution in your home.” The ads are presented in Web Appendix
= .026, η2p = .012). Planned contrasts revealed that while there was no 4. After reading the message, participants filled out five behavior
effect of message on behavioral intentions when the message was about intention items (e.g., “Following this ad, how likely are you to consider
paper recycling (Mcombined = 4.00, Mcontrol = 4.25, F(1,396) = 1.15, p = using solar energy in your home?”; 1-definitely not, 7-definitely yes;
.285), participants who read a combined message about e-waste recy­ Cronbach α = 0.940). After that, participants filled out four items per­
cling were more likely to recycle (M = 4.71) compared with participants taining to message perceived informativeness (e.g., “After reading this
who read the control message (M = 4.26, F(1,396) = 4.07, p = .044, η2p ad I feel I can make a more informed decision about using solar energy”;
= .052). We also found a significant main effect of recycling context, 1-definitely not, 7-definitely yes; Cronbach α = 0.955). Finally, partic­
with intention to recycle e-waste being significantly higher (M = 4.47) ipants filled out demographic items.
than intention to recycle paper (M = 4.13, F(1,396) = 4.56, p = .033, η2p
= .011). We found no significant main effect of message type (p = .522). 3.7.2. Results
Mediation. To test the moderated mediation of perceived informa­ Three participants did not compete more than half of the study and
tiveness we employed PROCESS Model 8 regression add-in (10,000 their data was omitted from the analyses. As expected, participants we
bootstrapped samples; Hayes, 2013) with message type as predictor, significantly more knowledgeable about solar energy (M = 3.81) than
pro-environmental context as the moderator, perceived informativeness about geothermal energy (M = 2.38, F(2, 361) = 95.67, p < .001). Two-
as the mediator and behavior intentions as the outcome. We found that way ANOVA revealed a significant interaction of context (solar/
while the direct effect of the interaction of message type and context on geothermal energy) and message type and on behavioral intentions (F(2,
behavior intentions is significant (B = 0.643, SE = 0.321, t = 2.00, p = 361) = 3.37, p = .035, η2p = .018). Planned contrasts revealed that while
.04, 95% CI: [0.012, 1.274]), the conditional indirect effect is insignif­ within the solar energy condition there was no difference in behavioral
icant when perceived informativeness is entered into the model (B = intentions between the various messages, within geothermal energy
0.538, SE = 0.350, t = 22.54, p = .125, 95% CI: [-0.150, 1.226]). These behavioral intentions were significantly higher after the combined
results suggest that perceived informativeness mediates the effect of the message (M = 4.60) than after the encouraging message (M = 3.99, p =
interaction. Table 11 presents full results of this analysis and Fig. 2 .024) or the discouraging message (M = 3.76, p = .002). a similar
presents the mediation model. pattern was found for perceived informativeness: while there was not
significant difference between the messages within solar energy condi­
3.7. Study 5: testing the full moderated mediation model tion, for geothermal energy the combined message elicited significantly
higher perceived informativeness (M = 3.97), compared with the
Study 5 is an online study testing the full moderated mediation encouraging message (M = 2.27, p < .001) and the discouraging mes­
model whereby pro-environmental behavior (new/established) moder­ sage (M = 2.26, p < .001).
ates the effect of message (discouraging/encouraging/combined) on Mediation. We conducted a moderated mediation analysis employing
behavioral intentions via the mediation of perceived informativeness. In PROCESS Model 8 regression add-in (10,000 bootstrapped samples;
this study, participants read and responded to a message encouraging Hayes, 2013) with message type as a predictor, pro-environmental
them to implement geothermal (new)/solar (established) energy solu­ context as the moderator, perceived informativeness as the mediator
tions in their home. and behavior intentions as the outcome. We found that while the direct
effect of the interaction of message type and context on behavior in­
3.7.1. Procedure tentions was not significant (p = .234), the indirect effect was significant
Three hundred and seventy participants were recruited through (p = .003). Specifically, the path from the interaction to informativeness
Prolific Academic participant pool (Mage = 34.9, 60.4% female). They was significant (B = 0.597, SE = 0.203, t = 2.948, p = .003, 95% CI:
were randomly assigned into one of six conditions in a 2 (solar (new)/ [0.199, 0.996]) and the path from informativeness to behavior was
geothermal (established) energy) X 3 (message: discouraging/encour­ significant (B = 0.4341, SE = 0.0436, t = 9.951, p = .000, 95% CI: [
aging/combined) between subjects design. A power analysis using 0.348, 0.519]), whereas the direct path from the interaction to behavior
G*Power 3.1.9.7, specifying a small-medium effect of f = 0.25 and was not (B = 0.203, SE = 0.170, t = 1.193, p = .234, 95% CI: [-0.132,
numerator df = 2, revealed that a total sample size of 158 participants 0.538]), suggesting full mediation by informativeness.
would provide 80% power, at α = 0.05. Hence, the sample size was
appropriate for this study. 3.7.3. Discussion
Participants first filled out three items pertaining to their knowledge Study 5 implemented a full interaction design in order to test our
about solar/geothermal energy (“I am knowledgeable about solar en­ entire conceptual model. These results further support H2, suggesting
ergy”; 1-definitely not, 7-definitely yes; Cronbach α = 0.937). They then that perceived informativeness mediates the effect of the interaction of

