“Job change and self-control of waste pickers: evidence from a field experiment in the Philippines”

Environmental policies may have a negative side effect on employment, often in a specific industry in the short run. Workers in regulated industries can be affected by losses in job-specific human capital. The informal sectors in developing countries are often associated with environmental pollution and thus targeted by such policies. Welfare loss due to this side effect can be problematic in developing countries, since they often lack safeguarding schemes, including unemployment insurance. Inducing workers in informal sectors to change their jobs can mitigate these negative side effects. This study examines efficient methods of inducing informal workers to change jobs. An alternative job is offered to informal workers at a dumpsite in the Philippines and whether changing the scheme of wage payment increases the acceptance of the offer is examined. The impacts of changing payment schemes are evaluated by using a randomized field experiment. The sampled 112 waste pickers each randomly receive one of four offers for an alternative job, and the number of those who accept the offer is observed to evaluate the impact of less frequent payment (i.e., once every three days instead of daily). Piece rates and fixed wages are also compared. Those offered less frequent payment are more likely to accept the job offer compared with those offered daily payment. This preferred payment scheme can mitigate the side effects of environ mental policy and workers’ self -control problem related to savings, while minimizing moral hazard.


Introduction 2
Increasing environmental regulation may have a negative side effect on employment, often in a specific industry in the short run. Workers in regulated industries may be affected by job losses and/or losses in job-specific human capital. Job losses can be considered to be the costs of environmental policies if the affected workers suffer from costly job transitions such as decreased income and high adjustment costs for a new job (Walker, 2011). Further, welfare loss due to this side effect may be more problematic in developing countries, since they often lack safeguarding schemes, including unemployment insurance. The informal sectors in developing countries are often associated with environmental pollution and thus targeted by such regulations (Blackman & Bannister, 1998;Blackman et al., 2006). Inducing workers in informal sectors to change their jobs can mitigate the negative side effects of environmental regulation in many cases. Therefore, efficient methods to induce job changes by informal workers have gained policy prominence.
A common example of informal workers in many developing countries are waste pickers, namely individuals who collect recyclable waste at an open dumpsite (Hayami et al., 2006;Wilson et al., 2009) 1 . 3 Open dumping is a widely used waste disposal practice in developing countries. This practice causes several environmental problems such as bad smells and dirty leachate. Moreover, once this practice has been employed, the land cannot be used for other purposes such as agriculture for a long time, which means a real economic loss. Thus, agencies operating waste management services (e.g., municipalities) may plan to close open dumpsites to decrease those costs. However, changing to a more environmentally friendly practice such as sanitary landfilling would inevitably harm waste pickers (Medina, 2000;Paul et al., 2012). To minimize the social cost of solid waste including the negative side effect of environmental management, it is therefore important to safeguard waste pickers by providing alternative job opportunities before the closure of open dumpsites and inducing them to change their jobs (Medina, 2000).
This study empirically examines some such methods of inducing informal workers to change their jobs. In this study, an alternative job is offered to waste pickers and whether changing the scheme of wage payment increases the acceptance of the offer is examined. A field experiment is conducted at a dumpsite in Iloilo City, the Philippines, where 112 waste pickers are offered an alternative job. Each individual randomly receives one of four types of offer letters and the number of those who accept their offer is observed.
The experiment has three objectives. Based on evidence of demand for commitment devices (Bryan et al., 2010), including the finding by Ashraf et al. (2006) that individuals who exhibit timeinconsistent preferences in a survey are more likely to take up a commitment savings product that restricts customers' access to their funds, the first objective of the experiment is to evaluate the extent to which less frequent payment (i.e., payment every three days) affects the acceptance of a job offer by informal workers.
