Policy Research Working Paper 9108 What Aspects of Formality Do Workers Value? Evidence from a Choice Experiment in Bangladesh Minhaj Mahmud Italo A. Gutierrez Krishna B. Kumar Shanthi Nataraj Development Economics Knowledge and Strategy Team January 2020 Policy Research Working Paper 9108 Abstract This study uses a choice experiment among 2,000 workers value job stability: the average worker would be willing to in Bangladesh to elicit willingness to pay (WTP) for job forego a 27 percent increase in income to obtain a 1-year attributes: a contract, termination notice, working hours, contract (relative to no contract), or to forego a 12 percent paid leave, and a pension fund. Using a stated preference increase to obtain thirty days of termination notice. There method allows calculation of WTP for benefits in this set- is substantial heterogeneity in WTP by type of employment ting, despite the lack of data on worker transitions, and the and gender: women value shorter working hours more than fact that many workers are self-employed, which makes it men, while government workers place a higher value on difficult to use revealed preference methods. Workers highly contracts than do private sector employees. This paper is a product of the Knowledge and Strategy Team, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at snataraj@rand.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team What Aspects of Formality Do Workers Value? Evidence from a Choice Experiment in Bangladesh Minhaj Mahmud, Italo A. Gutierrez, Krishna B. Kumar, and Shanthi Nataraj JEL classification codes: J46, J32, O17 Keywords: informality, working conditions, choice experiment Minhaj Mahmud is a Senior Research Fellow at the Bangladesh Institute of Development Studies (BIDS); his email address is minhaj@bids.org.bd. Italo A. Gutierrez is a Senior Economist at Amazon; his email address is gitalo@amazon.com. He completed this work prior to joining Amazon. Krishna B. Kumar is a Senior Economist and Director of International Research at the RAND Corporation; his email address is kumar@rand.org. Shanthi Nataraj (corresponding author) is a Senior Economist at the RAND Corporation; her email is snataraj@rand.org. This document is an output from a project funded by the UK Department for International Development (DFID) and the Institute for the Study of Labor (IZA) for the benefit of developing countries (IZA/DFID GLM|LIC Programme, Grant Agreement GA-C3-RA1-299). The views expressed in this paper are not necessarily those of DFID or IZA. Kumar also acknowledges support from the Rosenfeld Program on Asian Development at the Pardee RAND Graduate School. We are grateful to Dr. Shamsul Alam, Member (Senior Secretary), General Economics Division, Bangladesh Planning Commission, for his input and support of this research. We thank participants at the World Bank Jobs Group Brown Bag Lunch, the BIDS Workshop, the Workshop on Formal and Informal Labor Nexus and Bangladesh’s Growth, IZA conferences in Oxford and Washington, and the Western Economic Association International meetings in San Diego, for their valuable comments. We also thank three anonymous referees for their constructive suggestions. 1. Introduction A large share of workers in developing countries are part of the “informal” economy – whether that is measured in terms of lack of firm registration, tax evasion, self- employment, employment in small firms, or off-the-books employment (ILO 2014, LaPorta and Shleifer 2008, Tybout 2000). The prevalence of informality presents a concern for a variety of reasons, including the fact that workers in the informal sector are often subject to poor working conditions, and receive few or none of the benefits that are available to formal workers such as written contracts or paid leave. Attempting to extend formal job protections to informal workers – or even to enforce all existing protection as per prevailing labor laws – is beyond the reach of regulatory agencies in many settings, and may actually encourage informal firms to move even further into the shadows. Thus, understanding which job benefits are most valuable to workers can help policymakers and regulators to focus their limited resources. This paper used a choice experiment involving approximately 2,000 workers in Bangladesh to elicit workers’ preferences for specific benefits associated with formal employment. In the choice experiment, each worker was presented with two alternative jobs, which differed in terms of the levels of five attributes: a written contract, termination notice, paid leave, working hours, and access to a retirement fund. The worker was also told the monthly income that would be associated with each job.i The worker was then asked to decide which job he or she would select, if given an opportunity to choose between the two alternatives. We used the tradeoffs between monthly income and each of the other attributes to estimate the willingness to pay (WTP) for that particular attribute (i.e., job benefit). 2 One key finding is that workers have a strong preference for job stability. The average worker would be willing to forego an increase of about 19 percent of monthly income for a 6-month contract, 27 percent for a 1-year contract and 44 percent for a long- term contract (relative to no contract); and to forego an increase of about 12 percent of monthly income in order to obtain thirty days of termination notice. A second important finding is that workers exhibit substantial heterogeneity in their preferences for job attributes. On average, women are willing to give up a greater amount of income than men, to avoid longer working hours. We also find preference heterogeneity across different types of employment: relative to private sector employees, government employees, on average, place a higher value on long-term contracts, and casual workers have a strong preference for higher income relative to other job attributes. A key contribution of this work is that it is one of only a handful of choice experiments that examine preferences for job attributes. Such stated preference methods have been used extensively in the marketing, environmental, health and transportation literature. Notably, however, a few papers have used choice experiments to provide a better understanding of workers’ valuations of job attributes. For example, Ubach et al. (2003) and Scott et al. (2015) used choice experiments to examine preferences for job attributes among health care workers. More recent examples include Mas and Pallais (2017), who estimated the value of scheduling flexibility and working from home; Wiswall and Zafar (2018), who estimated WTP among undergraduates at New York University for job attributes such as wage increases, bonuses, and job security; Maestas et. al (2018), who estimated WTP for a wide range of working conditions in the United States; and Assy et 3 al. (2018), who estimated youth preferences and WTP for different job attributes and support services in Kenya. The traditional approach to estimate WTP for job attributes is to apply hedonic methods to observable data on monetary and non- monetary job attributes. A key advantage of this revealed preference approach is that it is based on observable market transactions. However, it has been difficult to obtain reliable estimates using this approach, because the tradeoffs that individuals face in the labor market are not randomly assigned, but likely depend on the individuals’ labor market productivity and on the characteristics of the firms to which they can apply, neither of which are fully observable. For instance, high productivity workers are likely to receive offers with higher wages as well as more desirable non-wage job attributes such as greater flexibility in terms of working hours (Hwang et al. 1992). Suppose the true underlying tradeoff workers make is to accept lower wages in exchange for more flexible hours. If we cannot fully control for individual productivity, we will observe a positive correlation between earnings and flexibility of hours across individuals, which does not reflect this true tradeoff. In addition, in the presence of search frictions in the labor market, more cost-efficient firms will offer jobs that not only pay higher wages, but also offer better non-wage attributes, because they face greater opportunity costs when they have job vacancies that go unfilled (Hwang et al. 1998). Thus, if it is not possible to fully control for differences in cost-efficiency across firms, the WTP estimate from hedonic wage regressions will be biased.ii An important advantage of using a choice experiment – a stated preference method in which respondents are asked to choose between two hypothetical options, each with different levels of attributes – is that the attributes can be randomized across individuals, 4 thus mitigating the concern that the preferences underlying the observed tradeoffs are correlated with unobserved characteristics. A second advantage of choice experiments is that they allow us to estimate the valuation for job attributes that do not vary substantially across workers, and thus may be difficult to identify using revealed preference data (Hanley, Wright and Adamowicz 1998). This can happen either because certain attributes are relatively standard (e.g. paid vacations) or relatively uncommon (e.g. working from home or unlimited vacation). Relatedly, choice experiments also allow the estimation of valuations of novel attributes that are not included in standard surveys. Another important difference, as pointed out by Mas and Pallais (2017), is that valuations from hedonic wage regressions reflect preferences of marginal workers in jobs with and without the attribute being examined; in contrast, choice experiments allow us to recover the average valuation of attributes in the study population. The potential disadvantage in using a stated preference method is that it is possible that what workers say they will do in a choice experiment might not reflect what they would actually do if faced with the choice in real life. This concern is particularly important when respondents are faced with choices they are not familiar with, and are asked to make tradeoffs over which they have “an absence of preferences” (Diamond and Hausman 1994). We took several steps to mitigate this concern. First, we carefully identified the types and levels of job benefits to reflect realistic conditions that workers in Bangladesh were likely to face in the labor market, based on a review of the Bangladesh Labour Law, discussions with policymakers and other stakeholders in Bangladesh, and focus groups with workers. Second, as discussed in more detail below, we limited the survey to cover only working adults, since they are more likely to have had real-world experiences in making tradeoffs 5 between different job attributes. Third, we tested the choice experiment through a pilot survey, and modified the attribute levels based on feedback regarding the appropriateness and realism of the scenarios. Finally, in the actual survey, we used practice questions to ensure that respondents understood the choice exercise. More broadly, there is encouraging validating evidence that when choice experiments are well-designed, they can produce findings that are in line with observed market transactions. Mas and Pallais (2017) used a choice experiment, embedded in a real hiring process, to estimate job applicants’ valuations for scheduling flexibility and the option of working from home. In addition, they used a similar choice experiment to elicit preferences through a survey of a different population (unrelated to the hiring process). They found that the survey-based valuations were close to the valuations from the experiment that was embedded in the real hiring process. Further evidence in support of the consistency of choice experiments is that recent studies suggest that individuals who have a given amenity in their current job tend to value that amenity more highly than workers who do not have that amenity (Maestas et al. 2018). As discussed below, we also find this pattern; most notably, government employees in our sample, who have the highest job stability (Gutierrez et al. 2019) and are much more likely to have long-term contracts, exhibit the highest valuation for these contracts. Existing estimates of WTP for non-wage job attributes vary substantially. Traditional methods using hedonic wage regressions have generally found small valuations (Bonhomme and Jolivet 2009); as noted above, there results are likely biased downwards due to unobserved heterogeneity among workers and firms as well as search frictions. Alternative approaches to estimate WTP from market transactions include the analysis of 6 job durations and the estimation of job search models. Studies using these approaches have found considerably higher WTP values for non-wage job attributes. For instance, Gronberg and Reed (1994) found that white male workers in the United States are willing to pay 28 percent of their average salaries for jobs without extreme work conditions.iii Similarly, using Dutch longitudinal data, Van Ommeren et al. (2000) found that the average WTP for one hour of commuting is estimated to be equal to almost half of the hourly wage rate. Dey and Flinn (2008) found that, in the United States, households would be willing to have the husband (wife) take a 41 percent (45 percent) pay cut to get a job with health insurance coverage. Using data from eight European countries, Bonhomme and Jolivet (2009) found that the marginal WTP for job attributes are usually between 20 percent and 40 percent of earnings, depending on the job attribute and the country in question.iv Similarly, the recent choice experiment literature has found substantial WTP for various job attributes. For example, Maestas et al. (2018) estimated that the full bundle of amenities included in their experiment is valued at 56 percent of wages. Mas and Pallais (2017) found that workers are willing to give up 20 percent of wages to avoid employer discretion in work scheduling. This study complements, and contributes to, this strand of literature in several ways. Most workers in Bangladesh, as in many developing countries, are informally employed or self-employed, and thus there are few data on their current job attributes, on which to base revealed preference estimates. Even among formally employed workers, longitudinal data are scarce, limiting the extent to which more sophisticated revealed preference methods such as search models can be used. Our study demonstrates how a choice experiment can convincingly be used to estimate WTP in such a setting. 7 By creating hypothetical tradeoffs between specific job benefits and wages, and keeping everything else equal between alternative job options, we are able to calculate quantitative estimates of workers’ valuations of specific benefits, and identify which benefits are valued the most. In addition, by including formal, informal, and self-employed workers in the survey, we can examine the extent to which preferences for specific attributes differ by individual characteristics and type of current employment. This study also contributes to the policy discussion regarding informality. While many employment protections, such as basic occupational safety and paid leave, are supposed to apply to all workers, in reality such protections are more likely to be enforced in larger, formal firms (and even there not on a uniform basis) and government agencies may not have sufficient resources to extend enforcement efforts to numerous, small firms. Moreover, attempting to extend additional formal protections or to enforce all existing protections on informal firms may encourage them to become even less visible to regulators. This study therefore views formality as a continuum, and our findings shed light on which benefits attached to formal employment might be more valuable to workers, and thus should have higher priority in policy discussion. The rest of this paper is organized as follows. Section 2 provides a brief overview of the survey methods, while Section 3 describes the choice experiment design. Section 4 presents results, and Section 5 concludes. 