Table 11
Moderated mediation for behavioral intentions (Study 4).
Antecedent Coefficient SE t p LCI UCI

X (Message) − 1.24 0.52 − 2.41 0.01 − 2.2589 − .2301


W (Context) − 1.08 0.50 − 2.14 0.03 − 2.0685 − .0908
M (Perceived Informativeness) .1067 0.05 2.32 0.02 .0164 .1971
Interaction 0.64 0.32 2.00 0.046 .0116 1.2737
Constant 5.97 0.84 7.12 0.000 4.3220 7.6205
Model Summary R2 = .0372
F(4,395) = 3.815, p = .005

10
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

Fig. 2. Perceived informativeness mediates the greater effect of a combined message on new pro-environmental behaviors (Study 4).

message and pro-environmental context on behavior. 3.8.2. Results


Main analyses. We found a significant difference between the con­
ditions on intentions to consume cultured meat (F(2, 338) = 5.95, p =
3.8. Study 6: using perceived informativeness to improve compliance with .003, ŋ2 = 0.26). Planned contrasts also revealed the predicted effect. As
message promoting new pro-environmental behaviors can be seen in Fig. 3a, the data supported our prediction that the
encouraging message with no information was inferior to both the
The purpose of Study 6 was to replicate the mediating role of combined and the encouraging plus information messages (t(338) =
perceived informativeness in the effect of a combined message on a 22.9, p < .001, Cohen’s d = 0.384). No significant differences were found
different new pro-environmental behavior – consumption of cultured between the combined and the encouraging plus information message. A
meat. Further, rather than directly measuring perceived informative­ similar pattern was found with regards to perceived informativeness,
ness, in this study we manipulated it. Finally, we tested whether adding suggesting that the combined message is deemed as informative as the
more information to a sub-optimal message – an encouraging message – encouraging message plus information (F(2, 338) = 32.9, p < .001;
can reach levels of compliance similar to a combined message. planned contrasts: t(338) = 28.1, p < .001, Cohen’s d = 1.14). See
Fig. 3b.
3.8.1. Procedure Additional analyses. Participants in the different conditions did not
We prepared three ads promoting cultured meat, corresponding with differ on previous knowledge about cultured meat (p = .655) or on
three conditions: a combined message, an encouraging message, and an innovativeness (p = .789). Current habits among the participant’s circle
encouraging message containing additional information about cultured significantly predicted intentions to consume cultured meat (p = .001),
meat. We expected the encouraging + information message to perform but did not interact significantly with condition (p = .575). We found a
as well as the combined message. The ad read, “Cultured meat is marginal effect of innovativeness on intentions to consume cultured
beneficial to our environment. Open (Combined message: Don’t close) meat (p = .079) but no interaction with condition (p = .873).
your eyes to our world. Include cultured meat in your meal!” in the Mediation of perceived informativeness. We conducted a mediation
additional information condition, a line appeared in small letters after analysis using the PROCESS macro model 4 (Hayes, 2013), with message
the first phrase and read, “Cultured meat is real meat produced via cell type as a predicting variable, intentions to consume cultured meat as an
cultures in the laboratory, rather than from a whole slaughtered ani­ outcome, and perceived informativeness as the mediator. The model
mal.”. The three ads are presented in Web Appendix 5. suggested a significant mediation (B = 0.381, SE = 0.064, t = 12.2, 95%
Three hundred forty one undergraduate students from a North­ CI: [0.259, 0.508]). Further, the direct effect of message on intentions in
eastern American university participated in this study for course credit the model that included perceived informativeness became
(mean age 19.5, 48.5% women). A power analysis using G*Power non-significant (B = − 0.0472, SE = 0.0879, t = − 0.5364, 95% CI:
3.1.9.7, specifying a small-medium effect of f = 0.25 and numerator df = [-0.220, 0.126]), suggesting a full mediation.
2, revealed that a total sample size of 158 participants would provide
80% power, at α = 0.05. Hence, the sample size was appropriate for this 3.8.3. Discussion
study. Study 6 replicated the findings in Studies 1–5, showing the superi­
Participants first indicated their familiarity and knowledge about ority of the combined message in eliciting intentions to consume
cultured meat on four 7-pt. scale items (e.g., “I am knowledgeable about cultured meat. We also find that perceived informativeness mediates this
cultured meat”; Cronbach’s α = 0.844). Next, participants were effect. In this study we compared the combined message with an
randomly assigned to inspect one of the three ads for cultured meat. encouraging message, and with an encouraging message with informa­
After that, they filled out a five-item 7-pt. scale of intentions to consume tion, and were able to show that the combined message elicits percep­
cultured meat (e.g., “Following this ad I will plan to consume cultured tions of informativeness, as much as the encouraging message with
meat in the future”; Cronbach’s α = 0.904), as well as a five-item 7 = pt. information. As it is not always feasible to provide information in
scale of the extent to which they felt informed after reading the ad (e.g., communication, the combined language in our message can serve as
“After reading this ad I feel more informed about cultured meat”, useful proxy for perceived informativeness.
Cronbach’s α = 0.843). Participants also indicated their level of inno­
vativeness on a four-item 7-pt. scale (e.g., “In general, I am among the 4. General discussion
first to buy a new product when it appears”, Cronbach’s α = 0.822) and
whether they, or someone else in their circle, currently consumed The question how to encourage people to act more responsibly or
cultured meat (binary: yes/no). discourage them from irresponsible behavior (in contexts such as health

11
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

Fig. 3a. Message language effect on intentions to consume cultured meat. The vertical axis reflects average ratings of intention to consume cultured meat on a 1–7
scale, following each of the messages. The figure shows higher intentions to consume cultured meat following the combined message.