The job offered in the experiment is the production of solid fuels, and an important feature of this production is the easy measurement of the output. According to the theory of optimal compensation (Lazear, 1995(Lazear, , 1998, if the measurement of the output is costless, then piece rate pay with a price equal to the marginal cost achieves an efficient level of effort by a worker and minimizes moral hazard, whereas fixed wages do not. The second objective of the experiment is to compare the extent to which piece rate pay and fixed wages affect the acceptance of the offer. Several studies empirically examine the relationship between compensation systems and labor supply (Prendergast, 1999). For example, Shearer (2004) estimates the gain in the productivity of workers in a tree-planting firm in Canada when they are paid piece rates rather than fixed wages, finding that productivity is higher for workers who receive piece rate pay. While existing studies focus on the impact of changing the compensation system on labor supply conditional on job choice, however, the present study compares the effects of piece rates and fixed wages on occupational choice.
Finally, although piece rate pay can provide an incentive for workers to provide their maximum effort, an employer cannot raise its profit as long as the price of the output equals the marginal cost. For instance, Lazear (1998) provides a piece rate pay scheme with a fixed salary (i.e., a draw) to raise profit, while maintaining an incentive for workers. The third objective of the present experiment is thus to investigate whether piece rate pay with a draw, which is beneficial for an employer, induces the acceptance of the offer, as well as pure piece rate pay does.
The results show that changing payment frequency from daily to once every three days increases the probability that an individual accepts a new job offer, which is counterintuitive to the monotonicity assumption of the discount rate. Using piece rate pay paid once every three days, 27 percent of the recipients are induced to accept the offer. This acceptance rate is slightly higher than the fixed wage of around half of the average daily earnings in collecting waste, which is paid once every three days. There is no positive or negative effect for adding a draw, implying that it may be possible to raise profit without decreasing the acceptance rate.
The findings of this study make several contributions to the field of environmental and resource economics. First, this study extends the literature on environmental policies and unemployment as its side effect (Bovenberg & van der Ploeg, 1994). Several empirical studies find evidence of such a side effect (Greenstone, 2002;Walker, 2011Walker, , 2013 4 For empirical studies of the informal sector, the collection of micro data is challenging. For example, many waste pickers do not have mailing addresses or personal identification. Furthermore, the population of waste pickers working at a dumpsite is difficult to identify. Therefore, random sampling is difficult (Hayami et al., 2006). Owing to these difficulties, few econometric analyses have been carried out to investigate the association between informal sectors and environmental management. An exception is the seminal work by Blackman and Bannister (1998) who use primary data on 76 informal brick makers in Mexico to study the determinants of the adoption of clean fuel. While there are a few empirical studies of the behavior of informal workers, all rely on nonexperimental data. Of particular concern when using nonexperimental data is the possible endogeneity problem. One way in which this problem can be solved is to use data from a field experiment. Although the sample size of the experiment is relatively small and careful consideration of the external validity of the results is required, the internal validity of the results is guaranteed, since treatments are randomly assigned to informal workers.
The remainder of this paper proceeds as follows. Section 1 describes the experimental context and design. Section 2 presents the empirical results. The last section discusses the limitations of the study and concludes.

Study setting.
In Iloilo City in the Philippines, a private contractor collects approximately 220 tons of municipal solid waste daily using trucks and hauls it to the Calajunan-controlled disposal facility (Paul et al., 2012). Since there is no household recycling program in Iloilo City, all waste is mixed together. Because of these conditions, which are common in many developing countries, including the Philippines, informal workers gather at dumpsites to collect recyclable waste and sell it to nearby junk shops.
As the random sampling of informal workers is difficult, the true population of the waste pickers working at this dumpsite is unknown. In 2010, the German International Cooperation (GIZ) estimated that there were approximately 300 waste picking households. This estimate suggests that the size of the informal sector at this dumpsite is not very large, making it plausible to conduct a census. In this regard, Iloilo is an ideal research site.
