Journal of Environmental Economics and Management 76 (2016) 1–13
Contents lists available at ScienceDirect
Journal of Environmental Economics and Management journal homepage: www.elsevier.com/locate/jeem
Can indifference make the world greener? Johan Egebark a, Mathias Ekström a,b,n a b
Research Institute of Industrial Economics (IFN), Stockholm, Sweden Department of Economics, Norwegian School of Economics (NHH), Helleveien 30, N-5045 Bergen, Norway
a r t i c l e i n f o
abstract
Article history: Received 24 February 2015 Available online 26 November 2015
We conducted a natural field experiment to evaluate two resource conservation programs. One intervention consisted of a moral appeal message asking university employees to cut back on printing in general, and to use double-sided printing whenever possible. The other intervention tested whether people's tendency to stick with pre-set alternatives is applicable to resource use: at random points in time we changed the default setting on the university printers, from single-sided to double-sided printing. Whereas the moral appeal had no impact, the default change cut paper use by 15 percent. Further analysis adds two important insights. First, we show that defaults influence behavior also in the longer run. Second, we present results indicating that resource efficient defaults have the advantage of avoiding unintended behavioral responses. Overall, our findings send a clear message to anyone concerned about resource conservation: there are potentially large gains to be made from small interventions. & 2015 Elsevier Inc. All rights reserved.
JEL classification: C93 D03 Q50 Keywords: Resource conservation Default option Moral appeal Natural field experiment
Introduction Depletion of natural resources, such as water, food and forests, constitutes a severe threat to the global environment, and to sustained economic growth (United Nations, 1992; European Union, 2011). In order to address this threat policy makers need to understand what types of actions will have an impact on resource conservation, and what types of actions are ineffective. In recent years, non-price based behavioral interventions that preserve choice — the so-called nudges — have gained increasing attention, both from academic scholars and among politicians (Thaler and Sunstein, 2008; Shafir, 2012). It has been argued that nudges have a particularly important role to play within the environmental domain, not least since regular price adjustments have been difficult to implement (Allcott and Mullainathan, 2010; Carlsson and JohanssonStenman, 2012; Sunstein and Reisch, 2014). However, empirical evidence is still scarce, and to the extent that it does exist the results are mixed.1 In this paper, we provide new evidence by studying the causal effect of two different behavioral interventions aimed to lower the consumption of paper. The activity that we consider is universal, frequent, and consumes a vast amount of resources every year: document printing. Estimates suggest that U.S. office workers use roughly five million metric tons of paper annually, amounting to around 20 million metric tons of wood. To illustrate the potential, reducing this amount by n
Corresponding author. Fax: þ 47 55959100. E-mail address:
[email protected] (M. Ekström). 1 Allcott and Rogers (2014) show that when using feedback to reduce energy consumption it is crucial to repeat the treatment, since otherwise the effect decays quickly. Furthermore, Kallbekken et al. (2013) find that disclosing information on lifetime energy costs can affect purchases of durable goods, but only for certain product categories and when the information is coupled with training of the sales staff. http://dx.doi.org/10.1016/j.jeem.2015.11.004 0095-0696/& 2015 Elsevier Inc. All rights reserved.
2
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
only five percent would save roughly six and a half million trees, free 6500 acres (5000 football fields) of forest for other productive use, and prevent the annual greenhouse gas emissions equivalent to that of 140,000 cars.2 When sending a document to a printer, a user can typically choose whether to print on both sides of a sheet of paper (duplex) or to print on only one side (simplex). Duplex printing reduces the number of sheets in the production of a document and is thus less resource intensive. We use this functionality in a field experiment at a large Swedish university to evaluate two different paper saving programs. One of the interventions tests whether moral appeal increases environmental responsibility. We designed an e-mail campaign, in collaboration with the university, which encouraged employees to cut back on paper use in general and to use duplex printing whenever possible. The other intervention examines the importance of how alternatives are presented. A common feature of most modern printers is the existence of a default option — a pre-set alternative that people obtain if they do not actively make another choice. By randomly changing the default, from simplex to duplex printing, we test whether people's tendency to stick to the pre-set alternative can help save resources. There are several motives for studying these two interventions. Seminal (observational) studies have documented substantial default effects on retirement savings (Madrian and Shea, 2001) and on organ donor registration (Johnson and Goldstein, 2003). However, it is still an open question whether, and to what extent, these effects carry over to the environmental domain. Löfgren et al. (2012) is, to the best of our knowledge, the only published paper that experimentally tests if default rules influence people to behave pro-environmentally. The authors find that the pre-set alternative is completely irrelevant when participants at a conference in environmental and resource economics are asked whether they want to offset travel-related CO2-emissions or not.3 Researchers are currently acquiring a better understanding of the limitations of defaults (see e.g., Beshears et al., 2010), and there could be many potential factors behind this specific null result. There are, however, reasons to expect attenuated default effects when it comes to environmental issues in general and resource conservation in particular. One reason is the wide-spread public debate concerning the environmental threats the world is currently facing. Environmental awareness is likely to be high, and people may have strong motives for their current behavior. It could also be that the warm-glow that comes with behaving pro-environmentally is enough to move borderline decision makers — who are likely to drive default effects — from passively accepting default options to making an active choice. Notably, strong intentions have been shown to reduce the default effect (Bronchetti et al., 2013). Another reason to suspect smaller effects of mandated “green” default options, especially in settings like ours, is that people and businesses in charge of setting default options already face economic incentives to consume less of the resource. Thus, economic agents who have actually chosen a more wasteful default option presumably have good reasons to do so, and their underlying behavior cannot be expected to be easily altered by a change in the default option. There is, however, a large literature on what has been referred to as “the energy paradox” and “the energy-efficiency gap”, which suggests that people are not taking full advantage of improvements in energy efficient technology (Jaffe and Stavins, 1994). If this assertion is correct, changes in default options could potentially affect resource conservation. In addition to these concerns, it might also be that people adapt their behavior in response to resource efficient defaults by using more of the resource (Khazzoom, 1980). A relevant example capturing this kind of perverse behavior is Catlin and Wang (2013), who report an increase in paper use when the possibility to recycle is introduced. Currently, there is limited knowledge about the long-run impacts and unintended consequences of using green defaults, which is a serious concern given that default options (and other behavioral interventions) are increasingly being seen as valuable complements to traditional policy tools (Sunstein and Reisch, 2014). Our default intervention will provide coveted