A default nudge in waste management: assessing the impact of explicit consent for unaddressed mail

ABSTRACT

On 1 January 2018, the municipality of Amsterdam changed the system for the reception of unaddressed mail from presumed consent to explicit consent to reduce paper waste. This policy can be defined as a default nudge. The no-choice population received unaddressed mail in the presumed consent system but not in the new explicit consent system. Residents receive unaddressed mail only when they actively decide to put an opt-in sticker on their mailbox. This study assesses the effectiveness and social benefits of this nudge. The effect on paper waste is estimated using a difference-in-differences approach in which several other Dutch municipalities function as the control group. Our main finding is that the default nudge results in a reduction of paper waste between 5.3% and 11%. Social benefits of this reduction include, for example, lower carbon emissions for collection and transport for paper waste, which are equivalent to yearly benefits between approximately €135,000 and €285,000 in Amsterdam. If all Dutch municipalities implement the system of explicit consent for unaddressed mail, the yearly benefits are approximately between €14 million and €30 million. The default nudge is a low-cost policy to implement and, therefore, offers municipal policymakers a cost-effective way to reduce waste.

KEYWORDS:

• Default option
• green nudge
• natural experiment
• nudge
• waste management

JEL CODES:

• D91
• Q56
• Q58

Key Policy Highlights

• The change of presumed consent to explicit consent for unaddressed mail in Amsterdam allows for a unique opportunity to test a green nudge using a natural experiment.

• An econometric difference-in-differences approach shows that the implementation of the green nudge results in a 5.3% to 11% reduction in paper waste.

• Private and environmental benefits include lower costs from collecting, transporting and recycling paper waste.

• Changing the default option for junk mail is an efficient way for policy makers to reduce environmental pressure.

1. Introduction

This study assesses the effectiveness and social benefits of a default nudge for unaddressed mail on the amount of paper waste in the Netherlands. Over one million trees a year are cut around the globe to produce junk mail (McKenzie-Mohr et al. 2012). The growing world population, in combination with increasing urbanisation and expanding consumption, leads to increased competition for resources (George, Schillebeeckx, and Liak 2018). The consequences include environmental degradation and the depletion of natural resources (Hussain, Khan, and Zhou 2020). In a circular economy, reducing waste is an important strategy for reducing pressure on the earth’s stock of natural resources (Borrello et al. 2017). Lansink (2018) formulated a hierarchy on the most efficient ways to reduce waste and theorised that rethinking existing consumption patterns was the best way to avoid waste.

Consumption behaviour can be influenced using nudges (Sunstein 2014a). A nudge is a policy tool that makes a slight change in the choice architecture an individual faces, which might thereby lead to a change in behaviour (Sunstein 2014a). Nudges attempt to steer people to a choice instead of commanding them towards, for example, more environmentally friendly behaviour. Various green nudges, such as using clear bags for plastic waste (Akbulut-Yuksel and Boulatoff 2021), changing the default on computer printers to promote double-sided printing (Egebark and Ekström 2016) or receiving green energy as a default option (Pichert and Katsikopoulos 2008), have been proposed in the literature. The magnitude of the effect of these nudges differs and is context dependent. It is important to estimate the effects of green nudges in practice to understand whether the policy accomplishes its goal (Hansen and Jespersen 2013).

The municipality of Amsterdam implemented a nudge that is similar to the rethinking of consumption as Lansink (2018) described. Starting on 1 January 2018, Amsterdam replaced an opt-out mail system for an opt-in mail system with regard to the reception of unaddressed mail (City of Amsterdam 2020). The opt-in system means that a resident of Amsterdam has to explicitly give consent to receive unaddressed brochures, flyers and other junk mail in their letter box. Hence, the current status quo is that the resident does not receive unaddressed mail. In contrast, a resident had to explicitly refuse the reception of junk mail in the previous opt-out system. Thus, the default option for the reception of unaddressed mail has changed. When no choice regarding the reception of unaddressed mail is made, the previous no-choice option was that the resident would receive these flyers in the previous status quo. However, the new system prohibits sending junk mail when explicit consent is not given. The goal of this policy was to reduce the amount of paper and cardboard waste and thereby reduce environmental pressure (Vereniging van Nederlandse Gemeenten [VNG] n.d.). The reasoning behind this policy is that the no-choice population that previously received junk mail does not receive this junk mail in the new system.

This study estimates whether the default nudge accomplishes its goals of reducing paper waste and the extent to which it benefits society. It is expected that the change in the opt-in mail system negatively influences the amount of waste (Schmitz 2018), but the size of this effect and the gains for society are unclear as of yet. The estimated reduction in paper waste is expressed in terms of monetary value to assess the extent to which society benefits from the default nudge. This insight will allow policymakers to decide whether the opt-in mail system is effective and economically desirable or not. Policymakers from other municipalities who aim to reduce the environmental impact of their municipality may be motivated to adopt this policy as well if the default nudge accomplishes its goal.