The preliminary survey identified several features associated with the job of a waste picker in Iloilo City. First, waste pickers are not employed (see also Birkbeck, 1978), and they act individually 3 . 5 Second, the self-employment nature of this job suggests that waste pickers work whenever they want, and they can even bring their children to their workplace. Third, in general, they earn cash directly from the junk shops. In some cases, they even borrow money from junk shops, which can be interpreted as a patron-client relationship (Gill, 2007). Fourth, workers can obtain payment daily. Fifth, the earnings of each worker are determined by his or her output, which depends on the quantity and quality of sellable waste collected. This is akin to a piece rate scheme. Sixth, roughly half of workers had another job or business before they ventured into waste picking and more than 30 percent of those were farmers or fishers. Finally, the working conditions of waste pickers are low; they have to work outside on a hill of waste, where the smell is bad and flies are crawling elsewhere. Furthermore, in tropical countries such as the Philippines, the high temperature and intense heat of the sun take their toll on the health of workers.
In 2006, the local government of Iloilo City started several programs to support the livelihoods of waste pickers. These programs are assisted by the GIZ, Central Philippine University (CPU), and a nongovernmental organization, Love Our Own Brethren. Their objectives are to support workers at the dumpsite and help them find an alternative source of livelihood (Paul et al., 2012). One alternative jobs that the project has started to provide is the production of solid fuels made from wastepaper (i.e., paper briquettes). (Romallosa & Kraft, 2017). The wastepaper generated at CPU along with procured sawdust and carbonized rice husks are used as inputs for this fuel source. This product can be used as fuel for cooking and can serve as a substitute for charcoal and wood. In 2013, CPU decided to recruit new workers to engage in paper briquette production for three weeks. A randomized experiment was conducted at this time to examine whether changing the payment scheme from the baseline offer would increase the number of waste pickers who accepted the offer.
In addition to providing alternative livelihoods, the city government and GIZ developed an association of dumpsite waste pickers with optional membership. This association works as a workers' union, providing social security and personal identification, as well as buying recyclable waste to sell together. Almost 33 percent of those interviewed sell their recyclable waste to this association. The executives of the waste pickers' association supported the authors during their experiment. Furthermore, the association's center was used for the paper briquette production. This center is located approximately 100 m from the main gate of the dumpsite, and therefore the commuting cost for waste pickers who decide to pursue this job is negligible. Notes: The piece rate pay scheme provides 1 PHP per four pieces of production. The fixed wages scheme provides 65 PHP per day. The piece rate pay with a draw scheme provides a draw of 30 PHP per day and 1 PHP per four pieces of additional production greater than 150 pieces. .

Treatments.
Four types of wage payment schemes for paper briquette production are prepared. Table 1 summarizes the four groups of offer letters. Piece rate schemes with daily payments is our first offer (PR-Daily). In paper briquette production, employers can easily count the pieces produced by one worker per day, or even an hourly rate.
According to the theory of optimal compensation (Lazear, 1995(Lazear, , 1998, if the measurement of outputs is costless and the production cost (e.g., expenditure for materials) is variable, the piece rate equal to the price of the output minus the marginal cost of materials will achieve an efficient level of effort by a worker and minimize moral hazard. Based 6 on this theory, a piece rate scheme, namely 1 Philippine Peso (PHP) per production of four pieces of paper briquettes, is considered 4 . Following evidence of demand for commitment devices, it is hypothesized that waste pickers face a self-control problem related to savings and would prefer payment less frequently than every day. The second proposal (PR-Three) thus offers payment once every three days (i.e., twice each week). It is hypothesized that decreasing payment frequency from every day to once every three days increases the number of workers accepting the offer.
The third proposal (FW-Three) offers fixed wages of 65 PHP per day regardless of how many paper briquettes are produced. Based on the technical team's experience of using old equipment, the productivity of workers is expected to be around 300 pieces per day (five to six working hours). This estimation is used to determine the fixed rate. Note that the preliminary survey found that 15 percent of a waste picker's daily earnings is equal to or less than 65 PHP. This third offer also pays once every three days.