insights from a natural setting and thus shed light on the validity of the points outlined above. The moral appeal treatment is included in the study for two main purposes. First, it provides a natural baseline: it is the intervention first chosen by the university, and similar messages are commonly used in many everyday settings. Examples include appeals to hotel guests to reuse towels during their stay, and reminders about waste disposal in public parks. The common use of moral appeal messages suggests that policy makers perceive them as effective (or that they simply lack better alternatives). The second, and more substantive, reason for including the moral appeal treatment is that it can help us address some of the drivers behind a potential default effect. For example, it is often proposed that people follow the default option because it provides information about how to behave (McKenzie et al., 2006; Altmann et al., 2013). If this is the case in our setting, giving explicit information about how employees are expected to print should be enough in itself. The effect of the two interventions differed sharply. Using moral appeal to encourage people to act responsibly had no impact on paper use, not even the day the message was communicated. In contrast, we document a substantial and immediate effect of changing the default option. Daily paper consumption dropped by 15 percent on average, implying that the default alternative determined the layout of one third of all printed documents. Two other findings are important in light of the increasing interest in behavioral interventions. First, the default effect was equally large six months after the implementation. Second, the absence of unintended adverse behavior in our experiment suggests that it is possible to introduce resource efficient defaults without the risk of backfiring. Overall, the results suggest that people sticked to the default option to avoid thoughtful considerations and a tiny switching cost, which showcases the importance of carefully 2
The figures and the estimated impact come from the Environmental Paper Network (www.papercalculator.org). There are results from unincentivized hypothetical choice experiments showing that people are more prone to choose green energy (Pichert and Katsikopoulos, 2008) and compact fluorescent light bulbs (Dinner et al., 2011) if these options are presented as the default alternative. 3
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
3
choosing “no-action” options and that passively taking advantage of people's (lack of) preferences can be much more effective than actively trying to change them (see also Chetty et al., 2014). The rest of the paper proceeds as follows. The section “The experiment” carefully describes the interventions, the experimental design, the empirical strategy, and an ex-post survey in separate subsections. The section “Results” outlines the results whereas the section “Concluding discussion” provides conclusions and potential policy implications based on our findings.
The experiment The interventions We conducted a natural field experiment at a large public university in Sweden to measure the effect of two paper savings programs. The main intervention changes the default option on printers, from simplex to duplex printing. Duplex printing makes use of both sides of a sheet of a paper and is thus more resource efficient. The default is not binding since everyone is free to choose their preferred layout before sending the document to the printer. The cost of being active is close to zero as it constitutes the time spent navigating a pop-up window and executing one mouse click; something which is unlikely to take more than a few seconds irrespective of experience.4 Thus, we can rule out costly switching as an important driver in the studied setting. As pointed out by Thaler and Sunstein (2008), nudges have great potential in situations when people are relying on quick and intuitive judgment. As they see it, nudging is a tool that align people's actions with their own long-term interests in cases when they use automatic reasoning. Kahneman (2011) states several reasons why fast thinking is often justified, and provides examples of how a person faced with a decision problem can save effort by using heuristics and mental short cuts. A recipe for effective nudging is therefore to first identify the heuristic that people use and then alter the outcome of that particular heuristic (another alternative is to try to change the heuristic people use). Tasks that are relatively easy, performed frequently, and that does not render catastrophic consequences in case of failure (such as choosing course from a lunch menu) are typically well suited for the intuitive system. Since document printing is something most office workers do on a regular basis, it is reasonable that people in our setting rely heavily on automatic thinking. Hence, if people are fast thinkers — and use the default alternative to ease decision making — the small change in the decision environment could potentially have large effects on paper consumption. Notably, acting fast without deeper processing sometimes leads to suboptimal outcomes, at least in the short run. The long term success of a nudge is therefore linked to how well it is received, and ultimately whether people adapt their decision-making process or not. We look into this aspect by examining both the short and long run effects of the default intervention. It is sometimes argued that introducing a default alters people's beliefs about the alternatives in the choice set. A pre-set option may be interpreted as advice or as an implicit norm about how to behave, and hence make this choice more attractive (McKenzie et al., 2006; Altmann et al., 2013). To gain insights about the importance of this explanation we added a second intervention to the study. The intervention consisted of a message that tried to convince employees at the university to reduce their use of paper. The moral appeal was communicated via e-mail and was signed by the university's environmental coordinator. We deliberately had a person with some authority send the e-mail as previous evidence suggests that source credibility is important when using persuasive communication (DellaVigna and Gentzkow, 2010). The wording of the e-mail encouraged people to participate in the strive to reduce the university's impact on the environment, by cutting back on printing in general and by using duplex printing whenever possible. The employees were also reminded of how easy it is to use the duplex printing mode (an English version of the e-mail is included in the appendix). The impact of the moral appeal treatment is interesting to study in itself. These types of messages are used in many everyday settings, for example in hotels where guests are asked to reuse towels during their stay. However, the primary reason for including the appeal is that it aids the interpretation of the main intervention. It is not obvious that people are familiar with the green alternative, i.e., the possibility to use duplex on the printer. Moreover, changing the default option to duplex might influence people because it alters their expectations about what is the right thing to do. The message intervention jointly addresses these two explanations for a potential default effect: it explicitly informs users about the duplex alternative, and it states that they are expected to use duplex printing. The purpose of the moral appeal treatment was not to provide a deeper understanding of what type of messages is more effective than others (for readers interested in this topic we refer to Schultz et al., 2007; Goldstein et al., 2008; Fellner et al., 2013; Pruckner and Sausgruber, 2013; Dwenger et al., 2014).5 We do note, however, that Fellner et al. (2013) report that a one-shot letter can have a sizable impact on public good compliance irrespective of the content of the letter. 4
The print screens of the pop-up window included in the appendix show how easy it is to opt in or out of the default option. The use of an injunctive norm (information about the “correct” way to behave) as opposed to a descriptive norm (information about how other people behave) was, in this respect, an active design choice. 5
4
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
Fig. 1. Experimental design.