We believe that our study fills an important gap in the literature. The recycling and recovery rate of paper waste is relatively high, as approximately 75% of all paper waste can be recycled back into new paper (Holwerda, Tiekstra, and Hooimeijer 2019). For this reason, most studies on paper waste have focused on socioeconomic determinants of paper waste or the effect of paper recycling policies (e.g. Bruno, Bianchi, and Sánchez 2022; Khan and Burney 1989; Monavari et al. 2012). However, recycling also has its environmental costs, where a total reduction of waste results in less environmental impact compared to recycling (Lansink 2018). Reducing the demand for paper has a significant carbon-reducing potential (Szabó et al. 2009). However, the literature often overlooks paper waste-reducing policies, with the exception of Egebark and Ekström (2016) who tested the impact of a default on printers on paper use and found that defaults help avoiding unintended behaviour. This paper contributes to the literature of green nudges with a potentially large impact with the application on paper waste. Although research has been carried out on green nudges, few researchers have performed a natural experiment. Akbulut-Yuksel and Boulatoff (2021) are an exception, they applied a regression discontinuity design to estimate the effects of a clear bag policy in Canada. Existing studies on green nudges mostly perform an economic field experiment, such as adding stickers to mailboxes, using see-through bags for waste or changing defaults for computer printing (e.g. Liebig and Rommel 2014; Akbulut-Yuksel and Boulatoff 2021; Puig-Ventosa, Jofra-Sora, and Freire-González 2015; Egebark and Ekström 2016). The large scale of the opt-in mail policy allows for a natural setting for the nudge to be tested in. The importance and originality of this study are that it explores a natural experiment on an implemented policy using a difference-in-differences design. The advantage of this design is that it allows for more externally valid estimates of the effect of the default nudge than experimental designs do while still being able to estimate the counterfactual and draw causal conclusions.

The remainder of this article is structured as follows. Section 2 gives a policy background of the default nudge. Section 3 provides a literature review on green nudges. Section 4 describes the methods, including the econometric techniques, data used, and the approach for monetising paper waste. Section 5 presents the results, which are discussed in Section 6. Section 7 concludes.

2. Policy background

With respect to unsolicited mail, the United States and most countries in Europe use an opt-out system, where a household explicitly refuses junk mail (e.g. Liebig and Rommel 2014; Standley 2019). Unaddressed mail forms approximately 30–50% of all paper waste in urban areas in the Netherlands (Van Gerrevink, Uitgeverij Steden Driehoek, and All-In Verspreidingen 2019; Milieu Centraal n.d.). Amsterdam was the first municipality in the Netherlands to introduce the opt-in mail system on 1 January 2018. Before the implementation of this nudge, approximately 55% of all households in Amsterdam had no sticker to indicate consent or refusal for unaddressed mail on their mailbox (Schmitz 2018). The municipality outlined the benefits of the new system and the consequences of making a certain choice regarding unaddressed mail (City of Amsterdam 2021). This campaign also included a website on which wrongfully sent unaddressed mail could be reported. The households that made no choice regarding junk mail did receive junk mail before the default nudge. After the policy was implemented, the no-choice population suddenly stopped receiving unaddressed mail. Schmitz and Slot (2018) examined how many people were aware of the system change in Amsterdam. They found that 93% of all residents of Amsterdam were aware of the new opt-in mail system in 2018, meaning that the information campaign of the municipality accomplished its goal. Schmitz (2018) found in a random sample that 13.8% of all households in Amsterdam adopted the explicit consent sticker six months after its introduction. Meanwhile, the municipality of Amsterdam faced resistance from local entrepreneurs not able to reach their customers anymore (van Zoelen 2017). These entrepreneurs started a legal case against the municipality for this reason. In 2021, the Dutch Supreme Court decided that the default mail nudge is legitimate, which paved the way for other municipalities such as Utrecht, Rotterdam and Amersfoort to introduce the explicit consent mail system (Dutch Supreme Court 2021).

3. Literature review

3.1. Status quo bias

Nudges are an example of the application of behavioural insights on the economy. The nudge seeks to overcome the status quo bias (Schirrmacher et al. 2019), which means that individuals tend to value existing options, patterns or behaviour more than the alternatives that are known to be more preferable than the existing option (Polites and Karahanna 2012). People often choose the option with the least resistance (Thaler and Sunstein 2008). Therefore, a default nudge might change the status quo, allowing the individual to choose the option the policymaker deems as the most preferable. Lee and Joshi (2017) described three situations where the status quo bias is more persistent: when an individual is faced with a large set of options, where a system of behaviour is used for a long time and when the individual does not recognise the possibility of a choice. The last two situations apply to the reception of unaddressed mail. First, the opt-out mail system has been in use for a long time in Amsterdam, possibly resulting in individuals less willing to put effort into explicitly refusing the reception of unaddressed mail. Second, it may be possible that some individuals did not know about the possibility of the explicit refusal of the unaddressed mail, thus receiving mail when they did not wish to. Croson and Treich (2014) demonstrated that a green nudge that played into the status quo bias through a default option tended to have low costs while generating significant results. A green nudge steers individuals towards a more sustainable option, for example, the opt-in mail system in Amsterdam. Playing into the status quo bias by using nudges offers a relatively new approach to facing environmental challenges (Croson and Treich 2014).

3.2. Nudges aimed at reducing waste

Puig-Ventosa, Jofra-Sora, and Freire-González (2015) estimated the effect of a no junk mail sticker in Barcelona by conducting a field experiment. One household voluntarily put a ‘no junk mail’ sticker on their mailbox, and the same household proposed another household within the same building that functions as the control unit. This was an active choice against the reception of junk mail, and the control household actively chose to receive junk mail. The researchers only nudged the participants to actively make a choice, making the rejection of junk mail the default option. This is different from the Amsterdam opt-in mail nudge because Amsterdam implemented explicit consent stickers, whereas Puig-Ventosa et al. examined the effect of explicit refusal stickers. Puig-Ventosa, Jofra-Sora, and Freire-González (2015) found that households with the no junk mail sticker on their mailbox generated significantly less waste. The presence of the stickers is associated with 0.31–0.36 kg less waste per person per year, which can be translated to a 0.4% decrease in paper waste.