2. Survey Methods We conducted a survey of approximately 2,000 workers in the two major administrative divisions of Bangladesh.v We used a two-stage sampling design to select workers to interview in four districts: Dhaka, Gazipur, and Narayanganj districts in the 8 Dhaka division, and Chittagong district in the Chittagong division. We began with a list of the number of households in each of 1,971 mouzas (the lowest administrative level at which national surveys are conducted in Bangladesh and for which such data were available) in the four targeted districts. We used the mouza as our primary sampling unit (PSU) and selected 80 mouzas with probability proportional to size (PPS) where size was the number of households.vi Within each mouza, we selected 30 households for first-stage sampling, using a random walk method, as follows. Three enumerators, plus a supervisor, would arrive at the selected mouza. The supervisor first reviewed the general layout of households in the mouza, and then directed each enumerator to start at a different point, typically along the outside border of the mouza. The enumerator was instructed to visit each 10th household.vii At each selected household, we used a screener questionnaire to collect basic information about all household members, including age, gender, employment status and type of employment. We collected screener information for the whole household from any adult who was available at the time. Among the households included in the screener, there were a total of 10,690 household members, including 7,239 adults. Among these adults, 3,787 were reported to be working. We focused on working adults for several reasons. First, our aim was to examine the tradeoffs that individuals would make between different job attributes, conditional on having decided to participate in the labor force. In practice, this means that we exclude many women from the sample, since the Bangladeshi labor force participation rate (LFPR) for men was 81.7 percent in 2013, compared with an LFPR of 33.5 percent for women.viii In our results, we examine how preferences differ among currently working men 9 and women, but we anticipate that the tradeoffs that working women make may be different from the tradeoffs that non-working women would make. Second, for the purpose of the choice experiment, it was important that the respondents have some understanding of the various working conditions that we described. While the working adults may not have held jobs with each of the specific attributes, their labor market experience would give them some context they could use to make decisions between the two jobs offered. As we discuss in the conclusion, it is possible to develop vignettes that provide sufficient context for non-working adults to make realistic tradeoffs, but that would require a different choice experiment module than we used for working adults. Finally, we note that focusing on working adults makes our results more comparable with studies that use hedonic methods, since they rely on observed job histories, and thus are confined to working individuals. Therefore, among the adults identified in the roster, we focused on the 3,787 who were working at the time we conducted the screener. To fully reflect the prevalence of informal work, we used a broad definition of “working”; in the screener, we emphasized that “working” could include wage or salaried work, self-employment, or contributing to a household business (whether paid or unpaid).ix The target sample size was 2,000 workers. To select these 2,000 workers from among the 3,787 working adults in the household roster, we stratified the list by gender and by employment type.x Given the low number of working women, we aimed to include all of them in the sample. We also aimed to include all men in the employment type cells with few individuals (seasonal worker, domestic worker, and apprentice/intern/trainee). 10 Within the other employment type cells for men, we randomly selected 40 percent of the workers. Table 1 shows the total number of adults in the roster, as well as the number of working adults, by gender and reported employment type. As noted above, we only targeted working adults. Table 1 also shows the total number of respondents who were targeted, by gender and employment type, as well as the total number who were ultimately included in the sample. There are three main reasons for the discrepancies between the target and the sample. First, the person who responded to the screener did not always accurately provide the employment type for the selected individual. In addition, 15 women were initially reported as working in the screener, but were later found not to be working. Second, we were unable to locate 66 targeted individuals, and therefore replaced them with other, similar working individuals from the roster.xi Third, 36 targeted individuals were found, but failed to complete the survey. Thus, out of the target sample, we successfully completed 1,964 interviews (98 percent completion rate). 3. Choice Experiment Design and Model Estimation This section describes how the attributes and levels included in the choice experiment were identified, provides an overview of the experimental design, and presents the models used to analyze the choice experiment data. Appendix S1 of the supplementary appendix presents further details regarding the choice experiment design. Attributes and Levels 11 In the choice experiment, respondents were told that they would be presented with two hypothetical employment opportunities, each with different levels of the following attributes: a written contract, termination notice, paid leave, working hours, a retirement fund (Provident Fund), and monthly income.xii The alternatives were unlabeled; that is, we did not identify jobs as “formal” or “informal”, as doing so may have caused respondents to make assumptions about other aspects associated with the jobs, rather than focusing on the attributes listed above. They were asked to assume that all other attributes not presented in the scenario were identical between the two jobs.xiii The respondents were then asked to indicate which of the two jobs they would select if given a choice. Table 2 shows an example of a choice set. This could reflect a real-world situation in which a worker has to choose between two different job offers. Job A offers a 6-month written contract, 15 days of termination notice, working hours in the range of 30-45 hours/ week, and 10 days of paid leave, but no Provident Fund. The respondent is told that Job A would pay a monthly salary 20% above his or her current monthly income. Job B offers the same attributes in terms of termination notice, paid leave, and Provident Fund. However, this job offers a longer contract (1 year), longer working hours (45-60/week), and lower increase in income (10% increase over current income). The full set of attributes and levels is shown in Table 3. The specific attributes and their levels were initially selected based on a review of the Bangladesh Labour Law, and were refined through a series of discussions with policymakers and other stakeholders, as well as focus groups with workers, and a pilot survey. Since the survey included respondents who were self-employed or working for family businesses, and also included 12 respondents with a wide range of current incomes, we framed the monetary variable relative to current monthly income, rather than a fixed wage.xiv The full survey instrument, including the choice experiment module, is provided as Appendix S2. Experimental Design Given the numbers of attributes and levels, it was not possible to include all combinations of them in the choice experiment. Therefore, we used an efficient experimental design in order to create a reasonable number of choice sets. We developed an efficient design that would allow identification of all main effects, using Ngene software by ChoiceMetrics. The attributes were combined to create 48 choice sets in 8 blocks, each with 6 choice situations. To mitigate fatigue and cognitive burden, each respondent was presented with one block of 6 choice sets. The blocks were randomly assigned to respondents, and the choice sets were randomly ordered within each block. Further details regarding the design of the choice experiment are provided in Appendix S1. Model Estimation To estimate WTP using the choice experiment data, we follow standard practice by starting with a random utility model (McFadden 1973). Suppose that the utility an individual receives from a particular alternative j depends on the attributes xj associated with that alternative: U v x ε where εj reflects individual characteristics that are unobserved by the researcher. Each individual chooses the alternative that provides the maximum utility, and the probability that the individual selects alternative i from choice set S is given by: 13 , ∀ Assuming that the unobserved errors are independent and identically distributed with a Type 1 extreme value distribution, this yields a conditional logit model. Using a linear-in-parameters utility function, we can write the probability that individual n chooses alternative i as: exp ∑ ∈ exp The parameters can then be estimated using a standard maximum likelihood model. The ratio between the parameter estimates for any two attributes k and m yields the marginal rate of substitution (MRS) between them. Since one of the attributes is monthly income, the marginal value of any other attribute can be estimated by taking the ratio of the parameter on that coefficient βk and the parameter on income βw: The standard conditional logit model assumes that preferences are homogeneous. We use two methods to explore heterogeneity in preferences among respondents in our sample. First, we include interaction terms between the preference parameters and observable individual characteristics.xv Second, we estimate a latent class model, which allows the preference parameters to vary across groups, or classes, of individuals. The probability that individual n chooses alternative i, conditional on being in class c, is: exp | ∑ ∈ exp Following Swait (1994) and Boxall and Adamowicz (2002), we estimate the probability that individual n is in a particular class c as: 14 exp ∑ exp where zn are observable characteristics associated with individual n. The unconditional probability of observing individual n choosing alternative i is therefore equal to the probability that the individual is in class c, times the probability of choosing alterative i conditional on being in class c, summed over all classes: Pr ∗ Pr | We estimate the preference parameters () as well as the parameters on individual characteristics associated with class membership () using an expectation maximization algorithm. 4. Results This section presents basic demographics of the survey sample, before proceeding to the choice experiment results. Respondent Demographics Table 4 shows basic demographics for the 1,964 individuals who completed the survey, with and without sampling weights. The weighted and unweighted averages are quite similar, with the main differences driven by the fact that we attempted to include all working women in the sample, whereas for most categories of male employment, we sampled 40 percent of men.xvi Thus, while only 18 percent of the underlying population of working adults in our roster were women, fully 35 percent of respondents were women.xvii 15 The mean age of respondents was approximately 36 years. Most workers had some schooling, typically at the primary or secondary level. Because of our focus on the Dhaka and Chittagong areas, about 20 percent of the (weighted) sample consisted of workers in the garments industry, and another 15 percent were in other types of manufacturing industries. Consistent with the propensity of women to work in the garment sector, the share of garment workers was even higher (25 percent) in the unweighted sample. Panel (b) shows the distribution of respondents by gender and type of employment.xviii Both weighted and unweighted results are provided for men and for the overall sample; since all working women were targeted, the weighted and unweighted results for women are the same. About 40 percent of men, and 50 percent of women, were private sector employees. About 45 percent of men, and 25 percent of women, were either self-employed or family members working in a household business. About one-third of the self-employed men, but only about 7 percent of self-employed women, reported having non-family member employees in their businesses.xix Panel (c) shows basic summary statistics for the types of benefits reported by workers in their current jobs. This panel only reports benefits for wage workers (that is, we exclude workers who are self-employed or are family members working in household businesses). Workers are grouped into three categories: government employees (including workers in government and semi-government entities), private employees (paid employees in private entities), and casual workers (apprentices, seasonal workers, day laborers/casual workers, and domestic workers in private households). As we would expect, benefits were most prevalent among government employees; the degree of formality is decreasing across the columns. Written contracts, sick leave, 16 casual leave (that is, paid vacation), and access to a Provident Fund were nearly ubiquitous among government workers. About 75 percent of all government workers reported that they would receive termination notice if they were let go from their jobs. Private sector employees were also more likely to receive benefits than casual workers, with 25 percent reporting written contracts, and nearly 50 percent reporting verbal contracts. Sick leave and casual leave were also fairly common. Termination notice was less common, only being reported by 45 percent of workers, and only around 12 percent reported having access to a Provident Fund. Not surprisingly, almost none of the casual workers reported written contracts, but half did report verbal contracts, and over 60 percent reported sick leave. Casual leave was only reported by about 25-30 percent of casual workers. About 15-20 percent of casual workers reported that they would receive termination notice if they were let go from their jobs; most of these were domestic workers. Panel (d) reports the 25th, 50th, and 75th percentiles of monthly income reported by individuals.xx Wage workers are grouped into the three categories discussed above, and non-wage workers into three additional categories: self-employed workers who either do not have any employees or only have family employees, self-employed workers who have at least one non-family employee, and family members working in household businesses. As expected, median monthly income is highest for government workers, at about 25,000 Taka (319 USD). Interestingly, self-employed workers report the next highest monthly income, with those who have non-family employees reporting about 20,000 Taka (255 USD), and those who do not reporting about 15,000 Taka (191 USD), or 11,500 Taka 17 (147 USD) when not weighted. Private employees report a median monthly income of 9,000 Taka (115 USD), or 8,000 Taka (102 USD) when not weighted, while casual and family workers report only 7,500 Taka (96 USD).xxi Strictly Dominated Alternatives The attributes and levels for the choice experiment, and the pairing of alternatives in the choice scenarios, were developed with the aim of asking individuals to make relatively difficult tradeoffs. If the tradeoff between alternatives was too easy to make, then we would gain little information about how much people value each attribute. Thus, we expected to see individuals making different choices – in other words, we did not expect to see every respondent who was faced with the same choice scenario to select the same job. However, we did include three scenarios in which one of the alternatives was strictly dominated. That is, the strictly dominated alternative was paired with another alternative that was likely to be preferred in terms of all attributes. In these cases, we would expect few, if any, individuals to select the strictly dominated alternatives. The inclusion of strictly dominated alternatives does not provide substantive information about preferences, but does serve as a check on whether respondents understood the scenarios and made coherent choices. For each of the 48 choice scenarios, we calculated the percent of respondents who chose Job A and the percent who chose Job B, and then identified the minimum (for example, if 45 percent chose Job A and 55 percent chose Job B, we identified the minimum as 45 percent). A small minimum suggests an easy choice. 