Design and procedures To implement the default change intervention we asked the IT-support at the university for a list of all the printers that reported sufficient statistics online. This original sample consists of 54 printers distributed across 31 different departments. 19 of the departments had at least one printer that allowed for a change in the default settings from simplex to duplex. We contacted the heads of these 19 departments directly, asking whether they would be willing to participate in the study. The heads were informed about the intention of the experiment and that participation required the default change to be communicated to staff from within the department. (The e-mail employees received about the change in the settings is found in the appendix.) We stressed the importance of not informing any of the employees about the experiment, as this could ruin the benefits of conducting a natural field experiment (Harrison and List, 2004). 18 out of 19 departments accepted these terms and became part of the study. We interpret the high compliance rate as an illustration of the so-called energyefficiency gap. It appears as if the departments had not chosen their default settings in a rational and informed way, but rather stuck to the default which was set when the printer was first installed. The 18 departments had a total of 25 printers which we used in the experiment. Fig. 1 illustrates the experimental design. The experiment consisted of three phases spread evenly over 15 weeks. We started with a pre-treatment phase of five weeks when all printers had simplex as default, and we ended with a posttreatment phase of five weeks when all printers had duplex as default.6 Between these, there was a treatment phase of five weeks, in which the 18 departments were randomly assigned to one of six different pre-determined intervention dates. In the case that a department had more than one printer, we included all their printers in the experiment and treated all the printers at the same time. This feature of the design ensures that employees could not simply switch to another printer that still had simplex as default, and avoided the need to repeat the whole procedure within a department (which could have been perceived as strange). A random subset of the departments was exposed to the moral appeal treatment, prior to their change in default. The message went out once to all employees (300 þ) in these departments, and was sent on different days in the first week of the treatment phase, allowing for a four-week evaluation period. Identification strategy and data The staggered implementation of changed print settings means that our experiment constitutes a controlled event study: all printers are used for the default treatment, but the timing of the intervention is random. Given the design of the experiment, we measure the default effect by using all 25 printers, normalize time to zero, and compare mean outcomes before and after the intervention. Identification is more credible if there is a clear and visible shift in the outcome that appears close in time to the intervention. Since treatment occurs at different points in time, depending on the printer, a 6
All printers used simplex as default before the experiment started, and had been using it for as long as anyone could remember.
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
5
Table 1 Pre-treatment daily averages. Full sample
No message
Message
SP ratio Sheets Pages Duplex sheets Simplex sheets
0.85 170 221 51 119
0.87 144 180 37 107
0.83 204 274 69 135
N (Printers Days) Printers Departments
585 25 18
337 14 12
248 11 6
Moral appeal treatment
1
1
0.9
0.9 Sheets per page
Sheets per page
Default treatment
0.8 0.7
0.8 0.7 0.6
0.6
0.5
0.5 -25 -20 -15 -10
-5
0
5
10
15
20
25
-20
-15
Days since default change
-10
-5
0
5
10
15
20
Days since message
Fig. 2. The effect on resource efficiency.
potential treatment effect will not be confounded by other time events. The moral appeal treatment can be studied in the same fashion, except that we use a smaller sample of printers. Data were collected for each printer on a daily basis, implying that the unit of observation is printer and day. The two outcomes of primary interest are the ratio of sheets to pages, and the number of printed sheets. The first outcome measures how efficiently paper is used, whereas the second measures how much paper is actually saved. We are also interested in the number of printed document pages to account for the possibility that treatment affected printing demand. Table 1 gives some basic pre-treatment statistics, for all the departments combined, for the sample of departments that did not receive the message, and for the sample of departments that did receive the message. Looking at the full sample, we see that the average printer uses 170 sheets of paper to print 221 document pages on a typical day. These figures translate to a ratio of sheets to pages of 0.85, implying that there is considerably more simplex printing than duplex printing in the pretreatment period. In fact, less than one third of all printed sheets are duplex sheets. The relatively large differences in levels between the two subsamples are not statistically significant, which is expected considering the low number of departments in each category (12 and 6, respectively).7
Survey After collecting the data we decided to survey a random subset of the employees that participated in the experiment. The survey was carried out to collect information about people's expressed printing preferences, their general knowledge about default settings, and how they perceived the moral appeal message. We contacted a total of 249 employees and asked whether they wanted to participate in a web survey about printing habits. 119 employees completed the survey, which implies that the response rate was 48 percent. Two days before we invited employees to take the survey, the environmental coordinator sent the exact same moral appeal message that was used in the experiment. This was done since we wanted to include questions about the moral appeal message. Table A1 in the appendix shows the survey responses for the entire sample, and separately for men and women. We will refer the reader to the table when we elaborate on the results from the 7
Note that this is not a real concern since identification relies on within rather than between group comparisons.