Fig. 3b. Message language effect on perceived informativeness about cultured meat. The vertical axis reflects average ratings of perceived informativeness on a 1–7
scale, following each of the messages. The figure shows higher perceived informativeness following the combined message.

or the environment) has triggered mounting research efforts. However, an encouraging message, when communicators seek ways to boost
research so far juxtaposed, rather than combined, encouraging and intention to adopt new pro-environmental behavior.
discouraging communication approaches. Conversely, the current work
is inspired by educational psychology and suggests that it is a combi­
4.1. Theoretical implications
nation of an encouraging and a discouraging approach that is especially
valuable for promoting new pro-environmental behaviors that require
We study new pro-environmental behaviors. Such behaviors can be
extra educational efforts. We conducted two field experiments using
associated with adopting environmentally beneficial technologies and
Facebook advertising platform (Studies 1 and 2) and demonstrated that
products that are new to the market (sometimes labeled eco-
a message which combines an encouraging and a discouraging approach
innovations) such as electric cars and cultured meats or require adopt­
is more effective in driving online reactions and engagement, compared
ing new forms of activities to mitigate new environmental threats, such
with an only encouraging or an only discouraging message. In a third
as using public transportation instead of private cars or complying with
field experiment involving actual e-waste recycling behavior (Study 3),
a new environmental regulation. This suggests that our findings may be
we replicated the effect and then followed up with a field survey in the
useful for scholars of interested in pro-environmental communication
same location, which corroborated our findings. Study 4, in a controlled
and people’s response to it (e.g., Gulliver, Fielding, & Louis, 2021; Lee,
online setting showed that the main reason for the advantage of the
2011; White et al., 2019) and also for research on eco-innovation
combined message is that it bears higher educational value in the form
adoption (e.g., Carrilo-Hermosilla et al., 2009).
of higher perceived informativeness, compared with the encouraging or
In this work we focused on four different pro-environmental con­
a discouraging message. We also showed in this study that the effects we
texts: residential hydroponic agriculture, e-waste/paper recycling,
find do not hold in a more established (not new) pro-environmental
geothermal/solar home energy, and consumption of cultured meat.
context. Study 5, also a controlled online study, replicated the effect
While these contexts represent new pro-environmental behavior within
in a different new pro-environmental context (geothermal energy) as
the study locations and culture, they differ in some aspects that may be
part of a full moderated mediation model. Finally, Study 6, replicated
relevant for future research. For instance, residential hydroponic agri­
the mediation of perceived informativeness in another new pro-
culture, consumption of cultured meat or geothermal energy entail a
environmental behavior (consuming cultured meat) and demonstrated
more radical behavioral change, effort and cost on the one hand, yet, for
that the combined message is as good as directly adding information to
people who feel that they lack the specific environmental knowledge,