18 Figure 1 shows the empirical CDF of these minima. In general, it appears that there was a substantial amount of heterogeneity in choices, with many scenarios in which about one-third of respondents selected one job but the other two-thirds selected the other. The strictly dominated scenarios appear as triangles, and are all in the left tail of the CDF. That is, as we would expect, the strictly dominated jobs were selected by very few individuals, far fewer than those who selected one of the non-dominated jobs. Conditional Logit Results Table 5 presents results from the conditional logit model. In Column (1), all attribute levels are included as dummy variables. The excluded levels are: contract (none), termination notice (none), hours (30-45), paid leave (none), Provident Fund (none), and income (no change). Sampling weights are applied, and standard errors are clustered at the level of the respondent. In Column (2), we enter the percent change in income as a linear variable rather than as a set of dummies. A small number of respondents indicated that they did not receive a wage in their current jobs; since the percent change in income may have been meaningless for these respondents, we therefore excluded them in Column (3).xxii In Column (4), we enter notice, work hours, and leave linearly (where hours are set equal to the median number in each bin). Column (5), like Column (3), excludes those with zero reported income from their current jobs, using the linear specification of Column (4). In Column (6), we re-estimate the model in Column (4), without applying sampling weights, and confirm that results are similar. All of the coefficients are highly significant across specifications, and the signs of the coefficients are consistent with economic theory. First, as expected, the coefficients on greater monthly income are positive. Turning to the benefits, the valuation of a contract 19 increases as the length of the contract increases. Since we included the levels of contract as dummy variables, we tested whether the coefficients differed across different levels of the same attribute that were included as dummy variables (e.g., whether the coefficient on a 6-month contract was statistically distinguishable from the coefficient on a 1-year contract).xxiii Results from a joint test of the three levels included in the regression, as well as pairwise tests between each pair of included levels, indicate that the differences in valuation between levels were statistically significant at the 1 percent level. In Columns (1) through (3), we performed similar tests for the dummy variables for notice, leave, and working hours. The coefficients across levels were significantly different from each other at the 1 percent level in all but a few cases.xxiv In Columns (1) through (3), the valuation of notice generally increases with the number of days of notice. Similarly, workers are willing to give up more to avoid a 60-75 hour week than to avoid a 45-60 hour week, all else equal. These findings are consistent with the positive valuation on notice, and the negative valuation on hours, that we find in Columns (4) through (6), when these benefits are entered linearly rather than as dummies. Although we cannot always reject that different levels are valued equally, when leave is entered linearly in Columns (4) through (6), we find a clear and positive coefficient on the total number of days of leave. On the whole, and consistent with theory, the results indicate that respondents prefer longer contracts, more days of termination notice, greater amounts of paid leave, and fewer working hours (conditional on receiving the same income). They also prefer having a Provident Fund (pension fund). 20 Importantly, a benefit of including income as a job attribute is that the average respondent’s WTP for a marginal change in each of other attributes can be estimated. In Table 6, we use the coefficients from Column (4) of Table 5 to calculate the marginal value of each attribute (i.e. the marginal rate of substitution between the attribute and a one percent change in income, which is calculated by dividing the coefficient on each attribute by the coefficient on the percent change in income). The results suggest that the average worker would be willing to forego a 19 percent increase in monthly income in order to obtain a job with a 6-month contract (relative to no contract). This amount rises to 27 percent for a 1-year contract and 44 percent for a long-term contract. Workers’ preferences for job stability are also illustrated by the marginal value of termination notice. Since we modeled notice in terms of number of days, the value in Column (2) of Table 6 suggests that the average worker would be willing to forego an increase of 0.4 percent of monthly income for each additional day of notice. The typical requirement for notice in the Bangladesh Labour Law is 30 days; if we extrapolate the marginal value, we estimate that the average worker would be willing to forego an increase of about 12 percent of monthly income for 30 days of notice.xxv Similarly, the MRS in Table 6 indicates that the average worker would be willing to forego an increase of about 5 percent of monthly income to receive the 10 days of paid leave that are typically required under the Bangladesh Labour Law. Finally, Table 6 suggests that the average worker would be willing to forego an increase of around 18 percent of monthly income to have access to a job with a Provident Fund. Exploring Heterogeneity in Preferences 21 One advantage of including a wide variety of workers in the choice experiment is that we can examine how preferences vary across observable characteristics. First, we examined whether the valuation of benefits differed by gender or education level of the worker. To do so, we interacted each attribute with a female dummy. Separately, we also interacted each attribute with a dummy for workers who had at least some secondary level education. The WTP for each attribute for women is calculated by summing the coefficients on the baseline attribute plus the interaction term (attribute x female dummy), and dividing by the sum of the coefficient on income in the baseline plus the interaction term (income x female dummy). A similar calculation is used for those with at least some secondary education. Table 7 shows the coefficients and marginal values of each attribute (relative to a percent change in income) across these different groups. Columns (1) and (2) show the baseline coefficient (for men) and the related WTP. Column (3) shows the interaction terms for women, while Column (4) shows the estimated WTP for women. Most notably, women’s WTP for shorter hours is substantially larger in magnitude than the WTP among men – to work 10 fewer hours in a week, for example, men would be willing to forego a 3.8 percent increase in monthly income, whereas women would be willing to forego a 6.4 percent increase. While the interaction terms for contracts and access to a Provident Fund are also statistically significant, women’s WTP for these attributes is similar to those for men. Columns (5) through (8) show results by education level, and suggest that contracts are substantially more highly valued by those with higher levels of education. The WTP 22 for the other attributes is fairly similar between those with and without at least some secondary education. Similarly, Table 8 shows the coefficients and WTP for each type of benefit by employee type. We interacted each attribute with dummy variables for government employees, casual workers, and the self-employed (including family members). Columns (1) and (2) show results for private employees (the baseline category), while Columns (3) through (8) show the interaction terms and WTP for the other types of workers. The results in Table 8 suggest that government employees place a much higher value on contracts – particularly on long-term contracts – than do other types of employees. Of course, it is likely that the preference for such stability is what attracted these workers to the government sector in the first place. The valuations for contracts are fairly similar across the other types of employees. Although the interaction terms on 1-year and long-term contracts are negative and significant for the self-employed, the interaction term on income is also negative, and the overall WTP for contracts among the self-employed is similar to the WTP among private employees. We also explored preference heterogeneity by estimating a latent class model with three classes.xxvi A key advantage of the latent class model, relative to the interactions shown above, is that it allows us to sort individuals into classes based not only on their observable characteristics, but also based on the choices that they make in the experiment. Table 9 shows preference parameter estimates for each class. Columns (1), (3) and (5) show the coefficients on each attribute for the three classes, while Columns (2), (4) and (6) show the associated, average WTP for workers in each class. 23 We label Class 1 as “Stability Seeking” because the WTP among individuals in this class is characterized by a particularly high valuation for contracts. For these workers, the coefficient on leave is not significantly different from zero, suggesting that they do not place a high value on paid vacation. While the coefficient on a Provident Fund suggests a WTP of 12 percent, this coefficient is not statistically significant at conventional levels. We label Class 2 as “Income seeking,” as individuals in this class exhibit a high value for additional income; the coefficient on income (0.097) is nearly twice as high as the coefficient on income for the overall sample (0.051, Table 6). In contrast with the “Stability Seeking” class, those in the “Income Seeking” class exhibit a low WTP for contracts. The WTP for a long-term contract (19 percent) is less than half the WTP exhibited by the overall sample (43 percent, Table 6). Individuals in Class 3 exhibit a particularly high valuation for access to a Provident Fund; therefore, we label this class “Long-Term Planners.” Their valuations for contracts are between those exhibited by the other two classes. We caution that these labels are merely for illustrative purposes and do not capture all of the differences in preference parameters across the classes. For example, individuals in Class 3 also exhibit a relatively high valuation for termination notice, and a strong aversion to longer working hours. The labels above are based on the observed preferences of individuals in each class. We can also gain insights into the three classes by examining the individual characteristics associated with membership in the three classes, in two ways. First, we can use information from the class membership equations (shown in Appendix A1), which capture the probability that individuals with certain characteristics are in each class. Second, we can assign each individual to the class to which he or she has the highest probability of 24 belonging, and examine the distribution of key characteristics by class assignment. In doing so, we use the posterior probability, which is based on observable characteristics, as well as the actual choices that the individual is observed to make. Figures 2 and 3 show the percentage of individuals who are assigned to each class, by employment type and gender, respectively. Table 9 shows that nearly half of the workers in the sample are assigned to Class 2 (“Income Seeking”). In Appendix A1, the class membership results indicate that individuals with low levels of education are more likely to be in this class. Consistent with this finding, the average monthly income reported by individuals in this class is also lower than in the other classes, and Figure 2 shows that casual workers (those without steady jobs) are more likely than any other type of worker to be in this class. Thus, these individuals may have a relatively high marginal utility of money. About one-fifth of individuals are assigned to Class 1 (“Stability Seeking”). As shown in Appendix A1, membership in this class is associated with higher levels of education and larger enterprise size. Figure 2 also shows that government employees are more likely than private sector workers to be in “Stability-Seeking” class. Taken together, these findings suggest that workers who are more likely to have relatively stable employment, exhibit stronger preferences for such stability. The remaining one-third of individuals are assigned to Class 3 (“Long-Term Planners”). The average reported income of individuals in this class is between that of the “Stability Seeking” and “Income Seeking” classes; however, individuals in this class appear to be least sensitive to additional income. As shown in Figure 2, the self-employed are more likely than private sector employees to be in this class. The relatively strong 25 aversion to working longer hours exhibited by this class may also reflect the finding that women are more likely to be in this class than in the other classes (Appendix A1 and Figure 3). 