6
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
Default treatment
Moral appeal treatment Per printer and day
Per printer and day
250 200 150 100 50
300 250 200 150 100 50
0
0 -5
-4
-3
-2
-1
0
1
2
3
4
5
-4
-3
Weeks since default change Pages
Simplex sheets
Duplex sheets
Pages
-2
-1 0 1 2 Weeks since message Simplex sheets
3
4
Duplex sheets
Fig. 3. The effect on printing demand.
main experiment. We do note, however, that 83 percent of the respondents had read the e-mail containing the moral appeal message, and that 97 percent report that they know how to opt in and out of the default print settings.
Results Main findings We begin by studying the effect on resource efficiency. Fig. 2 shows daily means of the ratio of sheets to pages, before and after the respective treatments. The lower the ratio, the more duplex printing there is.8 The leftmost scatter plot shows the immediate default effect. The very day that we change the default settings, the ratio drops from around 0.85 to 0.73, and once the new level is reached it remains constant. A regression of the outcome on an indicator variable for the treatment period confirms that the difference in daily means is highly significant: tð1171Þ ¼ 6:26, p¼0.000.9 In sharp contrast, there is no visible impact of the moral appeal, not even on the day the message was communicated (see Fig. 2(b)). Regression analysis confirms that the difference in means is not statistically significant: tð388Þ ¼ 1:45, p¼ 0.147. The null effect of the moral appeal suggests that the impact of the default option is not due to implicit recommendation. The change in the default settings clearly improved resource efficiency. However, the improved efficiency does not automatically imply that less paper were being used in total. If people adapt their printing habits, for example by starting to print more in general, this will offset the efficiency gain (see, e.g., Khazzoom, 1980). This type of adverse behavior may be particularly relevant for nudging interventions because there is a risk that people feel manipulated (Amir and Lobel, 2009). It turns out that the number of printed pages is completely unaffected by the new default settings, as seen in Fig. 3(a). In terms of a formal test, the difference in means before and after the default intervention is far from significant: tð1249Þ ¼ 0:53, p¼0.593. This finding is important as it shows that users did not respond by printing more (or less) document pages; we hence conclude that printing demand is independent of the default setting.10 Instead, it is the relation between simplex and duplex printing that has shifted (gray and white bars). Both the reduction in the number of simplex sheets (tð1249Þ ¼ 4:29, p¼ 0.000) and the increase in the number of duplex sheets (tð1249Þ ¼ 2:75, p¼0.011) are highly significant. We also note that there is still a substantial amount of simplex printing after the duplex default has been implemented, suggesting that it is easy for people to opt out if they wish. The increase in duplex printing implies that paper consumption is substantially lower after the settings were changed; the number of sheets used per day has dropped by 15 percent on average (tð1249Þ ¼ 2:45, p¼0.014). Turning to the moral appeal treatment in the graph to the right there is further evidence that asking people to use duplex printing has no impact. In particular, when comparing bars on either side of the vertical line we see no drastic shift in the relation between duplex and simplex printing (the conclusion is confirmed by t-tests). This null result can be contrasted to the responses in the survey, where 35 percent of respondents said that the message would affect their own printing behavior and 75 percent 8 Documents of only one page will have a ratio of one irrespective of the printing mode. This implies that the lower bound is in practice strictly larger than 0.5 (document-level data from two printers in our sample suggest it is about 0.55). To ensure a balanced panel, we restrict the time window to five (four) weeks before and after the default change (moral appeal message). 9 The t-statistics in this section come from regressions of the respective outcomes on a dummy variable that equals one for the first 25 days after a printer's default change and zero for the 25 days prior to the change. Given that we have at most 25 clusters, we performed the regressions using both robust and cluster robust standard errors (by printer) and report t-statistics for the more conservative of the two. 10 The absence of a decrease in printed pages also suggests that employees did not switch to other printers as a response to the default change (for example by purchasing a private office printer), as this would imply a decrease in printed pages for the experimental printers. Estimates from two of the printers in the sample that report document-level data confirm that the length of documents and the number of documents are unaffected as well.
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
7
1
Sheets per page
0.9
0.8
0.7
0.6
0.5 -5
0
5
10
15
20
25
30
35
Weeks since default change Fig. 4. Long run default effect.