12
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

the adoption of such behaviors may not be perceived as only ‘environ­ readiness to introduce new behaviors.
mental’. This is especially true, since this behavior also relates to per­ Finally, a valuable theoretical lesson from our findings is that by
sonal health benefits (e.g., reduced amount of pesticides, reduced using a simple communication approach that involves slight changes in
amount of eating fat) or savings costs (via lower energy bills). In addi­ language, we can achieve perceptions of learning, informativeness and
tion, some contexts such as hydroponic agriculture might require some potentially higher knowledge, which may in turn result in higher
type of new skills (growing your own food if you didn’t do gardening engagement with, and adoption of, new pro-environmental behaviors.
before) or some personal taste issues (cultured meat might taste differ­ Purely informational messages are not always effective (Guitart &
ently than planet based protein like Tufu or real meat). Conversely, Stremersch, 2021) and can sometimes consume much space. Therefore,
while collecting and recycling e-waste clearly belongs to an environ­ there is special value in mere linguistic modifications to messages that
mental context, it is less of an innovation. On another level, participating replace informational texts, or education interventions.
in collecting e-waste be more complicated and time-consuming effort
than other recycling activities (carrying the equipment for example and 4.2. Practical implications
separate collection channels for separate types of e-waste). It is also
considered new compared with recycling activities that were already The current work highlights the role of the target audiences’ level of
practiced in the studied country (e.g., paper, glass, packages). Another knowledge as a moderator of the effectiveness of the communication
substantial difference of these contexts is related to real and perceived approach. Moreover, general education level may be a good proxy for
benefits on the one hand and risk (or chances to have limited environ­ knowledge levels in a specific topic such as sustainability (Grinstein &
mental impact or underperform) on the other hand. For example, resi­ Nisan, 2009). Therefore, a relatively pragmatic solution for communi­
dential hydroponic agriculture comes with tangible benefits of having cators may be segmenting the population by education level, as it can be
vegetables grown in your house with no pesticides or feeling positive useful to identify less knowledgeable people as the target audience for
about helping the environment (i.e., ’warm glow’; Sankar, Bhatta­ the combined messaging. From a practical point of view, when policy
charya, & Korschun, 2006). Yet, hydroponic agriculture bears the risk of makers and various organizations are looking to promote new
failure – one might fail if not following the exact instructions or meet pro-environmental behavior, our study can aid in designing campaigns
other conditions, whereas e-waste might be limited on the ‘warm glow’ that are both more effective and low-cost. More specifically, the
aspect but less risky. Further, although recycling e-waste is free in many campaign should include suggestion for a combined set of actions.
countries (for example in Europe15 and NY State16), people may perceive However, it might be especially important to provide a short facts
the process as costly, as they might believe that their electronic items, introduction on the topic in target before the messages itself. In our
even malfunctioning ones, have some value (Widmer, Oswald-Krapf, study we demonstrated the approach on multiple pro-environmental
Sinha-Khetriwal, Schnekkmann, & Böni, 2005). In this work, we tested contexts.
messages that seem orthogonal to the influences of perceived benefits Our results from Google Analytics suggest interesting implications
and risk, but we also did not directly compare the outcomes of the for practical use of analytical tools in digital communication. Learning
different campaigns. from field experiments conducted on social platforms and digital
Since we found relatively low level of knowledge about the new pro- advertising, policy makers, NGOs, and marketers need to carefully
environmental contexts studied, our messaging approach was based on choose how to analyze the raw results from online platform analytics
educational theory. This approach is most useful among people who tools and use a multi-method approach to corroborate and explain these
actually need it – people who feel that they lack knowledge. Accord­ results to improve insights and future predictions. For example, our
ingly, it is possible that among people who are more knowledgeable results were significant for many outcome metrics, but for some the
about the topic at hand, the combined messaging will be less effective differences did not reach significance. A careful analysis of an organi­
because these people may already be aware of both the problem and zation’s repeating successes of failure in eliciting particular online be­
possible solutions. These populations may consider the information haviors can shed light on its relationships with the audience and the best
offered in the combined message not new to them, and therefore this ways for it to encourage pro-environmental behaviors.
type of message will not motivate them to action. In sum, the audiences’
level of knowledge may be a moderator of the effectiveness of the 4.3. Limitations and future research
educative approach.
Other factors can explain the effects of combined messages on While our research advances the literature, it is not without limita­
behavior. For example, as the two-sided messaging literature suggests tions, which serve as opportunities for future research. First, as
(e.g., Rucker et al., 2008), these messages increase the link between mentioned earlier, in this research we tested messages that may be
attitudes and behavior. It is possible that a combined message can be orthogonal to the influences of perceived benefits and risks, or gains and
viewed as two-sided and result in higher attitude-behavior link because losses. Future research may focus on the relationship between perceived
it sounds more credible and interest-free. Additionally, we did not benefits and risks and reactions to encouraging and discouraging mes­
explore emotional responses to prohibition (discouraging) and permis­ sages as well as combined messages.
sion (encouraging). Classic research suggests that when feal appeal Second, framing this research within Regulatory Focus theory and its
messages are coupled with instructions (Leventhal, Watts, & Pagano, relation to language (e.g., Semin, Higgins, Gil de Montes, Estourget, &
1967), this may be an effective way to elicit compliance. Valencia, 2005), it may seem that encouraging messages elicit a pro­
Although our study was conducted before the onset of the COVID-19 motion mechanism and discouraging messages evoke prevention. This
pandemic, it relates to recent research following the pandemic. Specif­ may be so, but interestingly enough, in our research, the combined
ically, our results highlight the benefits of communication that provides message – that may have theoretical elements of both promotion and
perceptions of being better informed and having an outlet for action. prevention – is more effective than each of the mechanisms in isolation.
These communicational goals echo current research related to COVID19 Future investigation of the potentially evoked regulatory focus para­
and pro-environmental behavior, such as Tchetchik et al. (2020), who digm as an effect of our combined messages may shed light on additional
showed that people are more ready to increase recycling activities post explanations for its effects, beyond perceived informativeness.
COVID because of increased operational capacity for change and Third, we acknowledge that our field experiments, although offering
valuable external validity, may suffer from potential “noises”. For
example, in Study 3 we were able to get access to rich data for only some
15
https://ec.europa.eu/eurostat/web/waste/key-waste-streams/weee. of the studied areas (also limiting the ability to conduct rigorous
16
http://www.dec.ny.gov/chemical/65583.html. regression analysis) and in Studies 1 and 2 it was not feasible to rule out