5. Conclusion What specific aspects of formal jobs do workers value the most? This paper used a choice experiment to elicit workers’ WTP for contracts, termination notice, paid leave, preferred working hours, and access to a pension plan. The results suggest that among these attributes, workers most value job stability – that is, the guarantee of longer-term employment ensured by a contract. The baseline results show that the average worker would be willing to forego an increase of about 19 percent of monthly income for a 6- month contract, 27 percent for a 1-year contract and 44 percent for a long-term contract (relative to no contract). Similarly, the average worker would be willing to forego an increase of about 12 percent of monthly income in order to obtain thirty days of termination notice. The findings are consistent with previous work from other countries. Bonhomme and Jolviet (2009) estimated a job search model for eight European countries, and found that the marginal WTP for job security ranges from 30 percent to 92 percent, making it the highest ranked attribute in five out of the eight countries. Using a choice experiment, Wiswall and Zafar (2018) found that undergraduates at New York University would be willing to give up 2.8 percent of earnings to lower the probability of job dismissal by 1 percentage point. In their choice experiment, Assy et al. (2018) found that youth in Kenya would be willing to pay about 27 percent of the average earnings offered in exchange for additional job security.xxvii Thus, there is consistent evidence, arising from revealed and 26 hypothetical choice methods, that workers are willing to pay a substantial amount for more secure jobs. The average WTP estimates mask substantial heterogeneity among workers. Including interactions with observable characteristics in the conditional logit model, and using a latent class model, indicate that women are more averse to longer working hours than men. In addition, government employees are more likely to place a higher value on long-term contracts than private sector employees. In contrast, casual workers are substantially more likely than private employees to have a particularly strong preference for higher income, and a relatively low WTP for various benefits. The study lends support to the use of choice experiments to overcome the challenge of estimating WTP for specific job benefits from hedonic wage regressions or from observed job durations, especially in a setting in which longitudinal data are unavailable or in which many workers do not participate in the formal sector. The results from the choice experiment are consistent with economic theory, and the use of a stated preference method allows us to gauge the valuation of specific attributes by a wide range of workers – including casual workers and the self-employed, who may never have received some of those benefits. Despite the heterogeneity in observed preferences, we find a substantial amount of WTP for contracts and termination notice among all of the groups of workers we examine. To the extent that the capacity for enforcement of existing labor regulations is limited, it may therefore be valuable for policymakers to focus on aspects that improve job stability. The study also points to areas that may be fruitful for further research. First, our findings suggest potential sorting in the labor market – that is, employees with stronger 27 preferences for certain types of benefits will be more likely to take jobs that offer those benefits. They may also, however, point to loss aversion – workers who are in jobs that have certain types of benefits may seek to avoid losing those benefits. Disentangling these two possibilities would improve our understanding of the drivers of the preference heterogeneity that we document. Second, most research on preferences for job benefits, including this study, focuses on workers’ preferences. In a separate study, we also examine employers’ preferences and willingness to offer different types of job benefits, which could provide useful evidence for policymakers who seek to encourage provision of those benefits (Kumar et al., 2019). Finally, our sample was confined to adults who were working at the time of the survey; thus, we excluded the majority of adult females. The findings suggest that working women’s preferences for job attributes differ from men’s in at least one key way: women have a stronger preference for lower working hours, likely due to the social norm that they are responsible for household duties. Had we included non-working adults as well as working adults, it is possible that our overall results would have reflected a higher valuation of shorter working hours. Had we included additional, related job attributes, such as flexible work schedule and ability to work from home, these might also have yielded high valuations, particularly among women. A fruitful area for research would be to conduct a stated preference study of women who are not in the labor force, to identify what types of job attributes might induce them to join the labor force (or might induce their families to be more willing to allow them to participate). Such an experiment would have to be carefully designed to ensure that the vignettes provided give non-labor force participants a sufficient understanding of working conditions to allow them to make realistic tradeoffs. 28 Notes i Monthly income was presented as percentage increase in income compared to the respondent’s current monthly income from his or her main economic activity. ii Some early studies attempted to overcome these difficulties using panel data, with mixed results (Brown 1980, Duncan and Holmlund 1983). Another approach is to use information on job duration – namely, to estimate preferences for specific job benefits by examining whether workers stay longer in jobs with those benefits (Gronberg and Reed 1994, Reed and Dahlquist 1994). However, it is not clear that duration models solve the key problem of unobserved heterogeneity, since an important identifying assumption they make is that the involuntary separation rate is the same across jobs with different benefits. More recently, a number of studies have estimated search models to identify the total value of non-wage attributes (Sullivan and To 2014, Hall and Mueller 2018) and the contribution of non-wage attributes to variation in observed wages (Taber and Vejlin 2016, Sorkin 2018). However, only a few of these papers distinguish between different types of attributes: Dey and Flinn (2005, 2008) focus on health insurance, while Bonhomme and Jolivet (2009) use self-reported satisfaction with five job characteristics. iii These conditions include extremes of heat, cold, wetness, vibrations or hazards. The authors also found positive but not statistically significant WTP for jobs that do not require frequent stooping, kneeling, crouching or crawling, and for jobs that are not repetitive. iv The job attributes covered in their study are type of work, working conditions, working times, distance to job/commuting time, and job security. v The survey was conducted by the BRAC Institute of Governance and Development (BIGD), BRAC University. vi The sample included 50 mouzas from Dhaka, 7 from Gazipur, 4 from Narayanganj and 19 from Chittagong. vii The replacement rules for the household selection were as follows: If the enumerator was unable to talk to someone at the selected household, or if the potential respondent was busy, the enumerator would return up to 2 more times (up to 3 visits). If the household could not be reached after 3 visits, or refused to participate, the enumerator would choose the household next door. If this first replacement was also unsuccessful, the 29 enumerator would select the household next door to that one. If the second replacement was also unsuccessful, the enumerator would go on to the next original household (i.e., 10th household). viii LFP rates are from the 2013 Labour Force Survey (Bangladesh Bureau of Statistics, 2015). ix Our definition of “working” does exclude individuals who were in the labor force but not working at the time of the screener. However, the unemployment rate in Bangladesh is low (4.3 percent in the 2013 Labor Force Survey). x The employment types were: Paid employee in government; Paid employee in a private entity; Apprentice/Intern/Trainee; Seasonal worker; Day laborer / casual worker; Domestic worker in a private household; Self-employed / business owner with no employees; Self-employed / business owner employing only paid or unpaid family members; Self-employed / business owner employing some non-family members; and Paid or unpaid family member working in a household business. xi For women who could not be located, we replaced them with men, since we had already targeted all working women in the roster. xii In the manuscript, we use the terms “written contract” and “contract” interchangeably. For employees who have a Provident Fund, both employers and employees make equal contributions (based on a percentage of the employee’s salary), and the fund is administered by a board of trustees. xiii Although the survey enumerators emphasized that the respondents should assume that all other attributes were identical, we cannot completely address concerns about unobserved heterogeneity that may arise because of respondents’ prior beliefs about the correlation between included and non-included attributes; for example, if a respondent believes that working longer hours makes her a more likely candidate to get a contract in future, then our measure of the preference for working hours is picking up some preference for a (future) contract. We thank an anonymous referee for pointing this out. xiv When we piloted the choice experiment during the focus groups, we experimented with both decreases and increases in monthly income, relative to current income. When we offered decreases in monthly income, we found that participants were unwilling to consider any cuts to current income, regardless of what benefits were offered. In contrast, they were willing to consider tradeoffs between differential income increases and benefits. We hypothesize that this difference is driven by the fact that many of the participants are living at or just above a subsistence income, and are therefore unwilling to consider any cuts to current income. Even 30 for participants with higher levels of income, loss aversion is likely to be an issue. Therefore, we limited the levels for monthly income to current income or increases from current income. In future work, it would be worth exploring what levels of nonwage benefits would induce workers to take even small wage cuts. xv Individual characteristics cannot be included directly in the conditional logit model, because these characteristics do not vary across alternatives; they can, however, be included when interacted with regressors that vary across alternatives. xvi We attempted to include all men from several small employment categories, as discussed in Section 2; however, since these categories were so small, they do not affect the overall distribution of characteristics among men, or create substantial differences between the weighted and unweighted results. xvii The level of female labor force participation implied by our roster is lower than the national average for Bangladesh (33 percent, according to the 2013 Labour Force Survey). However, our sample was focused on two urban and peri-urban cities, and is not meant to be representative of Bangladesh as a whole. xviii We asked workers who had multiple jobs to focus on the job in which they spent the most hours. xix As noted above, the weighting makes little difference when examining the distribution of employment types by gender, since all women received a sampling weight of one, and nearly all men received a sampling weight of 2.52. xx Individuals either provided us with an estimate of their monthly income, or a range into which their income fell. For those who provided a range, we used the midpoint of the range. We were unable to collect information about income (an estimate or a range) for 36 individuals. Another 21 reported zero income. xxi As a comparison, the World Bank estimates that annual per-capita income in Bangladesh was about $1,359 USD in 2016, or about $113 USD per month. xxii Among the 65 respondents in our sample who indicated that they were household members working in a family business, 6 reported earning no income. 11 self-employed respondents also reported zero income, as did 1 domestic worker, 1 day laborer, and 2 paid employees. In the case of the 4 wage workers who reported zero income, it is not clear if this reflected a failure to be paid by their employers recently, or a refusal to respond to the income question. In addition, income information was missing for 36 respondents. xxiii We thank an anonymous referee for this suggestion. 31 xxiv These cases were: in Column (1), the coefficient on 5 days’ leave was not significantly different from the coefficient on 10 days’ leave at conventional levels, and was different from the coefficient on 15 days’ leave at the 5 percent level; the coefficient on 10 days’ leave was different from the coefficient on 15 days’ leave at the 5 percent level; and a joint test of the differences between the three levels of leave was significant at the 5 percent level. In Columns (2) and (3), the coefficient on 15 days’ notice was different from the coefficient on 30 days’ notice at the 5 percent level, and the coefficient on 5 days’ leave was not significantly different from the coefficient on 10 days’ leave at conventional levels; however, a joint test of the differences between the three levels of leave was significant at the 1 percent level, as was a joint test of the differences between the three levels of notice. xxv While the labor laws in Bangladesh make it legally difficult to fire private sector workers, enforcement is rather weak across all types of firms, and firing can be commonplace, especially when the termination notice period is not explicitly specified in the contract. Corroborating this, in our focus groups, many participants explained that they value job stability, and would strongly prefer a job with termination notice specified in the contract, as this gives them a modicum of protection. xxvi We experimented with different numbers of classes and found that three classes produced reasonable results. Sampling weights are not applied in this case. xxvii We calculate this WTP by taking the difference between the WTP coefficients on fixed earnings (-27 USD) and on fixed earnings with a contract that could end at any time (-67 USD), and dividing by the median monthly earnings offered in the experiment (148 USD): (-27+67)/148=27%. 32 References Assy, Angela Elzir, Tiago Ribeiro, David Robalino, Furio Rosati, Maria Laura Sanchez Puerta and Michael Weber (2018). “The Jobs that Youth Want and the Support They Need to Get Them: Evidence from a Discrete Choice Experiment in Kenya.” Working Paper. 2018 Jobs and Development Conference. Bangladesh Bureau of Statistics (2015). Labour Force Survey: Bangladesh 2013. Bonhomme, Stephane, and Gregory Jolivet (2009). “The Pervasive Absence of Compensating Differentials.” Journal of Applied Econometrics 24(5): 763-795. Boxall, Peter C. and Wiktor L. Adamowicz (2002). “Understanding Heterogeneous Preferences in Random Utility Models: A Latent Class Approach.” Environmental and Resource Economics 23: 421-446. Brown, Charles (1980). “Equalizing Differences in the Labor Market.” Quarterly Journal of Economics 94(1): 113-134. Dey, Matthew, Christopher Flinn (2005). “An equilibrium model of health insurance provision and wage determination.” Econometrica 73: 571–627. Dey, Matthew, Christopher Flinn (2008). “Household search and health insurance coverage.” Journal of Econometrics 145: 43–63. Diamond, Peter A., and Jerry A. Hausman (1994). “Contingent Valuation: Is Some Number Better than No Number?” Journal of Economic Perspectives, 8 (4): 45-64. Duncan, Greg J. and Bertil Holmlund (1983). “Was Adam Smith Right After All? Another Test of the Theory of Compensating Wage Differentials.” Journal of Labor Economics 1(4): 366-379. 