1
Default: Simplex Default: Duplex
1 0.9 Sheets per page
Sheets per page
0.9
0.8
0.7
0.8 0.7 0.6
0.6
0.5 0.5
Simplex
Academic departments
Administrative departments
Duplex
Duplex
Simplex
SDD
Simplex
SSD
Duplex
Simplex Message Duplex
SMD
Fig. 5. Default effect by type of department (a), and timing of intervention (b).
believed it would affect their colleagues (see Table A1 in the appendix). One potential explanation for the discrepancy is that people contemplate on the issue when asked, whereas their actual behavior reflects a more automatic response. While many behavioral interventions, such as feedback and social comparison, often have large short-run effects there is limited evidence on the impacts in the longer run (Allcott and Rogers, 2014). An important question therefore is whether the default effect that we observe endures over time. To investigate this, we collected additional data from the same set of printers during a second consecutive five-week period commencing 28 weeks after the intervention. Fig. 4 plots weekly averages of the ratio of sheets to pages. Strikingly, more than six months after the change the number of sheets per page is at the same low level as in the very short run. The fact that people did not revert to simplex printing shows that default rules can be powerful also in the longer run. Heterogeneous treatment effects This section takes a closer look at default effect heterogeneity. We first look at differences with respect to the type of department. There are both academic departments and administrative departments in the sample, with a 50 percent share of both. Heterogeneity in this dimension may inform us about the generalizability of the results: if the effect is mainly driven by academic departments, the results may be confined to universities around the world; if, on the other hand, the effect is driven by administrative departments the results may be of more relevance for government agencies and private businesses. As seen in Fig. 5(a) the default effect is independent of the type of department (the point estimate of the difference–in–differences is 0.004, and it is not statistically significant: tð1171Þ ¼ 0:10, p¼ 0.918). This result is consistent with a more general effect applicable to different types of workplaces. Aggregating the data and normalizing time to zero could potentially hide differences in the default effect across the departments. For example, we may think that the effect varies depending on the timing of the default change, and on whether the employees were exposed to the moral appeal before the change. To address this concern, Fig. 5(b) shows the default effects (and 95%-confidence intervals) separately for each of three different treatment groups. The first group (SDD) changed to duplex at the beginning of the second period (during week six), whereas the two other groups (SSD and SMD) changed to duplex at the beginning of the third period (during week 10). The SMD group is the one where the employees
8
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
1 Fraction duplex printing
Fraction duplex printing
1 0.8 0.6 0.4
0.8 0.6 0.4
0.2
0.2
0
0 Default: Simplex Females
Default: Duplex
Default: Simplex
Males
Age: 26 - 45
Default: Duplex Age: 46 - 68
Fig. 6. Default effect by gender (a), and age (b).
received the moral appeal message at the beginning of the second period (four weeks before changing to duplex). As expected, there is a significant drop in the number of sheets per page for the SDD group during the second period (tð232Þ ¼ 2:38, p ¼0.063), and for the SSD group during the third period (tð308Þ ¼ 3:97, p ¼0.005). For the SMD group we expect no drop in the second period as the moral appeal treatment had no effect. In the third period, however, we expect the default effect to kick in, and it should be of similar size as for the two other groups. Notably, this is exactly what we observe in Fig. 5(b). The fact that the default effect is equally large for the SMD group as for the other two groups shows that the employees who were subjected to the moral appeal message were able to print more duplex if they wanted to.11 It also strengthens the conclusion that the default effect is not explained by ignorance, or by implicit recommendation. From the figure we also draw the conclusion that there are no underlying time trends in printing behavior, as the outcome is stable for SDD (SSD) between periods 2 and 3: tð229Þ ¼ 0:59, p ¼0.550 (periods 1 and 2: tð296Þ ¼ 0:36, p¼0.732). We finally make use of two printers that report document-level data to provide tentative evidence on whether the default effect varies by individual characteristics. The sample consists of 44 employees (31 females) who printed 5980 documents in total. We first divide the effect by gender. As shown by the dashed line in Fig. 6(a), women are roughly 50 percent more likely to use duplex after the default change. What is more intriguing is that the effect is substantially larger for men, as shown by the steeper slope of the solid line in the figure. In fact, men are almost four times more likely to use duplex after the change. It is striking to see men's strong tendency to stick to the current default regime: when simplex printing is the default, men use simplex printing; when duplex is the default, they use duplex. This tendency is much less pronounced for women. Regression analysis confirms that the interaction effect is marginally significant: tð5980Þ ¼ 1:71, p¼0.095.12 Fig. 6(b) looks at the importance of age in explaining the default effect. An employee is defined as “old” if his or her age is at or above the sample median, which is 46. The point estimates suggest that the default effect is larger for employees above the median age, but the age difference is not statistically significant at conventional levels. The above results suggest that even though males (and to some extent the older employees) are not opposed to improvements in resource efficiency, they are less likely to actively take advantage of it. This could in turn reflect that male employees simply care less about the presentation of a document, or that they are less technologically savvy (and hence do not know how to opt out of the default option). To assess these different explanations we compare survey responses for men and women. Both men and women know how to opt in and out of the default settings (men: 95%; women: 97%) and both groups perceive duplex printing as proenvironmental (men: 98%; women: 97%). Hence, there seems to be no difference in technological skills or attitudes. When asked about their printing preferences (and their willingness to pay to have their preference realized), 60 percent of the men are indifferent between using simplex and duplex, compared to 50 percent of the women.13 Men are also more likely to state that the default option affects their printing behavior (Men: 55%; Women: 47%). Overall, the survey responses indicate that men tend to have weaker preferences than women, and that indifference seems to be a plausible explanation for the results in general. 11 12
The point estimate of the difference–in–differences is 0.010, and it is not statistically significant: tð920Þ ¼ 0:28, p¼ 0.779). We estimate a linear probability model of the following form: yi;t ¼ α þ β1 After t þ β2 Malei þ β3 Malei After t þ εi;t
where yi;t is a dummy variable that equals one if individual i used duplex to print a specific document at time t, After is a time dummy that equals one in the period after the default change, and Male is a gender dummy that equals one if the employee is a man. The coefficient of interest, β3, measures to what extent men are affected more than women by the default change (i.e., if they have a higher propensity than women to use duplex in the period after the change). When analyzing the importance of age we run the same model but with a dummy for age instead of gender. To account for serial correlation we clustered standard errors at the employee-level ðn ¼ 44Þ. 13 A person is classified as indifferent if he or she stated that it does not matter whether a document is printed simplex or duplex, or if they stated a preference but would not pay any amount to have their preference realized in a situation where the opposite printing alternative was free of charge.