13
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

the possibility that a small number of participants might have been Drukker, D. M. (2003). Testing for serial correlation in linear panel-data models. STATA
Journal, 3, 168–177.
exposed to multiple messages. We attempted to address these short­
Echegaray, F., & Hansstein, F. V. (2017). Assessing the intention-behavior gap in
comings via multiple robustness checks and analytical approaches. electronic waste recycling: The case of Brazil. Journal of Cleaner Production, 142,
Finally, future research may expand the results from our studies to 180–190.
other countries or cultures, to additional new pro-environmental be­ EIO. (2010). Future trends: Creating the environment for eco-innovation evolution accessed
on https://www.eco-innovation.eu/index.php/reports/h
haviors, to campaigns through other digital and off-line channels, and orizon-scanning-and-foresight-reports - last. (Accessed 5 March 2021).
employing other metrics. EIO. (2012). Eco-innovation observatory. Methodological Report. Available at:
https://www.eco-innovation.eu/index.php/reports/methodological-report.
Eriksson, N., Sjöberg, A., Rosenbröijer, C. J., & Fagerstrøm, A. (2019). Consumer brand
5. Conclusion post engagement on Facebook and Instagram–A study of three interior design
brands. In The 19th international conference on electronic business (Newcastle upon
The current work aimed to address a growing challenge: Promoting Tyne, UK).
European Commission. (2013). CIP eco-innovation first application and market replication
new pro-environmental behaviors. We relied on an educational projects [available at: http://ec.europa.eu/environment/eco-innovation/files/docs/
approach and tested the effectiveness of a combined messages that in­ faq/faq-call-2013.pdf. (Accessed 12 June 2019).
duces perceptions of informativeness, thereby helping overcome one of Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G* power 3: A flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavior
the main barriers to adoption of new pro-environmental behaviors – Research Methods, 39(2), Article 175e191.
perceived lack of knowledge. Our results serve as an initial step towards Feng, B., & MacGeorge, E. L. (2010). The influences of message and source factors on
understanding the ways to promote pro-environmental behaviors. advice outcomes. Communication Research, 37(4), 553–575.
Gamliel, E., & Herstein, R. (2011). To save or to lose: Does framing price promotion
affect consumers’ purchase intentions? Journal of Consumer Marketing, 28(2),
Author statement 152–158.
Garcia, R., & Calantone, R. (2002). A critical look at technological innovation typology
Ann Kronrod: Conceptualization; Data curation; Formal analysis; and innovativeness terminology: A literature review. Journal of Product Innovation
Management, 19, 110–132.
Investigation; Methodology; Project administration; Resources; Super­ Gastil, J. (1993). Democracy in small groups: Participation, decision making &
vision; Validation; Visualization; Roles/Writing - original draft; Writing communication. New Society Publishers.
- review & editing. Gifford, R. (2011). The dragons of inaction: Psychological barriers that limit climate
change mitigation and adaptation. American Psychologist, 66(4), 290.
Anat Tchetchik: Data curation; Formal analysis; Investigation; Google. (2019). Google analytics [available at: https://marketingplatform.google.com/a
Methodology; Resources; Software; Validation; Visualization; Writing - bout/analytics/?hl=en_US. (Accessed 19 January 2019).
review & editing. Gordon, B. R., Zettelmeyer, F., Bhargava, N., & Chapsky, D. (2019). A comparison of
approaches to advertising measurement: Evidence from big field experiments at
Amir Grinstein: Conceptualization; Data curation; Formal analysis; Facebook. Marketing Science, 38(2), 193–225.
Investigation; Methodology; Project administration; Resources; Valida­ Graham, A., Milner, P., Saul, J., & Pfaff, L. (2008). Online advertising as a public health
tion; Roles/Writing - original draft; Writing - review & editing. and recruitment tool: Comparison of different media campaigns to increase demand
for smoking cessation interventions. Journal of Medical Internet Research, 10(5), e50.
Lior Turgeman: Data curation; Formal analysis; Methodology. Grinstein, A., & Nisan, U. (2009). Demarketing, minorities, and national attachment.
Vered Blass: Funding acquisition; Conceptualization; Data collection; Journal of Marketing, 73(2), 105–122.
Supervision; Investigation; Resources; Writing - review & editing. Guitart, I. A., & Stremersch, S. (2021). The impact of informational and emotional
television ad content on online search and sales. Journal of Marketing Research, 58(2),
299–320.
Appendix A. Supplementary data Gulliver, R., Fielding, K. S., & Louis, W. R. (2021). Assessing the mobilization potential of
environmental advocacy communication. Journal of Environmental Psychology, 74,