33 Gronberg, Timothy J. and W. Robert Reed (1994). “Estimating Workers' Marginal Willingness to Pay for Job Attributes Using Duration Data.” Journal of Human Resources 29(3): 911-931. Gutierrez, Italo A., Krishna B. Kumar, Minhaj Mahmud, Farzana Munshi, and Shanthi Nataraj (2019). “Transitions between informal and formal employment. Results from a worker survey in Bangladesh.” IZA Journal of Development and Migration 9(3). Hall, Robert E. and Andreas I. Mueller (2018). “Wage Dispersion and Search Behavior: The Importance of Nonwage Job Values.” Journal of Political Economy 126(4): 1594-1637. Hanley, Nick, Robert E. Wright and Vic Adamowicz (1998). “Using Choice Experiments to Value the Environment.” Environmental and Resource Economics 11(3-4): 413- 428. Hwang, Hae-shin, W. Robert Reed, and Carlton Hubbard (1992). “Compensating differentials and unobserved productivity.” Journal of Political Economy 100: 835– 858. Hwang, Hae-shin, Dale T. Mortensen, and W. Robert Reed (1998), Hedonic Wages and Labor Market Search, Journal of Labor Economics 16(4): 815-847. International Labour Organization (ILO, 2014). “Global Employment Trends 2014: The Risk of a Jobless Recovery.” http://www.ilo.org/global/research/global- reports/global-employment-trends/2014/lang--en/index.htm Kumar, Krishna, Minhaj Mahmud, Shanthi Nataraj and Yoonyoung Cho (2019). “Employer and Employee Preferences for Worker Benefits: Evidence from a 34 Matched Survey on the Bangladesh Informal Sector.” IZA Discussion Paper No. 12064. La Porta, Rafael and Andrei Shleifer (2008). “The Unofficial Economy and Economic Development.” Brookings Papers on Economic Activity, Fall 2008, no. 2. Maestas, Nicole, Kathleen J. Mullen, David Powell, Till von Wachter, and Jeffrey B. Wenger (2018). “The Value of Working Conditions in the United States and Implications for the Structure of Wages.” RAND Education and Labor Working Paper WR-1287. Mas, Alexandre and Amanda Pallais (2017). “Valuing Alternative Work Arrangements.” American Economic Review 107(12): 3722-3759. McFadden, D. (1973). “Conditional Logit Analysis of Qualitative Choice Behavior” in Frontiers in Econometrics, edited by P. Zarembka. New York: Academic Press, pp. 105-142. Reed, W. Robert and Julie Dahlquist (1994). “Do Women Prefer Women’s Work?” Applied Economics 26: 1133-1144. Scott, Anthony, Julia Witt, Christine Duffield, and Guyonne Kalb (2015). “What do nurses and midwives value about their jobs? Results from a discrete choice experiment.” Journal of Health Services Research & Policy 20(1): 31 - 38. Sorkin, Isaac (2018). “Ranking Firms Using Revealed Preference.” Quarterly Journal of Economics 133 (3): 1331–1393. Sullivan, Paul, and Ted To (2014). “Search and Nonwage Job Characteristics.” Journal of Human Resources 49(2): 472–507. 35 Swait, J. R. (1994). “A Structural Equation Model of Latent Segmentation and Product Choice for Cross-Sectional Revealed Preference Choice Data.” Journal of Retailing and Consumer Services 1: 77–89. Taber, Christopher and Rune Vejlin (2016). “Estimation of a Roy/Search/Compensating Differential Model of the Labor Market.” NBER Working Paper No. 22439, July 2016. Tybout, James. R. (2000). “Manufacturing Firms in Developing Countries: How Well Do They Do, and Why?” Journal of Economic Literature 38(1), 11-44. Ubach, Cristina, Anthony Scott, Fiona French, Morag Awramenko, and Gillian Needham (2003). “What do hospital consultants value about their jobs? A discrete choice experiment.| BMJ 326(7404): 1432 - 1438. Van Ommeren, Jos, Gerard J. van den Berg, and Cees Gorter (2000). “Estimating the marginal willingness to pay for commuting.” Journal of Regional Science 40: 541– 563. Wiswall, Matthew and Basit Zafar (2018). “Preference for the Workplace, Investment in Human Capital, and Gender.” Quarterly Journal of Economics 133(1): 457-507. 36 Figure 1. Empirical CDF of Minimum Percent Selecting One Scenario Source: Authors’ analysis based on survey data. Note: CDF of the minimum percent of respondents who chose one of the two alternatives. A minimum of 50 percent would indicate that half of the respondents selected each alternative. Strictly dominated choices are shown as triangles. 37 Figure 2: Latent Class Results – Assigned Class by Employment Type Source: Authors’ analysis based on survey data. Note: Percent of respondents assigned to each class in the latent class analysis, by type of employment at the time of the survey. Respondents were assigned to the class to which they had the highest posterior probability of belonging. 38 Figure 3: Latent Class Results – Assigned Class by Gender Source: Authors’ analysis based on survey data. Note: Percent of respondents assigned to each class in the latent class analysis, by gender. Respondents were assigned to the class to which they had the highest posterior probability of belonging. Table 1. Sample Selection Male Female Roster Target Sample Roster Target Sample All Adults (18+) 3,695 N/A N/A 3,544 N/A N/A Working Adults (18+) 3,040 1,254 1,270 747 747 694 Wage workers 1,770 749 708 561 561 505 Self-Employed workers 1,270 505 562 186 186 189 Source: Authors’ analysis based on survey data. Note: Number of adults in the initial survey roster, number targeted for sampling, and number included in the final sample, by gender. 39 Table 2. Example Choice Scenario JOB A JOB B Written Contract 6 months initially 1 year initially Termination Notice 15 days 15 days Working hours 30-45 hours per week 45-60 hours per week Amount of paid leave 10 days 10 days (not including major government holidays / festival leave) Provident Fund No No Monthly income 20% higher than your current 10% higher than your current monthly income from main monthly income from main economic activity economic activity Source: Authors’ selection of levels from choice experiment used in survey. Note: Example of a choice set presented to respondents during the survey. 40 Table 3. Full Set of Attributes and Levels Attribute Levels Written Contract None 6 months initially 1 year initially Long term Termination Notice None 15 days 30 days 60 days Working hours 30-45 hours per week 45-60 hours per week 60-75 hours per week Amount of paid leave None (excluding government holidays and festival 5 days leave) 10 days 15 days Provident Fund Yes No Monthly income Same as now 10% increase over current income 20% increase over current income 30% increase over current income 40% increase over current income 50% increase over current income Source: Authors’ choice experiment used in survey. Note: Full set of attributes and levels from the choice experiment presented to respondents during the survey. 41 Table 4. Summary Statistics for Respondents Panel (a): Basic Demographics Unweighted Weighted (%) (%) Female 35.3 18.0 Age 0-25 25.0 21.9 26-35 35.5 35.6 36-45 21.8 22.8 46-55 11.6 12.8 56+ 6.1 6.9 Education Less than primary 18.2 16.5 Some primary 23.4 22.8 Some secondary 19.0 19.6 Some high school 14.6 16.0 High school degree 9.0 9.5 Bachelors degree or higher 15.5 15.4 Missing 0.3 0.3 Industry Garments 25.4 20.4 Other Manufacturing 13.5 15.5 Trade/Transportation 22.9 27.8 Other services 38.3 36.3 Panel (b): Type of Employment Male Female Total % Unweighted Weighted Unweighted Weighted Paid employee in government 4.2 4.3 5.5 4.7 4.5 Paid employee in semi 1.6 1.7 2.7 2.0 1.9 government entity Paid employee in a private 38.0 38.3 49.9 42.2 40.4 entity Apprentice/Intern/Trainee 0.2 0.1 0.3 0.2 0.1 Seasonal worker 0.6 0.2 0.7 0.6 0.3 Day laborer / casual worker 10.3 10.4 3.2 7.8 9.1 Domestic worker in a private 0.8 0.3 10.5 4.2 2.1 household Self-employed / business owner 19.5 19.7 16.9 18.6 19.2 with no employees Self-employed / business owner 6.9 7.0 4.5 6.1 6.5 employing only paid or unpaid family members Self-employed / business owner 14.9 15.0 1.9 10.3 12.7 employing some non-family members Paid or unpaid family member 2.9 2.9 4.0 3.3 3.1 working in a household business Total 100 100 100 100 100 42 Panel (c): Benefits by Employment Type Government employees Private employees Casual workers % % % Unweighted Weighted Unweighted Weighted Unweighted Weighted Written 90.8 91.9 23.8 25.0 1.6 1.6 contract Verbal 4.6 5.0 48.7 48.2 54.4 51.3 Contract Sick leave 98.4 98.4 87.7 87.8 67.9 63.7 Casual leave 96.0 96.6 61.1 61.3 29.5 26.0 (paid vacation) Provident 91.5 92.9 12.3 12.7 0.0 0.0 Fund Termination 70.6 73.5 44.5 45.3 19.5 14.4 notice Panel (d): Monthly Income Monthly Earnings in 2016 Taka (USD in parentheses) Employment Type 25% percentile 50% percentile 75% percentile Unweighted Weighted Unweighted Weighted Unweighted Weighted Government 15,000 15,000 25,000 23,000 30,000 30,000 employee (191) (191) (319) (293) (383) (383) 6,000 6,913 8,000 9,000 15,000 15,000 Private employee (77) (88) (102) (115) (191) (191) 3,500 6,000 7,500 7,500 10,500 12,000 Casual worker (45) (77) (96) (96) (134) (153) Self-employed 6,500 7,500 11,500 15,000 15,000 20,000 alone/with family (83) (96) (147) (191) (191) (255) Self-employed (with 15,000 15,000 20,000 20,000 30,000 30,000 non-family) (191) (191) (255) (255) (383) (383) 2,500 3,500 7,500 7,500 15,000 15,000 Family worker (32) (45) (96) (96) (191) (191) Source: Authors’ analysis based on survey data. Note: Summary statistics for individuals in survey. Results are shown with and without sampling weights. In Panel (c), we only include wage workers who were able to provide an answer. In Panel (d), if respondents did not provide an exact monthly income, but did indicate the bin into which their income fell, we used the median value from the bin for their income. 43 Table 5. Conditional Logit Coefficient Estimates (1) (2) (3) (4) (5) (6) All All Pos. All Pos. Income All, Income No Weights Contract - 6 months 0.84*** 0.81*** 0.81*** 0.97*** 0.97*** 0.95*** (0.062) (0.061) (0.062) (0.055) (0.055) (0.051) Contract - 1 year 1.34*** 1.38*** 1.39*** 1.37*** 1.39*** 1.31*** (0.065) (0.065) (0.066) (0.060) (0.060) (0.055) Contract - long-term 2.39*** 2.36*** 2.37*** 2.22*** 2.24*** 2.14*** (0.097) (0.096) (0.097) (0.092) (0.093) (0.086) Notice - 15 days 1.09*** 1.16*** 1.16*** (0.055) (0.054) (0.054) Notice - 30 days 0.93*** 1.04*** 1.05*** (0.055) (0.049) (0.050) Notice - 60 days 1.44*** 1.48*** 1.49*** (0.067) (0.065) (0.066) Notice 0.020*** 0.020*** 0.020*** (0.00089) (0.00090) (0.00084) Hours - 45-60 -0.20*** -0.17*** -0.17*** (0.039) (0.038) (0.038) Hours - 60-75 -0.58*** -0.50*** -0.50*** (0.044) (0.040) (0.041) Hours -0.021*** -0.022*** -0.023*** (0.0013) (0.0013) (0.0013) Leave - 5 Days 0.28*** 0.26*** 0.25*** (0.056) (0.056) (0.057) Leave - 10 Days 0.28*** 0.31*** 0.31*** (0.051) (0.048) (0.049) Leave - 15 Days 0.38*** 0.44*** 0.44*** (0.045) (0.042) (0.043) Leave 0.027*** 0.027*** 0.025*** (0.0025) (0.0025) (0.0024) Provident Fund 0.87*** 0.88*** 0.88*** 0.91*** 0.91*** 0.87*** (0.060) (0.059) (0.059) (0.054) (0.055) (0.052) Income - 10 Perc. Inc. 0.37*** (0.065) Income - 20 Perc. Inc. 0.99*** (0.085) Income - 30 Perc. Inc. 1.79*** (0.10) Income - 40 Perc. Inc. 2.03*** (0.10) Income - 50 Perc. Inc. 2.59*** (0.099) Perc. Change Income 0.054*** 0.054*** 0.051*** 0.051*** 0.050*** (0.0019) (0.0020) (0.0018) (0.0018) (0.0017) Observations 23,568 23,568 22,884 23,568 22,884 23,568 Source: Authors’ analysis based on survey data. 44 Note: Standard errors clustered at the respondent level. Sampling weights are applied. *** p<0.01, ** p<0.05, * p<0.1. Where applicable, excluded levels are: contract (none), notice (none), hours (30-45), leave (none), Provident Fund (none), and income (no change). Table 6. Marginal Values of Attributes Coefficient Marginal Value Labour Law Marginal Value (1) in Terms of % Requirement x Labour Law Income (βk/βw) (3) Requirement (2) (4) Contract - 6 0.97 19.0 months Contract - 1 1.37 26.9 year Contract - 2.22 43.5 long-term Notice (days) 0.020 0.39 30 days 11.7 Hours (median) -0.021 -0.41 Leave (days) 0.027 0.53 10 days 5.3 Provident Fund 0.91 17.8 (Yes) Percent change 0.051 1.0 in income Source: Authors’ analysis based on survey data. Note: Coefficients and associated marginal values of each attribute relative to income. Marginal value is calculated by dividing coefficient on attribute by coefficient on percent change in income. Labour law is based on requirement for typical workers given in the 2006 Bangladesh Labour Law. 45 Table 7. Conditional Logit Coefficient Estimates by Gender and Education Male Female Less than Primary Secondary Coefficient (se) WTP Interaction (se) WTP Coefficient (se) WTP Interaction (se) WTP (1) (2) (3) (4) (5) (6) (7) (8) Contract - 6 months 0.99*** 18.7 -0.093 19.5 0.86*** 16.9 0.19* 20.7 (0.065) (0.11) (0.085) (0.11) Contract - 1 year 1.43*** 27.0 -0.30*** 24.5 1.15*** 22.5 0.37*** 30.0 (0.071) (0.11) (0.090) (0.12) Contract - long-term 2.31*** 43.6 -0.45** 40.3 1.81*** 35.5 0.68*** 49.1 (0.11) (0.18) (0.14) (0.19) Notice 0.020*** 0.38 -0.0019 0.39 0.018*** 0.35 0.0028 0.41 (0.0010) (0.0018) (0.0014) (0.0018) Hours -0.020*** -0.38 -0.0093*** -0.64 -0.020*** -0.39 -0.0026 -0.45 (0.0015) (0.0027) (0.0021) (0.0027) Leave 0.028*** 0.53 -0.0066 0.46 0.028*** 0.55 -0.0028 0.50 (0.0029) (0.0049) (0.0040) (0.0051) Provident Fund 0.94*** 17.7 -0.20* 16.1 0.93*** 18.2 -0.032 17.7 (0.063) (0.11) (0.091) (0.11) Perc. Change Income 0.053*** 1.0 -0.0069** 1.0 0.051*** 1.0 -0.00030 1.0 (0.0021) (0.0035) (0.0029) (0.0037) Observations 23,568 23,496 Source: Authors’ analysis based on survey data. Note: WTP= willingness to pay. Columns (1) and (5) show the baseline coefficients for men and for workers with primary or lower education, while Columns (3) and (7) show interactions for women and for workers with a secondary or higher education, respectively. The WTP for each attribute for women is calculated by summing the coefficients on that attribute in Columns (1) and (3), respectively, and dividing by the sum of the coefficients on income in Columns (1) and (3). A 46 similar calculation is used for those with a secondary or higher education level. Omitted level for contract is none. Omitted level for Provident Fund is none. Standard errors are clustered at the respondent level. *, ** and *** represent statistical significance at the 10%, 5% and 1% levels, respectively. Table 8: Conditional Logit Coefficient Estimates by Employment Type Private employee Govt employee Casual worker Self-employed Coefficient (se) WTP Interaction (se) WTP Interaction (se) WTP Interaction (se) WTP (1) (2) (3) (4) (5) (6) (7) (8) Contract - 6 months 1.00*** 18.5 0.22 24.9 -0.0074 16.6 -0.090 19.0 (0.085) (0.23) (0.18) (0.12) Contract - 1 year 1.47*** 27.2 0.36 37.3 -0.042 23.9 -0.26* 25.3 (0.090) (0.27) (0.18) (0.13) Contract - long-term 2.33*** 43.1 0.88** 65.5 0.066 40.1 -0.37* 41.0 (0.14) (0.41) (0.30) (0.20) Notice 0.021*** 0.39 0.0026 0.48 0.0015 0.38 -0.0026 0.38 (0.0014) (0.0040) (0.0030) (0.0020) Hours -0.023*** -0.43 -0.0083 -0.64 -0.0039 -0.45 0.0047 -0.38 (0.0020) (0.0058) (0.0044) (0.0028) Leave 0.029*** 0.54 -0.0049 0.49 -0.0037 0.42 -0.0031 0.54 (0.0039) (0.011) (0.0084) (0.0055) Provident Fund 0.98*** 18.1 -0.038 19.2 -0.026 16.0 -0.15 17.4 (0.086) (0.23) (0.18) (0.12) Perc. Change Income 0.054*** 1.0 -0.0050 1.0 0.0058 1.0 -0.0062 1.0 (0.0028) (0.0082) (0.0057) (0.0040) Observations 23,568 Source: Authors’ analysis based on survey data. 47 Note: WTP= willingness to pay. Column (1) shows the baseline coefficients for private employees, while Columns (3), (5) and (7) show interactions for government employees, casual workers, and self-employed workers, respectively. The WTP for each attribute for government employees is calculated by summing the coefficients on that attribute in Columns (1) and (3), respectively, and dividing by the sum of the coefficients on income in Columns (1) and (3). A similar calculation is used for casual workers and self-employed. Omitted level for contract is none. Omitted level for Provident Fund is none. Standard errors are clustered at the respondent level. *, ** and *** represent statistical significance at the 10%, 5% and 1% levels, respectively. 