Finally, a combination of piece rates and fixed wages is considered. Although the former can provide an incentive for workers to provide their maximum effort, an employer cannot raise its profit since the price equals the marginal cost. Following Lazear (1998), a draw is thus added to the second proposal in the fourth proposal (PR-draw-Three). Workers are guaranteed a draw of 30 PHP per day, while no commission is paid until a worker produces more than 150 pieces. Once this threshold is achieved, an additional 1 PHP per four pieces is paid on top of the 30 PHP already earned. This compensation system can produce a profit of 7.5 PHP per worker per day if a worker produces more than 150 pieces. The PR-Daily and PR-Three groups are compared for objective 1 (see Table 1). The PR-Three and FW-Three (PR-draw-Three) groups are compared for objectives 2 and 3. Figure 1 shows the experimental design and sample size in each stage of the project. In August 2013, the team of researchers attempted to interview all adult (18 years and older) waste pickers working at the dumpsite in Iloilo with the assistance of four enumerators over a 10-day period. As the purpose is to offer an alternative job to adult waste pickers, child waste pickers are excluded from the analysis. In addition, those already involved in existing programs (e.g., handicraft manufacturing) are excluded. Envelopes containing the job offer letters were randomly distributed to waste pickers over four days in November 2013. First, the team found and identified a waste picker who was interviewed. Second, four cards were shuffled and shown to the individual face down. Third, the individual picked one card and, finally, the number on the card determined the job offer letter they received. Only one version of the offer letter was given to each waste picker.

Fig. 1. Experimental design and sample sizes
The team searched and tried to identify all those interviewed in August. However, identifying them was difficult. Finally, the team identified 112 waste pickers and distributed one envelope to each of them. Admittedly, the experimental subjects were not randomly selected. The association's center was used as the base of operations, and the search for waste pickers by the team always started from that location. The vice president of the waste pickers' association supported the researchers throughout this process. This does introduce bias into the sample selection, as it favors waste pickers who often work near the center or who live near the area, suggesting that caution is needed when interpreting the results of the study.
The offer letter constituted two pages. The first page described the job including the starting time, working time, and application process, while the second page set out the substance of the offer. The letters were written in Hiligaynon, the local dialect spoken in Iloilo. The research assistants explained the description of the offer in Hiligaynon if requested.
All workers were asked to join for three weeks from November 27, 2013, except for Sundays. Working hours were set to five hours, from 10 am to 4 pm (excluding a one-hour lunch break). A three-week contract is short for an alternative job. However, paper briquette production by informal workers is in a pilot phase compared with other livelihood programs such as handicraft manufacturing and the production of alternative fuels for cement manufacturing (Paul et al., 2012). Nevertheless, waste pickers expected the acceptance of this offer to lead them toward alternative livelihoods in the long run.
The offer letter mentioned the availability of more than 16 vacancies. An application form was attached to the offer letter. Those who wanted to accept the offer were instructed to fill out the application form, submit it to the association's center, and then attend the orientation scheduled on November 27. Accepting the offer was defined as an individual who attended the orientation. Those who accepted the offer and started producing paper briquettes received a salary from the second day according to their production rate and payment scheme. From November 29, paper briquette production commenced. Table A1 in the Appendix compares the 112 subjects with those interviewed, but not included in the experiment. Average hourly earnings are higher for the participants of the experiment than nonparticipants. In participant households, an income earner other than the subject is less likely to exist. Both average participants and nonparticipants are risk averse. Importantly, however, nonparticipants are more risk averse than participants. Based on these results, the external validity of the findings is discussed in section 2.  Notes: Means are reported. Standard deviations are in parentheses. Two observations in group 1 and one observation in group 3 lack data in the Willingness to accept closure category. Discount rate is assessed on a scale from one to six, with higher values meaning the discount rate of an individual is high and the individual is impatient. Risk aversion is assessed on a scale from one to eight, with higher values meaning an individual is highly risk averse. A risk-seeking individual would receive a value between one and three, a risk-neutral individual would receive a value of four or five, and a risk-averse individual would receive a value between six and eight. Willingness to accept closure measures the stated opportunity cost (in PHP) of closing the dumpsite for a month in December 2013. Table 2 presents the summary statistics of the four groups of the participants of the experiment, as well as the treatment balance for participants. Comparing the variable means across PR-Three and the other three groups using t-tests shows that only two variables differ at p < 0.10. Participants assigned to PR-Three are slightly more productive in the current business than those assigned to FW-Three and slightly more risk averse than those assigned to PR-draw-Three.    Table 3 shows the estimation results of equation (2). In columns 1-3, the marginal effects are calculated by setting all the independent variables to the mean values. The results show a significant and robust negative effect of daily payment, meaning that changing payment frequency from daily payment to once every three days increases the probability of accepting the job offer. The effect of changing from piece rates to fixed wages of 65 PHP per day is negative, although the result is not robust. The effect of adding a draw of 30 PHP until 150 briquettes are produced is not statistically significant.