Egebark and Ekström (2016) performed a field experiment where the impact of a moral message and a default to cut back printing have been tested. It was found that where a moral message to use double-sided printing had no impact on printing behaviour, the default option resulted in a 15% reduction of paper consumption at a Swedish university. This impact on behaviour also persisted in the longer run. Another study on the effects of a default nudge in the waste domain was conducted by Akbulut-Yuksel and Boulatoff (2021). They used a regression discontinuity design to estimate a nudge that attempted to promote recycling in Canada, the so-called clear bag policy. Plastic had to be recycled in see-through plastic bags to nudge households towards a more responsible separation of different types of waste. The policy introduced some form of social control, potentially resulting in less waste other than plastic that was recycled as plastic. This policy also has the potential to reduce waste in general (Akbulut-Yuksel and Boulatoff 2021). The discontinuity exploited by Akbulut-Yuksel and Boulatoff (2021) was the date the plastic bags were distributed. Several districts of the city Halifax in Canada received the plastic bags on different days of the week. This regression discontinuity design was possible because the policy was abruptly implemented and, therefore, allowed to exploit a sudden change of garbage collection between different city districts. The main result was that the nudge indeed reduced the total waste by 27% and increased recycling by 15%.

Rivers, Shenstone-Harris, and Young (2017) investigated the effects of a symbolic levy on plastic bags on waste. Nudges generally do not involve a price mechanism (Sunstein 2014b). Therefore, the levy on plastic bags was not a nudge in a typical way. However, the symbolic levy on disposable plastic bags only reminds the consumer of the environmental consequences of the plastic bag (Rivers, Shenstone-Harris, and Young 2017). Using longitudinal data in Toronto, they found that the levy on plastic bags significantly increased the use of reusable plastic bags. The main finding was that the policy was especially effective in encouraging people who were already using reusable bags to reuse them more frequently, while the effects on other people remained constant. People used 3.4%-points more reusable shopping bags.

Finally, Liebig and Rommel (2014) performed a study on the effect of a forced or active choice for the refusal of junk mail in a field experiment. Two types of treatment were used. The first treatment nudged active choice for or against junk mail by putting explicit refusal stickers in people’s mailboxes. The treated group had to add the sticker to their mailbox themselves. The second treatment is a default nudge, where explicit refusal stickers were halfway attached to the mailbox. It was forced to choose to remove or apply the sticker to the mailbox. Results indicate that forced choice significantly results in a larger rejection of junk mail than the free choice condition does. While 15.98% of the population treated with the active choice nudge chose to reject junk mail, of the population treated with the forced choice, 21.66% made the same choice. Liebig and Rommel (2014), therefore, concluded that forced choice resulted in significantly more choices against junk mail. This nudge differs from the default nudge in Amsterdam, since it examines the adoption of explicit refusal, where Amsterdam implemented a system of explicit consent. Moreover, Liebig and Rommel (2014) do not look into the eventual outcomes of these policies on paper waste.

4. Methodology

4.1. Estimation strategy

The policy intervention took place in Amsterdam on 1 January 2018, while the situation in the other municipalities remained constant. This timing offers a unique opportunity for a difference-in-differences estimation, in which the difference in the natural logarithm of the monthly waste per person in Amsterdam in the period before and after the intervention was compared with the difference in the control municipalities before and after the intervention. Since the default nudge is tested in a practical setting, results are more externally valid than other artificially constructed field experiments with a smaller sample (e.g. Liebig and Rommel 2014; Puig-Ventosa, Jofra-Sora, and Freire-González 2015). The effect of the default nudge was estimated using the following formula: (1) $\ln (kgpwastepp{)}_{it}=\theta +\gamma treatmen{t}_i+\lambda pos{t}_t+\delta (treatmen{t}_i\ast pos{t}_t)+X_{it}^{\mathrm{\prime }}\text{β}+a_i+\tau +\omega +{\phi }_{it}+{ϵ}_{it}$(1)

The specification of the equation is the result of the approach proposed by Wooldridge (2014). The dependent variable is the natural logarithm of the monthly paper waste per person in municipality i in month t. θ is the constant; γ represents the coefficient for the variable treatment, a dummy variable for the treatment group Amsterdam. The coefficient of dummy variable post for the period after the treatment is noted as λ. The difference-in-differences estimator is δ; this is the variable of interest in this study. X_it’β is the vector notation of relevant control variables for municipality i in month t (see Section 3.4). The error term is given by ϵ_it. Furthermore, it was relevant that effects specific to the municipality (α_i,), month (ω) and year (τ) and were included in this study because these might explain variations in paper waste. A Hausman test indicated that the fixed effects estimation fit the data the best. Environmental preferences are both individual and time specific (Perni, Barreiro-Hurlé, and Martínez-Paz 2020). For this reason, an interaction of the individual and year fixed effect is included as well, following the approach of Yamazaki (2017). The municipality-fixed effects were made to interact with year dummies to capture the effect of municipality-fixed effects that change over time (φ_it). The municipality dummies interact with the year dummies instead of the month dummies because this interaction coefficient represents the long-term time trends of paper waste, which allowed for the assessment of the impact of the treatment over several years. Next, there is heteroskedasticity in the data, as observations for a municipality are not fully independent across time. To correct for heteroskedasticity, a wild cluster bootstrap method has been applied, since this method tends to give more reliable results compared to cluster-robust standard errors when there are relatively few clusters in panel data (MacKinnon and Webb 2020).