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
9
Welfare One motivation for implementing a default option in any setting is that it is Pareto improving: those who prefer the default option and thus save on time and effort are made better off, while those who nevertheless have to choose actively are unaffected. The choice of which alternative to use as the default is, on the other hand, nontrivial. We have already shown that changing the default option to duplex did, indeed, save resources. However, we should also take into account the fact that the intervention may have inconvenienced people. That is, even though it was good for the environment, it may have increased the time and effort employees spent on printing. To analyze this aspect we use the opt-out behavior in the different default regimes as a proxy for printing preferences. From Fig. 4 it is clear that the number of sheets per page drops from around 0.85 to 0.7 as we change the default from simplex to duplex. This suggests that one third ½ð1 0:85Þ=0:45 ¼ 1=3 opts out when simplex is the default, and that one third ½ð0:70 0:55Þ=0:45 ¼ 1=3 opts out when duplex is the default. It therefore seems reasonable to assume that the fraction of users that prefer duplex is roughly similar to the fraction of users that prefer simplex. We have also looked closer at individual-level data for two of the printers in the sample in order to grasp whether the aggregate data grossly overestimate the share of people with duplex preferences. To the contrary, we find that 80 percent of the employees in this sample either have preferences for duplex printing or are what we refer to as indifferent (i.e., they follow the default option). A second important observation is that people do not revert to simplex printing in the longer run, as they eventually make more active decisions. Survey data support these findings. 77 percent of the respondents state that they prefer, or strongly prefer, duplex printing, whereas only 13 percent prefer to use simplex. Furthermore, 95 percent say that they would choose duplex as the default if they were in charge of the printers at the department. As skeptics of nudging have pointed out, it is not obvious that those in charge of setting defaults are all-knowing selfless beings that have nothing but the common good in mind (see Sugden, 2009). One message from the current experiment, with respect to that debate, is that people in charge actually might be too conservative when choosing the default alternative. In sum, we conclude that the change in the default was, in fact, welfare enhancing.
Concluding discussion We conducted a natural field experiment at a large public university to measure the causal effect of two paper saving programs. Changing printers' default option, from simplex to duplex printing, reduced paper consumption by as much as 15 percent. The default effect was immediate, universal across departments, and remained intact more than six months after the intervention. An intriguing finding is that the demand for printing was completely stable throughout the whole experiment period. The absence of undesired behavioral responses suggests that resource efficient defaults may offer an attractive benefit as a policy tool by avoiding adverse effects. The fact that people did not respond by taking active steps to reduce consumption of paper could suggest that green defaults induce pro-environmental behavior without affecting people's environmental awareness. This question, however, needs to be further explored in future research. In contrast to the substantial default effect, we find no effect of sending a moral appeal message that encouraged employees to cut back on paper use. The absence of any impact of the moral appeal leads us to conclude that the default effect is not explained by ignorant employees or misperception of the employer's intentions. Arguably, default options are more likely to serve as implicit advice when the decision is plagued by uncertainty, individual feedback is limited and norms about appropriate behavior are unclear (e.g. choosing a retirement savings plan). Notably, deciding whether to print duplex or simplex does not share any of these characteristics. A compelling, and consistent, explanation for the default effect in our setting is that users have weak preferences over the two alternatives. Weak preferences should induce people to use quick and intuitive judgment for choices that they often repeat and see as less important. Presumably, most people do not take the environment into account when deciding how to print inasmuch as their isolated action will have little (global) impact anyway. Survey data support this interpretation, as 60 percent of the men and 50 percent of the women who responded are indifferent between simplex and duplex printing. Indifference can also explain the fact that people did not revert back to simplex printing in the long run — a result that is difficult to reconcile with, for example, choice avoidance or inattention (as decision errors should diminish with experience). As always, one should be cautious in making far-reaching inferences from one single study. With respect to the moral appeal treatment, we do not claim that persuasive communication is useless in general. Public environmental awareness campaigns may, for example, have contributed to the positive views expressed in our survey, and the relatively large degree of duplex printing observed prior to the change in default. The null result does, however, indicate that there are limits to what pro-environmental reminders and campaigns can achieve in the short run. With respect to the default intervention it should be noted that the absence of behavioral demand responses may hinge on the fact that a majority of the employees were positively inclined to duplex printing. In contrast, recent research suggests that the support for nudges evaporates, and that behavior may even backfire, when people suspect that they might end up with outcomes that are inconsistent with their interests or their values (Amir and Lobel, 2009; Sunstein, 2015). Since resource efficient defaults typically save people money it may, nevertheless, be a result that generalize to other settings more often than not. We should also emphasize that other situations, that readily come to mind for similar interventions, are characterized by the fact that the green alternative
10
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
is associated with a monetary premium (e.g. ecological products and renewable energy). Whether the results extend to those situations are therefore still an open question (but Ebeling, 2013 reports affirmative results in a recent working paper). To be clear, there are still many situations where, as in the current study, saving resources comes at a zero or even negative monetary cost (e.g., turning off the lights when leaving a room or leaving electronic devices on standby). Our results point out that there could be large gains from carefully choosing, and creating, “no-action” options in such situations. Policy makers should be particularly supportive of this type of intervention since a socially preferred allocation can be reached without spending money and without limiting individual choice (Thaler and Sunstein, 2003). One area in which our results can be directly applied is the so-called go-paperless-initiatives by for example banks, government agencies and telephone operators. Another possibility is to stimulate suppliers, through mandates or incentives, to preprogram home appliances with energy efficient default options.