Supplementary data to this article can be found online at https://doi. Article 101563.
Hadar, L., Sood, S., & Fox, C. R. (2013). Subjective knowledge in consumer financial
org/10.1016/j.jenvp.2022.101945. decisions. Journal of Marketing Research, 30, 303–316.
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis:
References A regression-based approach. New York, NY: Guilford Press.
Helmig, B., & Thaler, J. (2010). On the effectiveness of social marketing—what do we
really know? Journal of Nonprofit & Public Sector Marketing, 22(4), 264–287.
Abdelbasir, S. M., Hassan, S. S., Kamel, A. H., & El-Nasr, R. S. (2018). Status of electronic Hitchon, J. C., & Thorson, E. (1995). Effects of emotion and product involvement on the
waste recycling techniques: A review. Environmental Science and Pollution Research, experience of repeated commercial viewing. Journal of Broadcasting & Electronic
25(17), 16533–16547. Media, 39(3), 376–389.
Adler, A. (1998). Understanding human nature. Minnesota: Hazleden. Hoch, S. J., & Ha, Y.-W. (1986). Consumer learning: Advertising and the ambiguity of
Baldé, K., Forti, V., Gray, V., Kuehr, R., & Stegmann, P. (2017). The global E-waste monitor product experience. Journal of Consumer Research, 13(2), 221–233.
– 2017. Bonn/Geneva/Vienna: United Nations University (UNU), International Jha, A. (2013). First lab-grown hamburger gets full marks for ’mouth feel. The Guardian
Telecommunication Union (ITU) & International Solid Waste Association (ISWA) [available at: https://www.theguardian.com/science/2013/aug/05/world-first-s
[available at: http://collections.unu.edu/eserv/UNU:6341/Global-E-waste_Monitor_ ynthetic-hamburger-mouth-feel. (Accessed 23 October 2018).
2017__electronic_single_pages_.pdf. (Accessed 21 May 2019). Johnstone, M.-L., & Tan, L. P. (2015). Exploring the gap between consumers’ green
Beise, M., & Rennings, K. (2003). Lead markets of environmental innovations: A rhetoric and purchasing behavior. Journal of Business Ethics, 132(2), 311–328.
framework for innovation and environmental economics. Discussion Papers, 52(1), Kamins, M. A., & Assael, H. (1987). Two-sided versus one-sided appeals: A cognitive
5–17. perspective on argumentation, source derogation, and the effect of disconfirming
Bockarjova, M., & Steg, L. (2014). Can protection motivation theory predict pro- trial on belief change. Journal of Marketing Research, 24(1), 29–39.
environmental behavior? Explaining the adoption of electric vehicles in The Keller, P. A., & Lehmann, D. R. (2008). Designing effective health communications: A
Netherlands. Global Environmental Change, 28, 276–288. meta-analysis. Journal of Public Policy and Marketing, 27(2), 117–130.
Buller, D. B., Pagoto, S., Henry, K. L., Baker, K., Walkosz, B. J., Joel Hillhouse, et al. Kelly, R. J. (1982). Behavioral reorientation of pedophiliacs: Can it Be done? Clinical
(2022). Effects of engagement with a social media campaign for mothers to prevent Psychology Review, 2(3), 387–408.
indoor tanning by teens in a randomized trial. Journal of Health Communication, Kvols, K. J. (1998). Redirecting children’s behavior. Parenting Press.
(August), 1–13. Lacroix, K., & Gifford, R. (2018). Psychological barriers to energy conservation behavior:
Byrka, K., & Kaminska, K. (2022). Doing laundry with biodegradable soap nuts: Can rare The role of worldviews and climate change risk perception. Environment and
and novel behaviors break bad habitual patterns? Journal of Environmental Behavior, 50(7), 749–780.
Psychology, 79, Article 101730. Landry, N., Gifford, R., Milfont, T. L., Weeks, A., & Arnocky, S. (2018). Learned
Carrilo-Hermosilla, J., Del Río González, P., & Könnölä, T. (2009). Eco-innovation. When helplessness moderates the relationship between environmental concern and
sustainability and competitiveness shake hands. Palgrave Macmillan. behavior. Journal of Environmental Psychology, 55(2), 18–22.
Cheng, M. J., Hung, S. W., Tsai, H. H., & Chou, Y. C. (2020). Fostering environmentally Lane, B., & Potter, S. (2007). The adoption of cleaner vehicles in the UK: Exploring the
responsible consumer behavior: A hierarchical approach toward smartphone consumer attitude–action gap. Journal of Cleaner Production, 15(11–12), 1085–1092.
recycling. IEEE Transactions on Engineering Management, 69(5), 2326–2336. Lee, K. (2011). The role of media exposure, social exposure and biospheric value
Dietz, T., Ostrom, E., & Stern, P. C. (2003). The struggle to govern the commons. Science, orientation in the environmental attitude-intention-behavior model in adolescents.
302(5652), 1907–1912. Journal of Environmental Psychology, 31(41), 301–308.
Dreikurs, R., & Soltz, V. (1991). Children: The challenge. New York: Plume: Penguin Leventhal, H., Watts, J. C., & Pagano, F. (1967). Effects of fear and instructions on how to
Group. cope with danger. Journal of Personality and Social Psychology, 6(3), 313–321.