48 Table 9: Latent Class Preference Estimates Class 1 Class 2 Class 3 (“Stability (“Income (“Long-Term Seeking”) Seeking”) Planners”) Coeff Coeff Coeff (se) WTP (se) WTP (se) WTP (1) (2) (3) (4) (5) (6) Contract - 6 months 2.62*** 44.4 0.55*** 5.7 0.86*** 29.7 (0.49) (0.19) (0.25) Contract - 1 year 3.87*** 65.6 1.24*** 12.8 1.05*** 36.2 (0.84) (0.19) (0.19) Contract - long-term 6.27*** 106.3 1.83*** 18.9 2.2*** 75.9 (0.84) (0.54) (0.25) Notice 0.017*** 0.29 0.019*** 0.20 0.034*** 1.17 (0.003) (0.002) (0.004) Hours -0.028*** -0.47 -0.041*** -0.42 -0.028*** -0.97 (0.01) (0.006) (0.005) Leave 0.004 0.07 0.056*** 0.58 0.018 0.62 (0.018) (0.008) (0.011) Provident Fund 0.73 12.4 0.75*** 7.7 1.64*** 56.6 (0.78) (0.23) (0.31) Perc. Change Income 0.059** 1.0 0.097*** 1.0 0.029** 1.0 (0.029) (0.007) (0.013) Class share 0.198 0.467 0.335 # (%) workers assigned to 367 949 648 class (18.7%) (48.3%) (33.0%) Source: Authors’ analysis based on survey data. Note: WTP= willingness to pay. Results from a latent class analysis. Sampling weights are not applied. Omitted level for contract is none. Omitted level for Provident Fund is none. *, ** and *** represent statistical significance at the 10%, 5% and 1% levels, respectively. 49 Appendix A1: Class Membership Parameter Estimates from Latent Class Analysis Table A1.1 shows results for the class membership parameters. In estimating the probability that an individual was in each class, we used gender, age, education, and the size of the enterprise in which the individual was working at the time of the survey. Class 3 is the omitted class, and all class membership parameters are normalized to zero for this class. Table A1.1: Latent Class Membership Results Class 1 Class 2 Female -0.36 -0.078 (0.63) (0.18) Age 0.01 0.002 (0.01) (0.006) Some Primary 0.24 0.0005 (0.49) (0.225) Some Secondary 0.26 -0.2 (0.63) (0.256) Some High School 1.07** -0.28 (0.43) (0.28) High School 0.31 -0.84*** (0.47) (0.29) Bachelors 0.85** -0.72*** (0.42) (0.27) Missing education 0.81 -0.59 (1.45) (1.31) 6-10 employees 0.84** 0.47 (0.4) (0.31) 11+ employees 0.62** 0.2 (0.25) (0.17) Missing enterprise size 1.84 0.26 (1.8) (1.73) Constant -1.66* 0.44 (0.97) (0.38) Source: Authors’ analysis based on survey data. Note: Class 3 is the omitted class. Omitted education level is none; omitted enterprise size is 1-5 employees. Sampling weights are not applied. *, ** and *** represent statistical significance at the 10%, 5% and 1% levels, respectively. 50 Supplementary Appendix S1: Further Notes on the Choice Experiment Design The choice experiment design involves creating choice sets by combining levels of attributes into choice alternatives. In a choice experiment, respondents make repeated choices between alternatives. However, given the number of attributes and the number of levels of each attribute, it is rarely feasible to include all potential choice sets (known as a full factorial design), as the task would be too cognitively demanding or would require too large a sample size. Therefore, researchers typically use fractional factorial designs, which include only a subset of the all potential choice situations (Louviere et al. 2000). A commonly used fractional factorial design is the orthogonal design, which seeks to minimize the correlation between the attribute levels in the choice situations, in order to allow estimation of all parameters. If prior information is known about the parameters, then an efficient design – which seeks to minimize the predicted standard errors of the parameter estimates – can outperform an orthogonal design (Carlsson and Martinsson 2003). Conceptually, an efficient design aims to minimize the standard errors of the parameters by using as much information as possible. In addition to orthogonality, a number of different criteria can contribute to an efficient design: these include minimal overlap, attribute level balance, and utility balance (Huber and Zwerina 1996). Minimal overlap is the concept of minimizing the chance that an attribute level appears more than once in a choice set, as this choice will therefore provide no information about the attribute. Attribute level balance occurs when all levels of each attribute appear an equal number of times in the design. Utility balance seeks to avoid choice sets in which one alternative clearly dominates the other, as this will not provide any information. However, it is worth noting that if both choice situations offer equal utility, this may result in respondents making random choices, which also does not provide information about preferences. Therefore, some degree of (but not perfect) utility balance provides the most information. As discussed in the text, our final design did include three choice situations in which one alternative likely strictly dominated the other. While these three situations provided little information about trade-offs, they did serve as a check on whether respondents were in fact making reasonable decisions. In practice, there are a variety of ways in which efficient designs can be developed. We selected the most commonly used efficiency method – the D-efficient design – which aims to minimize the determinant of the estimated asymptotic variance-covariance (AVC) matrix of the experimental design. To estimate the AVC matrix, we need to have priors about the parameter estimates (Hensher et al. 2015). While it is possible to assume that the parameters equal zero, using reasonable, non-zero priors can improve efficiency (Carlsson and Martinsson 2003). Thus, to design the experiment for the pilot phase of the survey, we used reasonable estimates based on theoretical priors about the signs and likely magnitudes of the parameters. For the design of the final survey, we relied on the parameter estimates from the pilot survey as priors. We used Ngene software by Choice Metrics to create the efficient design for our choice experiment. Below, we outline the specific design choices that we made, and we provide sufficient information so that other researchers can produce a similar design. However, as Hensher et al. (2015) note, given the number of potential choice situations, it is generally not possible to examine all potential designs, so the software uses an algorithm to cycle through many different experimental designs, and to select the one with the smallest D-error. Thus, the exact choice situations produced may differ to some extent, even if the same approach is followed. Model type: We selected the standard discrete choice model, namely the multinomial logit (MNL). Type of efficient design: As discussed above, we selected a D-efficient design, which seeks to minimize the D-error based on the determinant of estimated AVC matrix. Number of alternatives: We specified that each choice situation would have two alternatives (Job A and Job B). Number of choice situations and blocking: The efficiency of the design increases with the number of choice situations as additional situations provide more information; however, as noted above, it is generally not feasible to include all possible situations in any given experiment. While there is no specific guideline about the maximum number of choice situations that should be used, it is desirable that as many combinations as feasible be offered to respondents. To minimize cognitive burden, this is typically done by blocking the choice situations into smaller subsets, and offering each individual only one subset of choices. Given that we aimed to conduct the choice experiment with approximately 2,000 workers, we created 48 choice situations, which were divided into 8 blocks, each with 6 choice situations. The choice situations were randomly assigned within blocks, and each respondent was randomly assigned to a block. This resulted in approximately 250 respondents being assigned to each block. Constraints: In the design, we also used constraints on the attributes so that unreasonable situations could be avoided. For example, we did not allow a Provident Fund (pension) to be offered in jobs with a short contract duration. These constraints were based on input from the focus groups regarding which benefit combinations were likely to be viewed as unrealistic. The specific constraints are shown in the code below. The D-error that resulted from our design was 0.065007. Design Code / Specify the alternatives and their names, and then specify number of choice situations (48) and the number of blocks (8) / ;alts= job1, job2 ;rows=48 ;block=8 / Specify the model (MNL) and the efficiency measure (D-efficient) to be used / ;eff=(mnl,d) / Specify constraints on some attributes / ;cond: / If Provident Fund is offered, contract is long-term / if ( job1.pfc=1, job1.contract=3), if ( job2.pfc=1, job2.contract=3), / If contract is long-term, some termination notice must be given / if ( job1.contract=3, job1.notice=[1,2,3]), if ( job2.contract=3, job2.notice=[1,2,3]), / If contract is long-term, at least 10 days of leave must be given / if (job1.contract=3, job1.leave>1), if ( job2.contract=3, job2.leave>1), / Specify the utility functions. Prior parameters are shown in square brackets after each parameter (e.g., b2) and attribute levels are shown in square brackets after each attribute name (e.g., 0,1,2,3). Note that the prior parameter estimates and attribute levels only need to be defined for the first job, and are automatically applied to the second job. / ;model: u(job1)= b1[-0.1] + b2[1.1] * contract[0,1,2,3] + b3[0.8] * notice[0,1,2,3] + b4 [1.1] * leave [0,1,2,3] + b5[0.4] * hours [1, 2, 3] + b6[0.1] * pfc[0,1] + b7[1.5]* salary[0,1,2,3,4,5] / u(job2)= b2 * contract + b3 * notice + b4 * leave + b5 * hours + b6* pfc + b7*salary $ References Carlsson, Fredrik and Peter Martinsson (2003). “Design Techniques for Stated Preference Methods in Health Economics.” Health Economics 12: 281-294. Hensher, David A., John M. Rose, and William H. Greene (2015). Applied Choice Analysis. Cambridge: Cambridge University Press. Huber, Joel and Klaus Zwerina (1996). “The Importance of Utility Balance in Efficient Choice Designs.” Journal of Marketing Research XXXIII: 307-317. Louviere, Jordan J., David A. Hensher, and Joffre D. Swait (2000). Stated Choice Methods— Analysis and Application. Cambridge University Press, UK. Supplementary Appendix S2 BANGLADESH SURVEY OF FORMAL AND INFORMAL JOBS Conducted by RAND, USA and BIGD, BRAC University IDENTIFICATION OF THE SAMPLE AREA Sample Area Particulars Name Code No. Zila Upazila/Thana Union/Ward Mouza/Moholla PSU Number Questionnaire Number Interviewer Name Supervisor Name Name of Person Entering Data Date of Interview ENUMERATOR: PROVIDE INTRODUCTION / READ INFORMED CONSENT Will you be interested to participate in the second part of the survey, today? 1. Yes I would like to participate today 2. Yes I would like to participate another day: Date and Time 3. No, I would not like to participate Reason: ……………………………………………………………………………………………………………………… ……………………………………………………………………………………………………………………… ……………………………………………………………………………………………………………………… ……………………………………………………………………………………………………………………… ……………………………………………………………………………………………………….. C. JOB INFORMATION OF THE RESPODENT ENUMERATOR READS: FIRST, FOR THE PURPOSE OF RESEARCH ONLY, I WOULD NOW LIKE TO ASK YOU ABOUT YOUR EDUCATION, AND ABOUT YOUR CURRENT AND PAST WORK FROM WHICH YOU EARN YOUR MAIN INCOME. FIRST, I WILL START WITH SOME BASIC QUESTIONS ABOUT YOUR EDUCATION AND TRAINING. C1a_education What is the highest class you have passed? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST] 0. Pre-School 1. Class 1 2. Class 2 3. Class 3 4. Class 4 5. Class 5 6. Class 6 7. Class 7 8. Class 8 9. Class 9 10. Class 10 11. SSC 12. HSC 13. Diploma 14. Bachelor’s degree 15. Master’s degree/MPhil 16. PhD 17. No class passed 888. Don’t know 999. Refused C1b_vocationformal Have you received any vocational training as part of your formal education? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST] [MARK ALL THAT APPLY] 1. Yes, vocational SSC 2. Yes, vocational HSC 3. Yes, Business Management HSC 4. Yes, diploma-level course (for example, through Textile Institute) 5. Other (specify) 6. No 888. Don’t know 999. Refused C1c_vocationalother Have you received any vocational training outside the formal education system? 1. Yes 2. No à Go to C1f 888. Don’t know à Go to C1f 999. Refused à Go to C1f C1d_vocationlength For how long did you attend this vocational training outside the formal education system? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST. IF ATTENDED MORE THAN ONE TRAINING, PROVIDE THE LENGTH OF THE LONGEST TRAINING.] 1. Less than 1 week 2. 1-2 weeks 3. 2-4 weeks 4. 1-3 months 5. 3-5 months 6. More than 6 months 888. Don’t know 999. Refused C1e_vocationcert Did you receive a certificate for this vocational training outside the formal education system? 1. Yes 2. No 888. Don’t know 999. Refused C1f_numberemp How many jobs have you held over the last 15 years? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] CURRENT MAIN JOB ENUMERATOR READS: NOW, I WOULD LIKE YOU TO THINK ABOUT YOUR CURRENT MAIN WORK. BY MAIN WORK, I MEAN THE WORK WHERE YOU SPENT THE MOST TIME/HOURS IN THE PAST ONE MONTH. C2_emptype1 What describes the employment status of the MAIN job that you held over the past month? 1. Paid employee in government 2. Paid employee in semi-government 3. Paid employee in a private entity 4. Apprentice/Intern/Trainee 5. Seasonal worker 6. Day laborer / casual worker 7. Domestic worker in a private household 8. Self-employed / business owner with no employees 9. Self-employed / business owner employing only paid or unpaid family members 10. Self-employed / business owner employing some non-family members 11. Paid or unpaid family member working in a household business 888. Don’t know 999. Refused 1 C3_location1 Where do you mainly undertake your MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. At my home with no special work space 2. At my home with work space inside/ attached to the home 3. Factory, workshop, shop, kiosk, etc 4. Farm or individual agricultural plot 5. Home or workplace of client 6. Construction site 7. Market, bazaar stall, trade fair 8. Street pavement/ highway with fixed post 9. Office building 10. Employer’s home 11. Transport vehicle 12. No fixed location (e.g., mobile; door-to-door; street without fixed post) 13. Other (specify) 888. Don’t know 999. Refused C4_industry1 What is the industry of this MAIN job? [LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM INDUSTRY LIST. IF NOT CLEAR THEN ASK FOR MORE DETAILS. IF OPTION IS NOT ON THE LIST THEN WRITE DOWN WHAT THE RESPONDENT SAYS.] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C5_ownership1 What is the type of ownership of the enterprise in this MAIN job? 1. Government 2. Semi-government 3. NGO 4. Private enterprise 5. Small business 6. Private household 7. Other (Specify) 888. Don’t know 999. Refused C6_occupation1 What type of activity do you do in this MAIN job? [LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM OCCUPATION LIST. IF NOT CLEAR THEN ASK FOR MORE DETAILS. IF OPTION IS NOT ON THE LIST THEN WRITE DOWN WHAT THE RESPONDENT SAYS.] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] 2 C7a_startyear1 What year did you start this MAIN job? [YYYY] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C7b_startmonth1 In which month of the year did you start this MAIN job? 1. Jan 2. Feb 3. Mar 4. Apr 5. May 6. Jun 7. Jul 8. Aug 9. Sep 10. Oct 11. Nov 12. Dec 888. Don’t know 999. Refused C8_howfind1 How did you find this MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. By starting my own business 2. Contacted employer directly 3. Through a public employment agency 4. Through a private employment agency 5. Through friends or relatives 6. Through school/university employment center 7. By responding to a vacancy 8. Through a union/professional register 9. Through job training programs/courses 10. Other (specify) 888. Don’t know 999. Refused C9_numemp1 About how many persons (including yourself) work in the enterprise or household business where you are employed in this MAIN job? 1. 1 2. 2-5 3. 6-9 4. 10-25 5. 26-50 6. 51-100 7. 101-249 8. More than 250 888. Don’t know 999. Refused 3 C10_hours1 How many hours (on average) do you work during a normal WEEK in this MAIN job? 1. Fewer than 15 hours per week 2. 15-30 hours per week 3. 30-45 hours per week 4. 45-60 hours per week 5. 60-75 hours per week 6. More than 75 hours per week 888. Don’t know 999. Refused C11_hourspref Provided that you could make a free choice regarding your working hours and taking into account the need to earn a living: how many hours would you prefer to work during a normal WEEK in this MAIN job? 1. Fewer than 15 hours per week 2. 15-30 hours per week 3. 30-45 hours per week 4. 45-60 hours per week 5. 60-75 hours per week 6. More than 75 hours per week 888. Don’t know 999. Refused C12a_earnings_cash1 How much do you usually earn PER MONTH in cash from this MAIN job (in Taka)? [FOR SELF-EMPLOYED, EARNINGS REFER TO PROFITS, NOT REVENUES. FOR FAMILY MEMBERS WORKING IN A HOUSEHOLD BUSINESS, RESPONDENT SHOULD SAY HOW MUCH HE OR SHE EARNS, NOT HOW MUCH THE BUSINESS EARNS.] [If answered then go to C12c] [If don’t know then put “Don’t Know” and go to C12b] [If refused then put “Refused” and go to C12b] C12b_earnings_cash_range1 In what range does the amount you usually earn PER MONTH from this MAIN job fall? 1. Less than 5,000 Taka 2. 5,000-10,000 Taka 3. 10,000-20,000 Taka 4. 20,000-30,000 Taka 5. 30,000 Taka or more 888. Don’t know 999. Refused C12c_earnings_kind1 How much do you usually earn per month in kind (for example, free meals or clothing) from this MAIN job (estimated Taka equivalent)? [If no in-kind earnings, then put “0”, do not leave blank] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] 4 ENUMERATOR: IF C2_emptype1=1, 2, 3, 4, 5, 6 OR 7 (THAT IS, IF THE RESONDENT INDICATED THAT CURRENT JOB IS A PAID JOB OR APPRETICESHIP WORKING FOR SOMEONE ELSE), FILL OUT C14-C22 FOR CURRENT JOB. IF C2_emptype1=8, 9, 10 OR 11 (THAT IS, IF THE RESONDENT INDICATED THAT CURRENT JOB IS IN SELF-EMPLOYMENT OR AS A PAID OR UNPAID FAMILY WORKER IN A HOUSEHOLD BUSINESS), GO TO C24 FOR CURRENT JOB. C14a_contract1 Do you have a job contract or formal letter of agreement in this MAIN job? 1. Yes, permanent contract à Go to C15a 2. Yes, contract that is not permanent à Go to C14c 3. No 888. Don’t know 999. Refused C14b_verbal1 Do you have a verbal agreement about terms and conditions in this MAIN job? 1. Yes, verbal agreement that includes permanent appointment à Go to C15a 2. Yes, verbal agreement that does not include permanent appointment à Go to C14c 3. No 888. Don’t know 999. Refused C14c_contractlength1 What is the length of your contract or verbal agreement? [If length not specified in contract or verbal agreement, put “Not specified” and go to C15a] [If don’t know then put “Don’t Know” and go to C15a] [If refused then put “Refused” and go to C15a] C14d_contractunit1 [Specify unit in which length of contract or agreement is given] 1. days 2. weeks 3. months 4. years 888. Don’t know 999. Refused Go to à C15a 5 C14e_temp1 What is the main reason that you do not have a permanent contract or permanent verbal agreement in this MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT ALL THAT APPLY.] 1. This employer did not offer me a job with a permanent contract 2. I did not want to have a job with a permanent contract 3. I am unable to do continuous work 4. I was not able to find any job that offered a permanent contract or verbal agreement 5. Other (specify) 888. Don’t know 999. Refused C15a_sick1 How many days of sick leave does your employer provide PER YEAR in this MAIN job? [If “As needed” then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C15b_casual1 How many days of PAID earned or casual leave does your employer provide PER YEAR in this MAIN job? [If “As needed” then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C15c_holiday1 How many days of holiday does your employer provide EACH WEEK in this MAIN job? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16a_maternityactual1 Have you taken maternity leave from this MAIN job? 1. Yes 2. No à Go to C17 888. Don’t know à Go to C17 999. Refused à Go to C17 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16b_maternityactuallength1 How many weeks of maternity leave (paid or unpaid) did you take? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16c_maternityactualpaid1 How many weeks of those weeks were paid? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] à Go to C18 6 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17a_maternity1 Would your employer allow you to take maternity leave from this MAIN job if you need it? 1. Yes 2. No à Go to C18 888. Don’t know à Go to C18 999. Refused à Go to C18 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17b_maternitylength1 How many weeks of maternity leave (paid or unpaid) would your employer allow you to take if you needed it? [If as needed then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17c_maternitypaid1 For how many of those weeks would your employer pay you? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C18_overtime1 Does your employer give you extra pay for overtime work in this MAIN job? 1. Yes 2. No 3. Do not work overtime 888. Don’t know 999. Refused C19_bonus1 Does your employer offer any bonus payments in this MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C20a_providentfund1 Does your employer offer a Provident Fund in this MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C20b_gratuity1 Does your employer offer gratuity (lump sum) in this MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused 7 C20c_pension1 Does your employer offer a pension in this MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C21d_termination1 If you are terminated from this MAIN job, will you be given notice? 1. Yes 2. No à Go to C22 888. Don’t know à Go to C22 999. Refused à Go to C22 C21e_termlength1 How many days of notice will you be given? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C21f_termpay1 How many weeks of pay will you receive on termination? [If don’t know then put “Don’t Know” and go to C22] [If refused then put “Refused” and go to C22] C22_training1 If you need to learn more skills to do this MAIN job, where do you get them from? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. My employer provides training for me 2. I find the information I need on my own 3. I learn from co-workers or on-the-job 4. I learn from family and friends 5. I have not had to learn more skills to do my job 888. Don’t know 999. Refused 8 ENUMERATOR: IF C2_emptype1=8, 9, 10 OR 11 (IF THE RESONDENT INDICATED THAT CURRENT JOB IS IN SELF-EMPLOYMENT OR AS A PAID OR UNPAID FAMILY WORKER IN A HOUSEHOLD BUSINESS), FILL OUT C24-C31 FOR CURRENT JOB. OTHERWISE GO TO PREVIOUS JOB #1. C24_selfemptime1 How long have you been running or working in this business? 1. Up to 6 months 2. 7-12 months 3. 1-3 years 4. 3-5 years 5. 5-10 years 6. More than 10 years 888. Don’t know 999. Refused C25a_registered1 Is this business registered in any national or local agency? 1. Yes 2. In the process of being registered 3. No à Go to C26 888. Don’t know à Go to C26 999. Refused à Go to C26 C25b_registeredorg1 Which organization is the business registered with (or in the process of registering with)? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Joint Stock Company 2. Board of Investment (BOI) 3. Bangladesh Small and Cottage Industries Corporation (BSCIC) 4. Bangladesh Export Processing Zones Authority (BEPZA) 5. Department of Inspection for Factories and Establishments (DIFE) 6. Cooperatives 7. City Corporation / Municipalities / Union Parishad (UP) 8. NGO Affairs Bureau 9. Other (Specify) 888. Don’t know 999. Refused 9 C26_whyselfemp1 Why did you start or why do you work in this business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Because I was unable find other work 2. Because I wanted independence/ be my own boss 3. Because I wanted to have flexible working hours 4. Because I wanted higher income 5. Because I wanted to be close to home 6. Because parents, relatives or friends have their own business 7. Because I wanted do grow professionally 8. Because I wanted to have job security 9. Because family members wanted me to work in the business 10. Other (specify) 888. Don’t know 999. Refused C27_difficulty1 What is the main difficulty in running this enterprise/ business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. Acquisition of capital/credit 2. Lack of knowledge 3. Registration 4. Legal requirements 5. Acquisition of new technologies 6. Finding qualified workers 7. Acquisition or rent of a place 8. Difficulty getting paid for good or services provided 9. Lack of electricity 10. Lack of fuel 11. Lack of other infrastructure (roads, buildings) 12. No difficulties 13. Other (specify) 888. Don’t know 999. Refused C28_familyemp1 Please tell me the exact number of family members who work in this business (not including yourself) [Do not leave blank. If 0 then put “0”. If don’t know then ask respondent to estimate. If still don’t know then put “Don’t know”; if refused then put “Refused”. Please check that anyone who reports being self-employed with no employees (C2_emptype1=8) SHOULD NOT report anyone here. Anyone who reports being self-employed and employing paid or unpaid family members (C2_emptype1=9) SHOULD report someone here.] C29_nonfamilyemp1 Please tell me the exact number of nonfamily members who work in the business [Do not leave blank. If 0 then put “0”. If don’t know then ask respondent to estimate. If still don’t know then put “Don’t know”; if refused then put “Refused”. Please check that anyone who reports being self-employed with no employees or only employing family members (C2_emptype1=8 or 9) SHOULD NOT report anyone here. Anyone who reports being self- employed with some non-family members (C2_emptype1=10) SHOULD report someone here.] 10 C30_trainingself1 If you need to learn more skills to run or work in this business, where do you get them from? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. I take formal training courses 2. I find the information I need on my own 3. I learn from co-workers or on-the-job 4. I learn from family and friends 5. I have not had to learn more skills to run or work in this business 888. Don’t know 999. Refused 11 PREVIOUS MAIN JOB #1 [ENUMERATOR: ASK ABOUT PREVIOUS MAIN JOB #1 IF THE RESPONDENT HAS BEEN AT THE CURRENT JOB FOR LESS THAN 15 YEARS (IF THE START DATE FOR CURRENT MAIN JOB, C7a_startyear1, IS AFTER 2000. OTHERWISE GO TO SECTION D.] ENUMERATOR READS: NOW I WOULD LIKE TO ASK YOU ABOUT THE MAIN JOB YOU HELD BEFORE YOUR CURRENT JOB. C2_emptype2 What describes the employment status of the MAIN job you held just before your current main job? 1. Paid employee in government 2. Paid employee in semi-government 3. Paid employee in a private entity 4. Apprentice/Intern/Trainee 5. Seasonal worker 6. Day laborer / casual worker 7. Domestic worker in a private household 8. Self-employed / business owner with no employees 9. Self-employed / business owner employing only paid or unpaid family members 10. Self-employed / business owner employing some non-family members 11. Paid or unpaid family member working in a household business 888. Don’t know 999. Refused C3_location2 Where did you mainly undertake your previous MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. At my home with no special work space 2. At my home with work space inside/ attached to the home 3. Factory, workshop, shop, kiosk, etc 4. Farm or individual agricultural plot 5. Home or workplace of client 6. Construction site 7. Market, bazaar stall, trade fair 8. Street pavement/ highway with fixed post 9. Office building 10. Employer’s home 11. Transport vehicle 12. No fixed location (e.g., mobile; door-to-door; street without fixed post) 13. Others, specify 888. Don’t know 999. Refused C4_industry2 What was the industry of this previous MAIN job? [LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM INDUSTRY LIST. IF NOT CLEAR THEN ASK FOR MORE DETAILS. IF OPTION IS NOT ON THE LIST THEN WRITE DOWN WHAT THE RESPONDENT SAYS.] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C5_ownership2 What was the type of ownership of the enterprise in this previous MAIN job? 1. Government 2. Semi-government 3. NGO 4. Private enterprise 5. Small business 6. Private household 7. Other (Specify) 888. Don’t know 999. Refused C6_occupation2 What type of activity did you do in this previous MAIN job? [LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM OCCUPATION LIST. IF NOT CLEAR THEN ASK FOR MORE DETAILS. IF OPTION IS NOT ON THE LIST THEN WRITE DOWN WHAT THE RESPONDENT SAYS.] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C7a_startyear2 What year did you start this previous MAIN job? [YYYY] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C7b_startmonth2 In which month of the year did you start this previous MAIN job? 1. Jan 2. Feb 3. Mar 4. Apr 5. May 6. Jun 7. Jul 8. Aug 9. Sep 10. Oct 11. Nov 12. Dec 888. Don’t know 999. Refused 1 C8_howfind2 How did you find this previous MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. By starting my own business 2. Contacted employer directly 3. Through a public employment agency 4. Through a private employment agency 5. Through friends or relatives 6. Through school/university employment center 7. By responding to a vacancy 8. Through a union/professional register 9. Through job training programs/courses 10. Other 888. Don’t know 999. Refused C9_numemp2 About how many persons (including yourself) worked in the enterprise or household business where you were employed in this previous MAIN job? 1. 1 2. 2-5 3. 6-9 4. 10-25 5. 26-50 6. 51-100 7. 