Acknowledgments We wish to thank the editor Fredrik Carlsson, and three anonymous referees for helpful comments and suggestions that significantly improved this paper. Hunt Allcott, Steffen Altmann, Rachel Croson, Tom Cunningham, Stefano DellaVigna, Peter Fredriksson, Magnus Johannesson, Therese Lindahl, Erik Lindqvist, Peter Martinsson, Rupert Sausgruber, Peter Skogman Thoursie, Bertil Tungodden and Robert Östling provided many valuable comments on earlier drafts. We would also like to thank seminar participants at the Norwegian School of Economics, the University of Bergen, and the EAERE Conference in Toulouse 2013. Finally, we want to express our gratitude to Anne Liv Scrase for excellent proofreading, to Jenny Lilliehöök, Fredrik Holm and Miroslav Lustinsky for administrative support, and to the departments that accepted to be part of the study. Financial support from the Jan Wallander and Tom Hedelius Foundation is gratefully acknowledged. All errors are our own.
Fig. A1. Simplex default.
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
11
Fig. A2. Duplex default.
Appendix A A.1. Print screens See Figs. A1 and A2. A.2. Default change notice Dear Colleague, In order to save the environment and resources, we have decided to change the default setting from simplex to duplex printing on printer X. The change enters into force tomorrow, x/x-2012. The possibility to print simplex remains, and if you absolutely want one-sided print jobs as a pre-selection, you can change your personal default to simplex. This is done under “Printer Settings” on your particular computer. If you have any problems in the transition or if you want help with your own printer settings, you can contact the IT media helpdesk by phone: XX-XXXXXX. Best regards X X, Head of Department A.3. Moral appeal message Dear Colleague, At X we strive to reduce our impact on the environment, and as an employee you can be part of this process. An easy way to save resources is to keep paper consumption low. You can contribute by choosing to print on both sides of a sheet. Duplex printing is available on most of the printers at X, and you can also make this printing mode the default option on your
12
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
computer. If you need help with setting up your own printer settings, you are welcome to contact the IT media helpdesk by phone: XX-XXXXXX. Thank you for your cooperation! Best regards
Table A1 Survey responses. QUESTION
All
Men
Women
QUESTION
All
Men Women
b
4. Willingness to pay: Simplex (%)
1. E-mail Read e-mail? (% answered yes) E-mail will affect my own printing? (% answered yes) E-mail will affect others' printing? (% answered yes) Perceived e-mail: (% of all respondents) Very positive/encouraging/needed
84 35
81 35
86 35
More than 1 kr Less than 1 kr but more than 50 öre
0.0 0.0 6.7 0.0
0.0 10
75
77
74
Less than 50 öre but more than 1 öre
27
20
30
8.5 7.1
9.2
Less than 1 öre but more than 0 kr Nothing, I would print double-sided
6.7 0.0 60 80
10 50
Positive/encouraging/needed
49
60
43
5. Knowledge (% answered yes)
Neither positive or negative
37
26
43
97
95
97
Negative/intrusive/irritating
5.1
7.1
4.0
80
85
77
Very negative/intrusive/irritating
0.0 0.0
0.0
Know how to change from single-sided/doublesided? Change some of the pre-set settings before printing? Made changes to your local computer's print settings?
59
64
55
50
55
47
92
93
91
95 98
95 98
95 97
20 12 73 83 6.8 4.8
25 67 7.9
26
30
2. Printing preferences (% of all respondents)
6. Default
Strongly prefer duplex
36
31
38
Prefer duplex
41
43
40
Does not matter Prefer simplex
11 10
14 9.5
9.2 11
2.5 2.4
2.6
4.4 10 30 10 46
3.4 6.8 34 12 44
Strongly prefer simplex a
3. Willingness to pay: Duplex (%) More than 1 kr Less than 1 kr but more than 50 öre Less than 50 öre but more than 1 öre Less than 1 öre but more than 0 kr Nothing, I would print simplex
6.5 16 23 6.5 48
Default will affect my own printing? (% answered yes) Default will affect others' printing? (% answered yes) Choice of default option? (% answered duplex) Is duplex printing pro-environmental? (% answered yes) Default on your main printer? (% of respondents) Simplex Duplex Do not know Change printer if your main printer changed default? (% answered yes)
19
a Question: Assume that printing simplex documents was free of charge, but that it costs a certain amount every time you printed a duplex document. How much would you be willing to pay, per sheet, to print a duplex document? This question was asked to those who stated that they preferred, or strongly preferred, duplex (n¼ 90). b Question: Assume that printing duplex documents was free of charge, but that it costs a certain amount every time you printed a simplex document. How much would you be willing to pay, per sheet, to print a simplex document? This question was asked to those who stated that they preferred, or strongly preferred, simplex (n ¼15).
X X, Environmental coordinator, Faculty X X, Environmental coordinator, The Environmental council A.4. Survey responses Table A1.