14
A. Kronrod et al. Journal of Environmental Psychology 86 (2023) 101945

Lohse, B. (2013). Facebook is an effective strategy to recruit low-income women to Shamai, H. (2017). Electronic waste in Israel. Adam Teva Vadin [available at: http://www.
online nutrition education. Journal of Nutrition Education and Behavior, 45(1), 69–76. adamteva.org.il/_Uploads/dbsAttachedFiles/EWASTEHEB.pdf. (Accessed 21 May
Marr, B. (2019). The future of food: Amazing lab grown and 3D printed meat and fish. 2019).
Forbes. Jun 28, 2019. Available at: https://www.forbes. Sierzchula, W., Bakker, S., Maat, K., & Van Wee, B. (2014). The influence of financial
com/sites/bernardmarr/2019/06/28/th incentives and other socio-economic factors on electric vehicle adoption. Energy
e-future-of-food-amazing-lab-grown-and-3d-printed-meat-and-fish/?sh Policy, 68, 183–194.
=30bc6c8446f6 last. (Accessed 19 February 2021). Snyder, L. B., Hamilton, M. A., Mitchell, E. W., Kiwanuka-Tondo, J., Fleming-Milici, F., &
Mazzanti, M. (2018). Eco-innovation and sustainability: Dynamic trends, geography and Proctor, D. (2004). A meta-analysis of the effect of mediated health communication
policies. Journal of Environmental Planning and Management, 61(11), 1851–1860. campaigns on behavior change in the United States. Journal of Health Communication,
McGregor, S. (2005). Sustainable consumer empowerment through critical consumer 9(S1), 71–96.
education: A typology of consumer education approaches. International Journal of Sood, A., & Kumar, V. (2017). Analyzing client profitability across diffusion segments for
Consumer Studies, 29(5), 437–447. a continuous innovation. Journal of Marketing Research, 54(6), 932–951.
Menon, R. V., Sigurdsson, V., Larsen, N. M., Fagerstrøm, A., Sørensen, H., Statista. (2020). Number of monthly active Facebook users worldwide as of 2nd quarter 2020
Marteinsdottir, H. G., et al. (2019). How to grow brand post engagement on [available at: https://www.statista.com/statistics/264810/number-of-monthly-act
Facebook and twitter for airlines? An empirical investigation of design and content ive-facebook-users-worldwide/.
factors. Journal of Air Transport Management, 79, Article 101678. Steiner, R. (2015). Consumer preferences towards electronic waste recycling programs using
Mochon, D., Johnson, K., Schwartz, J., & Ariely, D. (2017). What are likes worth? A advanced discrete choice modeling. Thesis submitted to the porter school of environmental
Facebook page field experiment. Journal of Marketing Research, 54(2), 306–317. studies and earth sciences towards masters degree in environmental studies. Israel: Tel
Möhlman, M. (2015). Collaborative consumption: Determinants of satisfaction and the Aviv University.
likelihood of using a sharing economy option again. Journal of Consumer Behavior, 14 Stivers, T. (2005). Non-antibiotic treatment recommendations: Delivery formats and
(3), 193–207. implications for parent resistance. Social Science & Medicine, 60(5), 949–964.
Monbiot, G. (2020). Lab-grown food will soon destroy farming – and save the planet. Jan 8 Stivers, T. (2006). The interactional process of reaching a treatment decision in acute
2020. The Guardian (Available at: https://www.theguardian. medical encounters. Communication in Medical Care: Interactions between Primary Care
com/commentisfree/2020/jan/08/lab-grown-food-destroy-farming-save-planet last. Physicians and Patients, 279–312.
(Accessed 19 February 2021). Strahan, E. J., White, K., Fong, G. T., Fabrigar, L. R., Zanna, M. P., & Cameron, R. (2002).
Morren, M., & Grinstein, A. (2021). The cross-cultural challenges of integrating personal Enhancing the effectiveness of tobacco package warning labels: A social
norms into the theory of planned behavior: A meta-analytic structural equation psychological perspective. Tobacco Control, 11(3), 183–190.
modeling (masem) approach. Journal of Environmental Psychology, 75, Article Tchetchik, A., Kaplan, S., & Blass, V. (2021). Recycling and consumption reduction
101593. following the COVID-19 lockdown: The effect of threat and coping appraisal, past
Mullen, P. D., Simons-Morton, D. G., Ramírez, G., Frankowski, R. F., Green, L. W., & behavior and iInformation. Resource Conservation & Recycling Journal, 167, Article
Mains, D. A. (1997). A meta-analysis of trials evaluating patient education and 105370.
counseling for three groups of preventive health behaviors. Patient Education and Tchetchik, A., Zvi, L. I., Kaplan, S., & Blass, V. (2020). The joint effect of driving
Counseling, 32, 157–173. hedonism and ‘trialability’ on the choice between internal combustion engines,
Murphy, T. F. (1992). Redirecting sexual orientation: Techniques and justifications. The hybrid and electric vehicles: The hidden segment of environmentalist-hedonists.
Journal of Sex Research, 29(4), 501–523. Elsevier Technological Forecasting and Social Change, 151, Article 11981.
Niemiec, R., Jones, M. S., Lischka, S., & Champine, V. (2021). Efficacy-based and Ter Mors, E., Weenig, M. W., Ellemers, N., & Daamen, D. D. (2010). Effective