4.2. Sampling strategy and operationalisation

Data were obtained from nine of the 20 largest municipalities in the Netherlands,¹ including Amsterdam. The sample follows a monthly time frame from 1 January 2015 to 1 December 2019 and covers three years before the policy intervention in Amsterdam and two years after the intervention. The panel data are slightly unbalanced, as not all data are available for each month in every municipality. These missing values are problematic when the missingness is nonrandom and correlated with outcomes. The missing observations are 2015 for Rotterdam and Utrecht, 2018 for Groningen, 2015 and 2016 for Breda and the first half of 2015 for Arnhem. These municipalities stated that the data for these years were not available, as they started measuring their paper waste after 1 January 2015. Earlier data are, therefore, not available. There is currently no reason to believe that the fact that data are missing is correlated with any different paper waste per person in a municipality.²

Aggregated data for the monthly paper waste per person were retrieved by contacting the separate municipalities about their monthly paper waste. The measured waste was the total paper waste collected at the waste collection site of the municipality. This monthly paper waste includes paper waste from both households and businesses. A large part of this paper waste could, therefore, be part of the business packaging material, which potentially results in a smaller effect than when only household paper waste is measured. Paper waste was measured only as the total at the waste collection site of the municipality, making it impossible to disentangle household waste from business paper waste. Next, this monthly paper waste was divided by the number of residents of the municipality retrieved from the Dutch Bureau of Statistics (CBS 2020a). The change in paper waste likely follows a nonlinear path (Sharma and McBean 2007); the reasoning is that a reduction in paper waste has diminishing marginal returns, and paper waste declines at a much higher rate when paper waste is already high. Moreover, when paper waste is low, a reduction in paper waste can be a lot smaller. Including the dependent variable in the log is, therefore, a way to estimate the nonlinear effect the opt-in mail system is likely to have on paper waste. The control variables were included in the log as well to reduce the impact of outliers and ease the interpretation of the coefficients as elasticities.

4.3. Common trend assumption

The difference-in-differences estimators provide the effect of the adoption of the explicit consent stickers in Amsterdam if the common trend assumption holds. With the common trend assumption, it is assumed that the difference between the treatment and control group would be constant over time in absence of the treatment (Yamazaki 2017). Figure A.1 in Appendix I gives the time trend for monthly paper waste per person for all municipalities. The default nudge was implemented on 1 January 2018. It becomes apparent that there is a large heterogeneity in paper waste between municipalities, indicating that municipality fixed effects need to be applied in the regression. In addition, different municipalities follow different time trends with respect to their paper waste, which implies that municipality × year fixed effects may be of additional value in the regression (Yamazaki 2017; Mummolo and Peterson 2018). Monthly paper waste has been normalised using a logarithmic transformation in Figure A.2 in Appendix I. The average of the control groups has been compared to paper waste in Amsterdam. Prior to the policy intervention, the control and treatment group seem to follow a similar pattern, which shows strong seasonality, which motivates the inclusion of month fixed effects in the regression.

To control for such seasonality, Figure 1 shows the natural logarithm of the average monthly paper waste before and after the implementation of the green waste nudge. The red line indicates the development of the dependent variable in Amsterdam over time. It stands out that the average paper waste per person is already lower compared to the control group before the policy was implemented. The blue dotted line represents the time trend of the average paper waste in the control group. In a difference-in-differences design, it is assumed that the treated group (Amsterdam) would follow that trend absent of the treatment (Wooldridge 2014). The difference between this counterfactual and the observed outcome for Amsterdam is the treatment effect of the default nudge. It can be seen that the trend in both groups is slightly downwards sloping prior to the implementation of the default nudge, although this negative trend is stronger in Amsterdam. The policy intervention seems to amplify the negative time trend of paper waste.

Figure 1. Natural logarithm of the average paper waste per person for Amsterdam and the control group before and after the policy intervention.

According to Eggers, Tuñón, and Dafoe (2021), besides visual evidence, evidence for the common trend assumption can be shown by using a placebo test. ‘Fake’ treatment groups are created using a placebo test in periods before and after the intervention. This placebo treatment group is certain to be not affected by the programme. When comparing the placebo treatment group with the control group, one is thus expected to find no impact of the treatment. If impact has been found, there must be an underlying difference in the trend between these groups (Eggers, Tuñón, and Dafoe 2021). When the common trend assumption holds, the placebo treatment should not affect outcomes (World Bank n.d.). When there is a bias present in the data (i.e. differences between municipalities that result in differences in paper waste), it will become visible when treatment significantly affects the outcome variable (Angrist and Pischke 2008). Angrist and Pischke (2008) empirically performed this placebo test by using the lag and lead values of the treatment status. An interaction effect on the outcomes in the current period was estimated for the treatment status two periods ago and two periods in the future. The insignificance of these interaction terms provides additional evidence for the common trend assumption (Angrist and Pischke 2008). This test is shown in Table A.1 in Appendix I. The coefficients of the interaction terms of the lead and lagged values of the opt-in variable are not significant, as outcomes are not dependent on the treatment status in another period. This means that the treatment group does not significantly differ from the control group in other periods than when treatment is assigned. In summary, on the basis of the visual inspection in Figure 1 and the placebo test, we conclude that the common trend assumption holds, which makes it possible to estimate the causal effect of the default nudge using a difference-in-differences estimator.

4.4. Control variables

The literature has shown several determinants of the amount of paper waste in a municipality. The included control variables with their respective literature base and data source are included in Table 1. The descriptive statistics for the included variables can be found in Appendix II. Bach et al. (2004) argued that tourism had a positive effect on paper waste and used the number of overnight stays per person to measure this effect. Since data on overnight stays in municipalities were not widely available in the Netherlands, the number of tourist jobs per 1,000 residents was used as the proxy for tourism (LISA 2020). However, both proxies for tourism and population density strongly correlate with the independent variable of interest in this study. These variables cannot be included at the same time in the regression to avoid multicollinearity.³

Table 1. Definition and source of the included variables.