References Allcott, H., Mullainathan, S., 2010. Behavior and energy policy. Science 327 (5970), 1204–1205. Allcott, H., Rogers, T., 2014. The short-run and long-run effects of behavioral interventions: experimental evidence from energy conservation. Am. Econ. Rev. 104 (10), 3003–3037. Altmann, S., Falk, A., Grunewald, A., 2013. Incentives and information as driving forces of default effects. IZA Discussion Papers No. 7610. Amir, O., Lobel, O., 2009. Stumble, predict, nudge: how behavioral economics informs law and policy. Columbia Law Rev. 108, 2098–2138.
J. Egebark, M. Ekström / Journal of Environmental Economics and Management 76 (2016) 1–13
13
Beshears, J., Choi, J., Laibson, D., Madrian, B., 2010. The limitations of defaults. Technical Report, National Bureau of Economic Research. Bronchetti, E.T., Dee, T.S., Huffman, D.B., Magenheim, E., 2013. When a nudge isn't enough: defaults and saving among low-income tax filers. Natl. Tax J. 66 (3), 609–634. Carlsson, F., Johansson-Stenman, O., 2012. Behavioral economics and environmental policy. Annu. Rev. Resour. Econ. 4 (1), 75–99. Catlin, J.R., Wang, Y., 2013. Recycling gone bad: when the option to recycle increases resource consumption. J. Consum. Psychol. 23 (1), 122–127. Chetty, R., Friedman, J.N., Leth-Petersen, S., Nielsen, T.H., Olsen, T., 2014. Active vs. passive decisions and crowd-out in retirement savings accounts: evidence from Denmark. Q. J. Econ. 129 (3), 1141–1219. DellaVigna, S., Gentzkow, M., 2010. Persuasion: empirical evidence. Annu. Rev. Econ. 2 (1), 643–669. Dinner, I., Goldstein, D.G., Johnson, E.J., Liu, K., 2011. Partitioning default effects: why people choose not to choose. J. Exp. Psychol.: Appl. 17 (4), 332–341. Dwenger, N., Kleven, H., Rasul, I., Rincke, J., 2014. Extrinsic and intrinsic motivations for tax compliance: evidence from a field experiment in Germany. Ebeling, F., 2015. Non-binding defaults and voluntary contributions to a public good—clean evidence from a natural field experiment. working paper no: 66 published by the University of Cologne in its "Working Paper Series in Economics". https://ideas.repec.org/p/kls/series/0066.html. EU, 2011. A Resource-efficient Europe—Flagship Initiative Under the Europe 2020 Strategy. European Union. Fellner, G., Sausgruber, R., Traxler, C., 2013. Testing enforcement strategies in the field: threat, moral appeal and social information. J. Eur. Econ. Assoc. 11 (3), 634–660. Goldstein, N.J., Cialdini, R.B., Griskevicius, V., 2008. A room with a viewpoint: using social norms to motivate environmental conservation in hotels. J. Consum. Res. 35, 472–482. Harrison, G.W., List, J.A., 2004. Field experiments. J. Econ. Lit. 42 (4), 1009–1055. Jaffe, A.B., Stavins, R.N., 1994. The energy-efficiency gap: what does it mean?. Energy Policy 22 (10), 804–810. Johnson, E.J., Goldstein, D., 2003. Do defaults save lives? Science 302, 1338–1339. Kahneman, D., 2011. Thinking, Fast and Slow. Farrar, Straus and Giroux, New York. Kallbekken, S., Sælen, H., Hermansen, E.A.T., 2013. Bridging the energy efficiency gap: a field experiment on lifetime energy costs and household appliances. J. Consum. Policy 36 (1), 1–16. Khazzoom, J.D., 1980. Economic implications of mandated efficiency in standards for household appliances. Energy J. 1 (4), 21–40. Löfgren, Å., Martinsson, P., Hennlock, M., Sterner, T., 2012. Are experienced people affected by a pre-set default option: results from a field experiment. J. Environ. Econ. Manag. 63, 66–72. Madrian, B.C., Shea, D.F., 2001. The power of suggestion: inertia in 401(k) participation and savings behavior. Q. J. Econ. 116, 1149–1187. McKenzie, C.R., Liersch, M.J., Finkelstein, S.R., 2006. Recommendations implicit in policy defaults. Psychol. Sci. 17 (5), 414–420. Pichert, D., Katsikopoulos, K.V., 2008. Green defaults: information presentation and pro-environmental behaviour. J. Environ. Psychol. 28, 63–73. Pruckner, G.J., Sausgruber, R., 2013. Honesty on the streets: a field study on newspaper purchasing. J. Eur. Econ. Assoc. 11 (3), 661–679. Schultz, P.W., Nolan, J.M., Cialdini, R.B., Goldstein, N.J., Griskevicius, V., 2007. The constructive, destructive, and reconstructive power of social norms. Psychol. Sci. 18 (5), 429–434. Shafir, E. (Ed.), 2012. The Behavioral Foundations of Public Policy, Princeton University Press, Princeton, NJ. Sugden, R., 2009. On nudging: a review of nudge: improving decisions about health, wealth and happiness by Richard H. Thaler and Cass R. Sunstein. Int. J. Econ. Bus. 16 (3), 365–373. Sunstein, C.R., 2015. Do people like nudges? Available at SSRN. Sunstein, C.R., Reisch, L.A., 2014. Automatically green: behavioral economics and environmental protection. Harv. Environ. Law Rev. 38, 128–158. Thaler, R.H., Sunstein, C.R., 2003. Libertarian paternalism. Am. Econ. Rev.: Pap. Proc. 93 (2), 175–179. Thaler, R.H., Sunstein, C.R., 2008. Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press, New Haven, CT. UN, 1992. Agenda 21. United Nations.