normative interventions for facilitating the diffusion of conservation behavior communication about complex environmental issues: Perceived quality of
through social networks. Conservation Biology, 35(4), 1073–1085. information about carbon dioxide capture and storage (CCS) depends on stakeholder
Nygrén, N. A., Kontio, P., Lyytimäki, J., Varho, V., & Tapio, P. (2015). Early adopters collaboration. Journal of Environmental Psychology, 30(4), 347–357.
boosting the diffusion of sustainable small-scale energy solutions. Renewable and Van Der Wal, A. J., Van Horen, F., & Grinstein, A. (2018). Temporal myopia in
Sustainable Energy Reviews, 46, 79–87. sustainable behavior under uncertainty. International Journal of Research in
Olsen, M. C., Slotegraaf, R. J., & Chandukala, S. R. (2014). Green claims and message Marketing, 35(3), 378–393.
frames: How green new products change brand attitude. Journal of Marketing, 78(5), Varadarajan, R. (2017). Innovating for sustainability: A framework for sustainable
119–137. innovations and a model of sustainable innovations orientation. Journal of the
Oryan, S., & Gastil, J. (2013). Democratic parenting: Paradoxical messages in democratic Academy of Marketing Science, 45(1), 14–36.
parent education theories. International Review of Education, 59(1), 113–129. Vossenaar, M., Solomons, N. W., Valdés-Ramos, R., & Anderson, A. S. (2011). Agreement
Osbaldiston, R., & Schott, J. P. (2012). Environmental sustainability and behavioral between dietary and lifestyle guidelines for cancer prevention in population samples
acience: Meta-analysis of proenvironmental behavior experiments. Environment and of Europeans and mesoamericans. Nutrition, 27(11–12), 1146–1155.
Behavior, 44(2), 257–299. Wagner, D. E., Fernandez, P., Jordan, J. W., & Saggese, D. J. (2019). Freedom from chew:
Ozaki, R. (2011). Adopting sustainable innovation: What makes consumers sign up to Using social branding to reduce chewing tobacco use among country peer crowd
green electricity? Business Strategy and the Environment, 20(1), 1–17. teens. Health Education & Behavior, 46(2), 286–294.
Passyn, K. A., & Sujan, M. (2006). Self-accountability emotions and fear appeals: Walker, D., Gutierrez, J. P., Torres, P., & Bertozzi, S. M. (2006). HIV prevention in
Motivating behavior. Journal of Consumer Research, 32, 582–589. Mexican schools: Prospective randomised evaluation of intervention. BMJ, 332
Pérez-Belis, V., Bovea, M. D., & Simó, A. (2015). Consumer behaviour and environmental (7551), 1189–1194.
education in the field of waste electrical and electronic toys: A Spanish case study. Wang, Z., Guo, D., Wang, X., Zhang, B., & Wang, B.( (2018). How does information
Waste Management, 36(2014), 277–288. publicity influence residents’ behaviour intentions around E-waste recycling?
Rucker, D. D., Petty, R. E., & Briñol, P. (2008). What’s in a frame anyway?: A meta- Resources, Conservation and Recycling, 133(1), 1–9.
cognitive analysis of the impact of one versus two sided message framing on attitude White, K., Habib, R., & Hardisty, D. J. (2019). How to shift consumer behaviors to be
certainty. Journal of Consumer Psychology, 18(2), 137–149. more sustainable: A literature review and guiding framework. Journal of Marketing,
Sankar, S., Bhattacharya, C., & Korschun, D. (2006). The role of corporate social 83(3), 22–49.
responsibility in strengthening multiple stakeholder relationships: A field White, K., & Simpson, B. (2013). When do (and don’t) normative appeals influence
experiment. Journal of the Academy of Marketing Science, 34(2), 158–166. sustainable consumer behaviors? Journal of Marketing, 77(2), 78–95.
Saphores, M. J.-D., Oladele, O. A., & Shapiro, A. A. (2012). Willingness to engage in a Widmer, R., Oswald-Krapf, H., Sinha-Khetriwal, D., Schnekkmann, M., & Böni, H. (2005).
pro-environmental behavior: An analysis of e-waste recycling based on a national Global perspectives on E-waste. Environmental Impact Assessment Review, 25(5),
survey of U.S. households. Resources, Conservation and Recycling, 60, 49–63. 436–458.
Schultz, P. W., & Kaiser, F. G. (2012). Promoting pro-environmental behavior. In Yoon, G., Li, C., Ji, Y., North, M., Hong, C., & Liu, J. (2018). Attracting comments: Digital
S. D. Clayton (Ed.), The Oxford handbook of environmental and conservation psychology engagement metrics on Facebook and financial performance. Journal of Advertising,
(pp. 556–580). Oxford University Press. 47(1), 24–37.
Schwartz, J., Luce, M. F., & Ariely, D. (2011). Are consumers too trusting? The effects of Zhang, H., Liang, X., & Wang, S. (2016). Customer value anticipation, product
relationships with expert advisers. Journal of Marketing Research, 48(SPL), innovativeness, and customer lifetime value: The moderating role of advertising
S163–S174. strategy. Journal of Business Research, 69(9), 3725–3730.
Semin, G. R., Higgins, T., Gil de Montes, L., Estourget, Y., & Valencia, J. F. (2005).
Linguistic signatures of regulatory focus: How abstraction fits promotion more than
prevention. Journal of Personality and Social Psychology, 89(1), 36–47.

15

You might also like