Variable	Definition	Literature base	Data source
kgwastepp	Total monthly paper waste in kilograms per person in a municipality		Directly retrieved from the respective municipalities, the population from CBS (2020a)
ln(kgwastepp)	Natural logarithm of the monthly paper waste per resident of the municipality	Sharma and McBean (2007)
municipality	The individual number for each municipality
post	Dummy variable for the period after 1/1/2018
year	Year
treatment	Dummy variable = 1 if the municipality has implemented the default nudge		VNG (2020)
ppbin	Recycling convenience, number of residents in a municipality per paper recycling container	Rosenthal and Linder (2021); Martin, Williams, and Clark (2006)	population/number of paper bins (directly retrieved from the respective municipalities)
pphousehold	The average number of persons per household	Khan and Burney (1989)	CBS (2020b)
popdenstiy	Population density	Khan and Burney (1989); Wowrzeczka (2021)	CBS (2020b)
medianworkerincome	Mean income of a worker in a municipality	Monavari et al. (2012)	CBS (2020b)
medianfamilywealth	Mean wealth of a household	Monavari et al. (2012)	CBS (2020b)
staysper1000	Number of overnight stays in a municipality per 1,000 residents	Bach et al. (2004)	CBS (2021a), CBS (2020a)
touristjobsper1000	Number of tourist jobs in a municipality per 1,000 residents	Bach et al. (2004)	LISA (2020), CBS (2020a)
farmsper1000	Number of farms in a municipality per 1000 residents	Bach et al. (2004)	CBS (2021b), CBS (2020a)

Finally, the population, population density, number of people sharing one paper recycling container, mean income and the median wealth of the municipalities were measured only on 1 January each year. Linear interpolation was applied to these variables to retrieve monthly observations. The main assumption is that the trend of these variables should be linear through the year. Sensitivity analysis examines how excluding the interpolated variables influences the main results. The Ramsey RESET test (Ramsey 1969) was performed on the model with the covariates included, and the test indicated that there was no indication of an omitted variable bias or misspecification of the functional form.

4.5. Monetising environmental benefits

To estimate the social benefits of a green policy, the environmental benefits of the possible reduction in paper waste should be monetised first. Only limited literature is available on the social costs of paper waste. Hence, they are approximated by the private and social costs of municipal solid waste, assuming that the social costs for paper waste are similar to the social costs for solid waste in general. Environmental costs for waste can be divided into private costs and social costs (Kinnaman 2009). Private costs are the costs for the municipality to collect and transport waste, which are incorporated into municipal tax (Niaounakis and Blank 2017).

Private marginal costs of waste can vary between $70 per tonne in the United States and $200 per tonne in more densely populated areas (Kinnaman 2009). Kinnaman (2009) used the Netherlands as an example of a densely populated country with private marginal costs of waste collection of approximately $200 per tonne or €166.28, using the latest exchange rates (OECD 2021). Moreover, external costs of waste such as carbon dioxide emissions or environmental degradation are typically not incorporated into the price of waste collection (Kinnaman 2009). The external costs of solid waste are between $5.38 and $8.76 per tonne of waste (Kinnaman 2006) or between €4.48 and €7.29, using the latest exchange rates (OECD 2021). Kinnaman (2006) estimated these external costs by comparing several studies on attributable greenhouse emissions for waste disposal.

Another study performed by Dijkgraaf and Vollebergh (2004) examined the external costs of waste in the Netherlands by estimating differences between the costs of incinerating waste or landfilling it. Variables they included in their calculation were emissions of greenhouse gases, energy recovery from incineration, harm to water and soil systems and chemical pollution. They valuated external costs for a tonne of a landfill at €3.54 and €14.97 for incineration but assumed limited possibilities for recycling. This valuation might lead to upwards biased estimates, as paper waste also has a material value because it can be recycled. Kinnaman (2009) calculated that the per-bag fee should be approximately $2.15 (€1.79) in the Netherlands. This fee includes both private and social costs of waste and assumes that a bag of waste weighs 20 pounds (approximately 9 kg), meaning that the costs of a tonne of municipal waste should be [1000/9 × 1.79  = ] €198.88.

A potential downside of this calculation is that this calculation is made for all types of waste and not just paper waste. Paper waste is typically recycled more than other kinds of waste, neglecting the material value of paper waste and leading to an upward biased estimate of the price of a tonne of paper waste. Since some paper waste can be recovered and reused, the costs of paper waste might be lower than the costs of general waste. However, recycling also has its costs. Samakovlis (2003) found that paper waste and electricity were complementary goods, meaning that higher use of paper waste resulted in more energy consumption. This result may be attributed to the relatively energy-consuming process of recycling paper waste. Furthermore, the recycling of paper waste also comes with other environmental costs, such as carbon emissions from the collection, transport and processing. A detailed explanation of the calculation of environmental benefits of paper waste reduction is presented in Appendix IV.

5. Results

5.1. Estimating the effect of the default nudge

The estimation of the treatment effect is shown in Table 2 for five models. First, Models 1 and 2 include only municipality-fixed effects. Models 3 and 4 also include year and municipality × year-fixed effects to correct for the location-specific effects that change over time. Model 5 is added as robustness test; month dummies are added to the model to check whether paper waste is determined by the time of the year.

Table 2. Fixed effects regression on the natural logarithm of the monthly paper waste per person.

Variables	(1)	(2)	(3)	(4)	(5)
	Model 1	Model 2	Model 3	Model 4	Model 5
post	−0.089***	−0.066**	−0.168***	−0.050	−0.070
	(0.023)	(0.025)	(0.003)	(0.128)	(0.349)
treatment	−1.044***	−1.048***	−1.077***	−0.708***	−1.666***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
post × treatment (DID)	−0.054**	−0.052*	−0.027***	−0.022	−0.172*
	(0.023)	(0.026)	(0.000)	(0.062)	(0.084)
ln(ppbin)				0.199	−0.172
				(0.107)	(0.105)
ln(popdensity)				0.204	1.024***
				(0.398)	(0.115)
ln(medianworkerincome)				2.108	−1.215
				(2.941)	(4.434)
ln(medianfamilywealth)				−0.189	−0.044
				(0.106)	(0.117)
ln(farmsper1000)				0.238	−0.037
				(0.151)	(0.032)
Constant	1.657***	1.761***	1.840***	−21.922	8.149
	(0.015)	(0.027)	(0.020)	(30.013)	(46.092)
Observations	474	474	474	438	438
Adjusted R-squared	0.167	0.428	0.681	0.547	0.738
Municipality fixed effects	X	X	X	X	X
Year fixed effects		X	X	X	X
Month fixed effects		X	X	X	X
Municipality × year fixed effects			X		X

Note: Robust standard errors in parentheses.