Impact of changing payment schemes
To precisely estimate the magnitude of the effect of changing from PR-Three to the other three payment schemes, the marginal effects are calculated with the model that includes the four demographic variables that are significant in column 3, setting those to their mean values; the payment scheme variables not of interest are set to zero. Columns 4-6 of Table 3 show the results, highlighting that changing from PR-Three to PR-Daily decreases the probability of accepting the job offer by 30.4 percentage points (column 4). Further, changing from PR-Three to FW-Three decreases the probability of accepting the job offer by 22.5 percentage points (column 5). There is no significant effect of changing from PR-Three to PR-draw-Three (column 6). Table 3 shows the correlations between several of the sociodemographic variables and the acceptance of the job offer. The transactional relationship with the waste pickers' association (Selling to the association) has a positive and highly significant effect on accepting the job offer. Women are more likely than men to change their jobs. The negative sign for the No other income earner in the household (HH) variable suggests that waste pickers who are the only income earners within a family are less likely to leave the dumpsite and work in a new job. Riskaverse waste pickers are less likely to accept a new job offer, although the result is not robust. The stated preference to accept a job offer during the interview survey does not explain their actual behavior (Wants to join PB production). To examine the heterogeneity of the treatment effects by preferences, interactions with the offer letter dummy variables and variables of time and risk preferences are added into equation (1). Table  4 reports the positive coefficient of the estimation of the interaction term of PR-Daily and the discount rate. Note that there is no significant correlation between a high discount rate and the acceptance of the job offer (see Table 3). This finding implies that, on average, changing payment frequency from once every three days to daily decreases the probability that an individual will accept the job offer. However, the magnitude of this decrease in probability is lower if an individual has a high discount rate.

Heterogeneous treatment effects
Column 2 in Table 4 reports that the coefficient of the interaction between the dummy variable of FW-Three and risk aversion is positive and significant. Note the slightly negative correlation of risk aversion and the acceptance of the job offer (see Table 3). Thus, risk-averse individuals are less likely to accept the job offer; however, the possibility that decreases with risk aversion falls if we change from piece rates to fixed wages. Similarly, the interaction of the dummy variable of PR-draw-Three and risk aversion is positive and significant. This finding implies that risk-averse individuals are less likely to accept the job offer with piece rate pay once every three days. However, the possibility that decreases with risk aversion falls if we add a draw, meaning that a draw of 30 PHP per day encourages risk-averse individuals to accept the offer. 7 Another interpretation of these results is possible. The impact of changing from piece rates to fixed wages of 65 PHP from the previous analysis is negative. Hence, the positive coefficient of the interaction of FW-Three and risk aversion suggests that the negative effect of 65 PHP per day reduces if an individual is more risk averse.

Conclusion
Environmental policies often force job changes for workers in various industries, and this can be considered to be a social cost. Considering the negative impact of such policy interventions, including the job losses caused by the closure of an open dumpsite, those who operate waste management services face an increasing normative expectation to handle the issues responsibly by providing alternative employment opportunities for waste pickers. This study involved a field experiment at a dumpsite in the Philippines to examine the effect of changing payment schemes on the number of waste pickers accepting the new job offer.