***p < 0.01.

**p < 0.05.

*p < 0.1.

The main explanatory variable of interest is the interaction effect between the post dummy and the treatment dummy, which is also referred to as the treatment effect or the DID estimator. Only the main independent variable is regressed on the natural logarithm of the monthly paper waste per person with municipality-fixed effects in Model 1. The default nudge significantly causes a reduction of 5.4% (p < 0.05) in the average paper waste per person in a municipality. However, the Adjusted R-squared is quite low in this case, where the model only explains 16.7% of the variation in the dependent variable. In Model 2, year and month fixed effects are added to control for the decreasing time trend observed in paper waste and the strong seasonality in paper waste. The treatment effect remains significant and is of similar size compared to Model 1, where the R-squared improves to 0.428. However, when year- and municipality × year-fixed effects are added in Model 3, the fit of the model substantially improves. The coefficient of the treatment effect is somewhat smaller compared to Model 1 and 2 and becomes significant at the 1% level, as the fixed effects reduce noise from the data by controlling for municipality specific time trends, such as environmental preferences (Yamazaki 2017; Mummolo and Peterson 2018).

Model 4 is the same as Model 2, only that several control variables are added. The DID estimator is no longer significant in this model, but when municipality × year fixed effects are added in Model 5, the coefficient becomes significant again. This coefficient is −0.172, which is larger compared to the previous models. Model 5 explains the largest fraction of variation in the monthly paper waste in a municipality. Monthly paper waste is 17.2% lower compared to other municipalities that did not implement the default nudge, ceteris paribus.

Other significant variables between the different specifications are the post variable, which indicates that monthly paper waste is on average lower after the treatment in all municipalities. This confirms the findings in Figure 1. The treatment variable is significant in all models, showing that monthly paper waste is on average lower in the municipality where the nudge has been implemented compared to the control municipalities. Again, this confirms the findings from Figure 1, where it is shown that Amsterdam already had lower paper waste prior to the policy intervention. Finally, population density is significant in Model 5, where a 1% increase in population density is associated with a 1.024% (p < 0.01) increase in paper waste, keeping all the other variables constant. This effect is in line with the effect Bach et al. (2004) and Wowrzeczka (2021) found in their study.

To test the robustness of the results, an additional analysis has been performed in Table 3. In these Models, the time window of the analysis has been reduced to 6 months or 12 months before and after the policy intervention, to potentially reduce the effect of unrelated events in the past or future that drive paper waste. In Models 1 and 2, the time frame of the analysis has been reduced to 6 months before and after the policy intervention, which leads to an effect size in the range of the found coefficients in Table 2. Despite the substantial reduction in the number of observations compared to Table 2, the coefficient of the DID estimator remains significant in Models 1 and 2. The time frame for Models 3 and 4 is 12 months before and after the policy intervention. These coefficients of the DID estimator are similar to Table 2 and insignificant when municipality × year fixed effects are left out of the analysis, but significant when they are included. Moreover, the adjusted R-squared of Models 1–4 in Table 3 is higher compared to the same specifications in Models 4 and 5 in Table 2. This outcome is remarkable, as the number of observations is more than halved. Reducing the time-frame of analysis, therefore, reduces the impact of unrelated events in the treatment and control group on paper waste. These findings contribute to the robustness of our results.

Table 3. Fixed effects regression on the natural logarithm of the monthly paper waste per person where the time period is reduced to 6 months (Models 1 and 2) or 12 months (Models 3–6) before and after the policy intervention. Models 5 and 6 only include control municipalities with similar paper waste levels as Amsterdam.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	6 months	6 months	12 months	12 months	12 months	12 months
	all	all	all	all	similar	similar
post	0.173	−0.043	−0.047	−0.251	−0.015	−0.005
	(0.146)	(0.090)	(0.071)	(0.135)	(0.035)	(0.119)
treatment	−2.739***	−3.094***	−1.953***	−1.978***	1.136***	−0.138
	(0.204)	(0.156)	(0.155)	(0.128)	(0.145)	(0.171)
post × treatment (DID)	−0.087**	−0.110***	−0.033	−0.053*	−0.032	−0.052**
	(0.029)	(0.006)	(0.043)	(0.026)	(0.030)	(0.009)
ln(ppbin)	−0.354	−0.985**	−0.035	−0.564**	−0.042	−0.084
	(0.352)	(0.314)	(0.155)	(0.211)	(0.132)	(0.196)
ln(popdensity)	2.336***	2.255***	1.223***	0.939***	5.291***	−0.335
	(0.122)	(0.062)	(0.089)	(0.124)	(0.778)	(0.916)
ln(medianworkerincome)	−9.127*	−8.486***	−0.035	2.316	−4.473	−2.567
	(4.060)	(1.286)	(2.747)	(2.898)	(3.683)	(7.221)
ln(medianfamilywealth)	−0.299	0.332	−0.124	0.281	−0.037	0.038
	(0.437)	(0.247)	(0.175)	(0.230)	(0.130)	(0.102)
ln(farmsper1000)	−0.016	−0.118***	0.098**	0.136	0.008	0.059
	(0.070)	(0.032)	(0.041)	(0.077)	(0.023)	(0.096)
Constant	86.316*	78.687***	−5.823	−28.624	2.916	31.543
	(41.695)	(13.455)	(28.446)	(30.304)	(36.174)	(75.535)
Observations	104	104	192	192	132	132
R-squared	0.766	0.782	0.724	0.745	0.683	0.703
Municipality fixed effects	X	X	X	X	X	X
Year fixed effects	X	X	X	X	X	X
Month fixed effects	X	X	X	X	X	X
Municipality × year fixed effects		X		X		X

Note: Robust standard errors in parentheses.

***p < 0.01.

**p < 0.05.

*p < 0.1.

Moreover, in Models 5 and 6, only municipalities with paper waste levels similar to Amsterdam are included as control group.⁴ This way, an artificial control group has been created which may be more similar to the treatment group. Additionally, this approach also helps testing whether the missingness of some observations is random, as described in Section 3.2. As the coefficient size of the DID estimator does not change between Models 3 and 4 as opposes to Models 5 and 6, we conclude that our results are robust to creating an artificial control group.

5.2. Estimating social benefits

The two models from Tables 2 and 3 with the highest adjusted R-squared find a treatment effect between −0.053 and −0.11, which indicates an average reduction in paper waste between 5.3% and 11% as result of the default nudge. In the old system, 55% of all households received unaddressed mail, while this is 13.8% in the new system (Schmitz 2018). As approximately 75% all paper waste is being separated (Milieu Centraal n.d.) and 30–50% of all paper waste consists of unaddressed mail, we believe that these effect sizes are reasonable.⁵ A 5.3–11% reduction in paper waste in Amsterdam indicates that the default nudge has reduced paper waste with approximately 1–2 million kg per year (Appendix IV). Applying the approach from Appendix IV, we find that the annual social benefits of implementing the nudge in Amsterdam are between €135,000 and €285,000. These social benefits include the reduced collection, transportation and disposal costs for paper waste and carbon emissions that come along with them. The environmental costs of recycling paper are also included in this estimation. Costs of implementing the opt-in mail nudge consist of the costs of communicating the change, enforcing the new system and possibly distributing the new explicit consent stickers and, in the case of Amsterdam, the legal costs to implement the system (van Zoelen 2017). These legal costs were incurred because local entrepreneurs protested against the opt-in mail system and decided to take the municipality of Amsterdam to court. In 2021, the Dutch Supreme Court decided that the default mail nudge is legitimate (Dutch Supreme Court, 2021). Other municipalities that want to implement the default nudge do not face these legal costs. When applying these findings to the entire Netherlands, annual benefits would be between €14 million and €30 million (Appendix IV).

6. Discussion

6.1. Discussion of the main findings in relation to the literature

This study attempted to find the effect of the default nudge Amsterdam implemented on 1 January 2018 by performing a difference-in-differences estimation. Under the common trend assumption, the default nudge caused a decrease of 5.3% to 11% in paper waste in Amsterdam. The found effect is robust to selecting a smaller control group and reducing the time frame of the regression. The adoption of the explicit consent stickers in Amsterdam resulted in annual social benefits between €135,000 and €285,000. Moreover, the costs of the implementation of the opt-in mail system nudge are low, as the municipality has to only communicate the system change and possibly print new explicit consent stickers to put on mailboxes (VNG 2020). It is possible that the costs of printing the stickers will be lower than those during the period of the system of explicit refusal. When fewer people need the explicit consent stickers rather than the explicit refusal stickers, the costs of the new stickers will be lower. However, the findings related to the monetary value of the avoided paper waste should be interpreted with caution. Up to our knowledge, the literature only provided estimates of the costs of waste in general instead of just paper waste. Using recycling and recovery rates for paper, we came to an approximation of the monetary value of paper waste. The uncertainty on the monetary value was reduced by providing an upper and a lower bound for the social costs of paper waste and validating the estimation with experts from the municipality of Amsterdam.

The contribution of this study to the literature is that it examines a practical application of a nudge on paper waste and to test this nudge in practice. There are two similar studies that have examined the effect of a nudge in the paper waste domain with other applications. First, Egebark and Ekström (2016) tested the impact of a double-sided printing default on university printers. Liebig and Rommel (2014) examined the impact of a nudge for the actual adoption of new explicit consent stickers instead of looking into the reduction in paper waste as a result of the default nudge. Furthermore, our study contributes to the existing literature on green nudges by estimating the outcome of a green nudge by performing a natural experiment rather than an economic laboratory experiment, as the nudge is tested in practice. This study on the default nudge leads to a higher external validity of the results than the studies of Liebig and Rommel (2014) and Puig-Ventosa, Jofra-Sora, and Freire-González (2015) because the difference-in-differences design of this study was performed in a less controlled setting.

This study finds an effect that is larger than those in other studies on green nudges. Puig-Ventosa, Jofra-Sora, and Freire-González (2015) examined the effect of a ‘no junk mail’ sticker in Barcelona, which resulted in a decrease of 0.4% in paper waste. This is substantially lower than the found effect in this study. However, Puig-Ventosa et al. examined explicit refusal stickers within a system of presumed consent for unaddressed mail, while this study investigated explicit consent stickers within a system in which unaddressed mail is generally forbidden. It can, therefore, be concluded that an opt-in mail system is more effective in preventing paper waste. Next, Akbulut-Yuksel and Boulatoff (2021) found that the clear plastic bag nudge in Canada resulted in a decrease of 27% in plastic waste. The nudge investigated by Rivers, Shenstone-Harris, and Young (2017) resulted in an increase of 3.4% in the use of reusable plastic shopping bags.