The data of ex-post interviews suggest an interpretation of the main finding that a change in payment frequency from once every three days to daily decreases the probability of accepting the job offer by about 30 percentage points. Four days after production started, two workers in the daily payment group asked us to change their payment frequency so that their salary would be paid two weeks later, claiming they wanted to generate greater savings. This anecdotal evidence suggests that the estimation results reveal evidence of waste pickers' self-control problem related to savings. In this sense, a new job with less frequent payment can serve as a commitment device. While alternative interpretations are possible, for example, daily payments are often associated with precarious jobs and this may lead to less acceptance, it is believed they are less likely, since the participants recognized that public sector organizations were offering these new jobs and that they had already implemented similar programs. That said, the findings do not imply that less frequent payments are generally better 5 . The authors believe this to be the case because the less frequent offer involved payment once every three days, which is more frequent than the common payment practice of once per month. Further studies are, however, needed to examine the preferred payment frequency for waste pickers as well as other low-income individuals in developing countries.
We found that waste pickers offered fixed wages of 65 PHP per day are less likely to accept a new job compared with those offered the piece-rate pay option. There are two potential explanations for this decision: (i) they simply prefer an uncertain offer, and (ii) they can successfully estimate their own productivity. In the production undertaken immediately after this field experiment, workers who received the second proposal (PR-Three) earned more than 65 PHP on average, suggesting that expected productivity in the production of paper briquettes is higher than 65 PHP, possibly because of the installation of new equipment. To mitigate inequality between the other groups, the fixed rate of pay for FW-Three was increased to 100 PHP from the first day. Nonetheless, average production by workers under the piece rate scheme was still twice as large as the group with fixed wages. This observation is consistent with existing studies (e.g., Shearer, 2004), implying that the piece rate scheme can decrease moral hazard and raise labor supply after a job change.
The fact that the transactional relationship with the waste pickers' association has a positive and highly significant effect on accepting the job offer makes sense, as the association supported the recruiting and provided the authors with a workplace for the alternative job. Indeed, evidence of a strong bond between waste pickers and buyers has also been demonstrated in previous studies (Hayami et al., 2006;Gill, 2007), suggesting that social networks such as buyerseller relationships affect the decision to change one's job.
Two concerns about the external validity of the authors findings should be mentioned here. First, there is concern about sample selection bias. Among the 240 workers surveyed before the intervention, only 112 participated in the experiment. Note that participants are less risk averse compared with nonparticipants (see Table A1 in the Appendix) and that the negative effect of changing from piece rates to fixed wages lowers with risk aversion (see Table  4). These results imply that the positive effect of piece rate pay may be overestimated. Second, there is concern about the specific sample problem (Peters et al., 2016). Since various programs have been conducted at the dumpsite in Iloilo City (Paul et al., 2012;Romallosa & Kraft, 2017), the participants are more familiar with these kinds of job offers, distinguishing them from the waste pickers at other dumpsites to some extent. However, although the participants recognized that they were in some kind of project driven by the local government, they might not have understood the intentions of the treatments. Thus, the authors believe that the potential hazards (related to the awareness of the study) to both the treatment group (the Hawthorne effect) and the control group (the John Henry effect) are less likely. Indeed, the impacts of changing payment scheme in the formal sector on job changes in the informal sector do not generate a general equilibrium effect. Thus, the proposed treatments can be scaled up despite external validity concerns.
Finally, it is more difficult to encourage men to change their jobs compared with women, especially when they are the only income earners in a household and show risk aversion. Note that this study considers the job change behavior of waste pickers to be a decision made by the individual alone. This assumption might need to be modified, as the decision to change one's job also concerns the entire household. Further studies are thus required to explore how to encourage those who did not accept the job offer in this study. Notes: Means are reported from the survey data for participants and nonparticipants. Standard deviations are in parentheses. See the notes of Table 2 for a description of the variables. One observation from nonparticipants lacks data on average hourly earnings. Two observations from nonparticipants and three observations from participants lack data on the willingness to accept closure category.