However, the outcome variable of paper waste on the municipality level indirectly measures the impact of the nudge. It is possible that spillovers occur between municipalities. Potentially, the distributor of unaddressed mail may have distributed all remaining folders in other municipalities, thus minimising the effect of the default nudge. We do observe a reduction in paper waste in Amsterdam, while the trend in the control municipalities remains constant (Figure 1). However, these control municipalities do not directly border on Amsterdam, so the possibility that the remaining unaddressed mail will be distributed in neighbouring municipalities remains. In an attempt to overcome this issue, the main distributor of unaddressed mail has been contacted, but indicated that production data were not available. Although insight in the production levels of unaddressed mail would have been of additional value, we do observe signals that distributors actually did downscale their production (Van Gerrevink, Uitgeverij Steden Driehoek, and All-In Verspreidingen 2019; van Zoelen 2018).

6.2. Policy implications

The policy implications of these findings are that an opt-in mail nudge can be added to the range of policy tools used to increase a municipality’s engagement in the circular economy and reduce waste. As other countries, such as Germany and the United States also have a system of explicit refusal (e.g. Liebig and Rommel 2014; Standley 2019), international implementation of this policy will result in a large reduction of paper waste, and therefore, global environmental pressure. The opt-in mail nudge is a relatively affordable way to decrease the paper waste of a municipality. Furthermore, private benefits of the nudge are lower garbage collection costs for the municipality, possibly resulting in lower municipal taxes. Social benefits are lower carbon emissions for producing paper, transport and recycling paper waste. The opt-in mail nudge offers municipal policymakers an affordable tool to reduce paper waste. Moreover, if the default nudge is implemented across the Netherlands, the social benefits will approximately be between €14 million and €30 million per year, where global benefits will be even larger.

Finally, the goal of the nudge needs to be clearly communicated and transparent to guide residents in making their own choice (Schubert 2017). The municipality outlined the benefits of the new system and the consequences of making a certain choice regarding unaddressed mail (City of Amsterdam 2021). Schmitz and Slot (2018) examined how many people were aware of the system change in Amsterdam. They found that 93% of all residents of Amsterdam were aware of the new opt-in mail system in 2018, meaning that the information campaign of the municipality accomplished its goal. The municipality of Amsterdam faced resistance from local entrepreneurs not able to reach their customers anymore (van Zoelen 2017). The municipality may step up to assist these entrepreneurs in engaging with their customers. Residents generally supported the opt-in mail nudge (van Zoelen 2017). Therefore, the default nudge contributes to the introduction of a circular economy in Amsterdam, avoiding unnecessary waste.

7. Conclusion

Green nudges are increasing in popularity, as they offer low-cost options to steer behaviour towards environmentally friendly choices. The effectiveness of these nudges differs depending on the kind of nudge and context. For this reason, empirical evaluations are required to identify effective nudge policies. This study estimated the effect of a default nudge that changed the system of explicit refusal to a system of explicit consent for unaddressed mail in Amsterdam by using a difference-in-differences design. Up to our knowledge, the impact of this particular green nudge has not been studied before. Moreover, our analysis monetises the benefits for society. This study differs from other studies on green nudges because of the difference-in-differences methodology, a design that tests the nudge in practice and, therefore, yields more externally valid results than other experimental settings.

Two main findings emerged from our analysis. First, the adoption of the default opt-in nudge significantly reduces paper waste with 5.3% to 11% compared to a situation where no nudge is implemented. Second, this reduction in paper waste yields important societal benefits. In particular, the yearly social benefits are between €135,000 and €285,000 in Amsterdam. When all the municipalities in the Netherlands change the presumed consent to explicit consent for unaddressed mail, as is currently possible due to a favourable court ruling, the annual social benefits will be between €14 million and €30 million. Currently, the most common system for unaddressed mail is the system of explicit refusal, which implies that introducing the system of explicit consent will result in large reductions of paper waste globally. The main conclusion is that the default nudge that changed the system of explicit refusal to a system of explicit consent for unaddressed mail is effective in both reducing paper waste and yielding societal benefits. A policy recommendation that follows from these results is, therefore, to expand the adoption of the default nudge to a global context, as other countries still use the explicit refusal system. Future research might examine the costs of the implementation of the green nudge in more detail to fully perform a cost–benefit analysis. Finally, future studies can use unaddressed mail as outcome variable and analyse the extent to which the observed effects of the nudge can be generalised to other municipalities than Amsterdam.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the REACHOUT project. The REACHOUT project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 101036599.

Notes

1 Groningen, Rotterdam, Utrecht, Breda, Tilburg, Zoetermeer, The Hague and Arnhem function as the control group. Eleven of the 20 largest municipalities did not have sufficient data on paper waste to be included.

2 The representative of the municipality of Groningen mentioned that paper waste in Groningen in 2018 was wrongly measured due to a miscommunication between the municipality and the waste disposal company. The miscommunication is exogenous to the paper waste in Groningen and is, therefore, also random. A sensitivity test in Table 1 concluded that the missingness of outcomes is random and does not bias the estimates.

3 Correlation coefficient larger than 0.8 (Franke 2010). See Appendix III.

4 Rotterdam, The Hague and Utrecht are included as control group, as they show similar paper waste levels prior to the implementation of the green waste nudge in Amsterdam (Figure A1 in Appendix I).

5 For example, when a household receives 100 kg of paper waste. 75 kg would be recycled and, therefore, measured. 30% to 50% of all paper waste is unaddressed mail, which implies 22.5 to 37.5 kg of unaddressed mail. Due to the default nudge 41.2 percentage points fewer households received unaddressed mail, which would imply an average paper waste reduction between 9.27 and 15.45 kg. Our estimates of a 5.3% to 11% reduction overlap with this range.