101-249 8. More than 250 888. Don’t know 999. Refused C10_hours2 How many hours (on average) did you work during a normal WEEK in this previous MAIN job? 1. Fewer than 15 hours per week 2. 15-30 hours per week 3. 30-45 hours per week 4. 45-60 hours per week 5. 60-75 hours per week 6. More than 75 hours per week 888. Don’t know 999. Refused C12a_earnings_cash2 How much did you usually earn PER MONTH in cash from this previous MAIN job (in Taka)? [FOR SELF-EMPLOYED, EARNINGS REFER TO PROFITS, NOT REVENUES. FOR FAMILY MEMBERS WORKING IN A HOUSEHOLD BUSINESS, RESPONDENT SHOULD SAY HOW MUCH HE OR SHE EARNED, NOT HOW MUCH THE BUSINESS EARNED.] [If answered then go to C12c] [If don’t know then put “Don’t Know” and go to C12b] [If refused then put “Refused” and go to C12b] 2 C12b_earnings_cash_range2 In what range did the amount you usually earned PER MONTH from this previous MAIN job fall? 1. Less than 5,000 Taka 2. 5,000-10,000 Taka 3. 10,000-20,000 Taka 4. 20,000-30,000 Taka 5. 30,000 Taka or more 888. Don’t know 999. Refused C12c_earnings_kind2 How much did you usually earn per month in kind (for example, free meals or clothing) from this previous MAIN job (estimated Taka equivalent)? [If no in-kind earnings, then put “0”, do not leave blank] [If don’t know then put “Don’t Know”] [If refused then put “Refused”] ENUMERATOR: IF C2_emptype2=1, 2, 3, 4, 5, 6 OR 7 (THAT IS, IF THE RESPONDENT INDICATED THAT PREVIOUS MAIN JOB #1 WAS A PAID JOB OR APPRETICESHIP WORKING FOR SOMEONE ELSE), FILL OUT C14-C23 FOR PREVIOUS MAIN JOB #1. IF C2_emptype2=8, 9, 10 OR 11 (THAT IS, IF THE RESPONDENT INDICATED THAT PREVIOUS MAIN JOB #1 WAS IN SELF-EMPLOYMENT OR AS A PAID OR UNPAID FAMILY WORKER IN A HOUSEHOLD BUSINESS), GO TO C24 FOR PREVIOUS MAIN JOB #1. C14a_contract2 Did you have a job contract or formal letter of agreement in this previous MAIN job? 1. Yes, permanent contract à Go to C15a 2. Yes, contract that is not permanent à Go to C14c 3. No 888. Don’t know 999. Refused C14b_verbal2 Did you have a verbal agreement about terms and conditions in this previous MAIN job? 1. Yes, verbal agreement that included permanent appointment à Go to C15a 2. Yes, verbal agreement that did not include permanent appointment à Go to C14c 3. No 888. Don’t know 999. Refused C14c_contractlength2 What was the length of your contract or verbal agreement? [If length not specified in contract or verbal agreement, put “Not specified” and go to C15a] [If don’t know then put “Don’t Know” and go to C15a] [If refused then put “Refused” and go to C15a] 3 C14d_contractunit2 [Specify unit in which length of contract or agreement is given] 1. days 2. weeks 3. months 4. years 888. Don’t know 999. Refused Go to à C15a C14e_temp2 What was the main reason that you do not have a permanent contract or permanent verbal agreement in this previous MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT ALL THAT APPLY.] 1. This employer did not offer me a job with a permanent contract 2. I did not want to have a job with a permanent contract 3. I was unable to do continuous work 4. I was not able to find any job that offered a permanent contract or verbal agreement 5. Other (specify) 888. Don’t know 999. Refused C15a_sick2 How many days of sick leave did your employer provide PER YEAR in this previous MAIN job? [If “As needed” then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”] C15b_casual2 How many days of PAID earned or casual leave did your employer provide PER YEAR in this previous MAIN job? [If “As needed” then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C15c_holiday2 How many days of holiday did your employer provide EACH WEEK in this previous MAIN job? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16a_maternityactual2 Did you take maternity leave from this previous MAIN job? 1. Yes 2. No à Go to C17 888. Don’t know à Go to C17 999. Refused à Go to C17 4 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16b_maternityactuallength2 How many weeks of maternity leave (paid or unpaid) did you take? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16c_maternityactualpaid2 How many weeks of those weeks were paid? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] à Go to C18 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17a_maternity2 Would your employer have allowed you to take maternity leave from this previous MAIN job if you needed it? 1. Yes 2. No à Go to C18 888. Don’t know à Go to C18 999. Refused [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17b_maternitylength2 How many weeks of maternity leave (paid or unpaid) would your employer have allowed you to take if you needed it? [If as needed then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17c_maternitypaid2 For how many of those weeks would your employer have paid you? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C18_overtime2 Did your employer give you extra pay for overtime work in this previous MAIN job? 1. Yes 2. No 3. Do not work overtime 888. Don’t know 999. Refused C19_bonus2 Did your employer offer any bonus payments in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused 5 C20a_providentfund2 Did your employer offer a Provident Fund in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C20b_gratuity2 Did your employer offer gratuity (lump sum) in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C20c_pension2 Did your employer offer a pension in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C21a_actualtermination2 Were you terminated from this previous MAIN job? 1. Yes 2. No à Go to C21d 888. Don’t know à Go to C21d 999. Refused à Go to C21d C21b_actualtermlength2 How many days of notice were you given? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] C21c_actualtermpay2 How many days of pay did you receive on termination? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] àGo to C23 C21d_termination2 If you had been terminated from this previous MAIN job, would you have been given notice? 1. Yes 2. No à Go to C23 888. Don’t know à Go to C23 999. Refused à Go to C23 C21e_termlength2 How many days of notice would you have been given? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] 6 C21f_termpay2 How many weeks of pay would you have received on termination? [If don’t know then put “Don’t Know” and go to C23] [If refused then put “Refused” and go to C23] C23_reasonleave2 Why did you leave this job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Terminated 2. Firm closed 3. Job completed 4. Found preferred job 5. Decided to start my own business 6. Decided to work for family business 7. Retired 8. Still work at this job but it is no longer main job 9. Other (specify) 888. Don’t know 999. Refused ENUMERATOR: IF C2_emptype2=8, 9, 10 OR 11 (THAT IS, IF THE RESONDENT INDICATED THAT PREVIOUS MAIN JOB #1 IS IN SELF-EMPLOYMENT OR AS A PAID OR UNPAID FAMILY WORKER IN A HOUSEHOLD BUSINESS), FILL OUT C24-C31 FOR PREVIOUS MAIN JOB #1. OTHERWISE GO TO PREVIOUS MAIN JOB #2. C24_selfemptime2 How long did you run or work in this business? 1. Up to 6 months 2. 7-12 months 3. 1-3 years 4. 3-5 years 5. 5-10 years 6. More than 10 years 888. Don’t know 999. Refused C25a_registered2 Was the business registered in any national or local agency? 1. Yes 2. In the process of being registered 3. No à Go to C26 888. Don’t know à Go to C26 999. Refused à Go to C26 7 C25b_registeredorg2 Which organization was the business registered with? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Joint Stock Company 2. Board of Investment (BOI) 3. Bangladesh Small and Cottage Industries Corporation (BSCIC) 4. Bangladesh Export Processing Zones Authority (BEPZA) 5. Department of Inspection for Factories and Establishments (DIFE) 6. Cooperatives 7. City Corporation / Municipalities / Union Parishad (UP) 8. NGO Affairs Bureau 9. Other (Specify) 888. Don’t know 999. Refused C26_whyselfemp2 Why did you start or work in this business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Because I was unable find other work 2. Because I wanted independence/ be my own boss 3. Because I wanted to have flexible working hours 4. Because I wanted higher income 5. Because I wanted to be close to home 6. Because parents, relatives or friends have their own business 7. Because I wanted do grow professionally 8. Because I wanted to have job security 9. Because family members wanted me to work in the business 10. Other (specify) 888. Don’t know 999. Refused C27_difficulty2 What was the main difficulty in running this enterprise/ business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. Acquisition of capital/credit 2. Lack of knowledge 3. Registration 4. Legal requirements 5. Acquisition of new technologies 6. Finding qualified workers 7. Acquisition or rent of a place 8. Difficulty getting paid for good or services provided 9. Lack of electricity 10. Lack of fuel 11. Lack of other infrastructure (roads, buildings) 12. No difficulties 13. Other (specify) 888. Don’t know 8 999. Refused C28_familyemp2 Please tell me the exact number of family members who worked in this business (not including yourself) [Do not leave blank. If 0 then put “0”. If don’t know then ask respondent to estimate. If still don’t know then put “Don’t know”; if refused then put “Refused”. Please check that anyone who reported that Previous Job #1 was self-employed with no employees (C2_emptype2=8) SHOULD NOT report anyone here. Anyone who reported that Previous Job #1 was self- employed and employing paid or unpaid family members (C2_emptype2=9) SHOULD report someone here.] C29_nonfamilyemp2 Please tell me the exact number of nonfamily members who worked in the business [Do not leave blank. If 0 then put “0”. If don’t know then ask respondent to estimate. If still don’t know then put “Don’t know”; if refused then put “Refused”. Please check that anyone who reported that Previous Job #1 was self-employed with no employees or only employing family members (C2_emptype2=8 or 9) SHOULD NOT report anyone here. Anyone who reported that Previous Job #1 was self-employed with some non-family members (C2_emptype2=10) SHOULD report someone here.] C31_reasonleave2 Why did you stop running or working in this business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Firm did not make enough profit 2. Found preferred job 3. Decided to start another business 4. Decided to work for another family business 5. Retired 6. Still work in this business but it is no longer main job 7. Other (specify) 888. Don’t know 999. Refused 9 PREVIOUS MAIN JOB #2 [ENUMERATOR: ASK ABOUT PREVIOUS MAIN JOB #2 IF THE RESPONDENT STARTED PREVIOUS JOB #1 LESS THAN 15 YEARS AGO (IF THE START DATE FOR PREVIOUS MAIN JOB #1, C7a_startyear2, IS AFTER 2000. OTHERWISE GO TO SECTION D.] ENUMERATOR READS: NOW I WOULD LIKE TO ASK YOU ABOUT THE MAIN JOB YOU HELD BEFORE THE JOB WE JUST TALKED ABOUT. C2_emptype3 What describes the employment status of the MAIN job you held just before the previous main job we talked about? 1. Paid employee in government 2. Paid employee in semi-government 3. Paid employee in a private entity 4. Apprentice/Intern/Trainee 5. Seasonal worker 6. Day laborer / casual worker 7. Domestic worker in a private household 8. Self-employed / business owner with no employees 9. Self-employed / business owner employing only paid or unpaid family members 10. Self-employed / business owner employing some non-family members 11. Paid or unpaid family member working in a household business 888. Don’t know 999. Refused C3_location3 Where did you mainly undertake your previous MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. At my home with no special work space 2. At my home with work space inside/ attached to the home 3. Factory, workshop, shop, kiosk, etc 4. Farm or individual agricultural plot 5. Home or workplace of client 6. Construction site 7. Market, bazaar stall, trade fair 8. Street pavement/ highway with fixed post 9. Office building 10. Employer’s home 11. Transport vehicle 12. No fixed location (e.g., mobile; door-to-door; street without fixed post) 13. Others, specify 888. Don’t know 999. Refused C4_industry3 What was the industry of this previous MAIN job? [LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM INDUSTRY LIST. IF NOT CLEAR THEN ASK FOR MORE DETAILS. IF OPTION IS NOT ON THE LIST THEN WRITE DOWN WHAT THE RESPONDENT SAYS.] [If don’t know then put “Don’t Know”] [If refused then put “Refused”] C5_ownership3 What was the type of ownership of the enterprise in this previous MAIN job? 1. Government 2. Semi-government 3. NGO 4. Private enterprise 5. Small business 6. Private household 7. Other (Specify) 888. Don’t know 999. Refused C6_occupation3 What type of activity did you do in this previous MAIN job? [LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM OCCUPATION LIST. IF NOT CLEAR THEN ASK FOR MORE DETAILS. IF OPTION IS NOT ON THE LIST THEN WRITE DOWN WHAT THE RESPONDENT SAYS.] [If don’t know then put “Don’t Know”] [If refused then put “Refused”] C7a_startyear3 What year did you start this previous MAIN job? [YYYY] [If don’t know then put “Don’t Know”] [If refused then put “Refused”] C7b_startmonth3 In which month of the year did you start this previous MAIN job? 1. Jan 2. Feb 3. Mar 4. Apr 5. May 6. Jun 7. Jul 8. Aug 9. Sep 10. Oct 11. Nov 12. Dec 888. Don’t know 999. Refused 1 C8_howfind3 How did you find this previous MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. By starting my own business 2. Contacted employer directly 3. Through a public employment agency 4. Through a private employment agency 5. Through friends or relatives 6. Through school/university employment center 7. By responding to a vacancy 8. Through a union/professional register 9. Through job training programs/courses 10. Other 888. Don’t know 999. Refused C9_numemp3 About how many persons (including yourself) worked in the enterprise or household business where you were employed in this previous MAIN job? 1. 1 2. 2-5 3. 6-9 4. 10-25 5. 26-50 6. 51-100 7. 101-249 8. More than 250 888. Don’t know C10_hours3 How many hours (on average) did you work during a normal WEEK in this previous MAIN job? 1. Fewer than 15 hours per week 2. 15-30 hours per week 3. 30-45 hours per week 4. 45-60 hours per week 5. 60-75 hours per week 6. More than 75 hours per week 888. Don’t know 999. Refused C12a_earnings_cash3 How much did you usually earn PER MONTH in cash from this previous MAIN job (in Taka)? [FOR SELF-EMPLOYED, EARNINGS REFER TO PROFITS, NOT REVENUES. FOR FAMILY MEMBERS WORKING IN A HOUSEHOLD BUSINESS, RESPONDENT SHOULD SAY HOW MUCH HE OR SHE EARNED, NOT HOW MUCH THE BUSINESS EARNED.] [If answered then go to C12c] [If don’t know then put “Don’t Know” and go to C12b] [If refused then put “Refused” and go to C12b] 2 C12b_earnings_cash_range3 In what range did the amount you usually earned PER MONTH from this previous MAIN job fall? 1. Less than 5,000 Taka 2. 5,000-10,000 Taka 3. 10,000-20,000 Taka 4. 20,000-30,000 Taka 5. 30,000 Taka or more 888. Don’t know 999. Refused C12c_earnings_kind3 How much did you usually earn per month in kind (for example, free meals or clothing) from this previous MAIN job (estimated Taka equivalent)? [If no in-kind earnings, then put “0”, do not leave blank] [If don’t know then put “Don’t Know”] [If refused then put “Refused”] ENUMERATOR: IF C2_emptype3=1, 2, 3, 4 5, 6 OR 7 (THAT IS, IF THE RESPONDENT INDICATED THAT PREVIOUS MAIN JOB #2 WAS A PAID JOB OR APPRETICESHIP WORKING FOR SOMEONE ELSE), FILL OUT C14-C23 FOR PREVIOUS MAIN JOB #2. IF C2_emptype3= 8, 9, 10 OR 11 (THAT IS, IF THE RESPONDENT INDICATED THAT PREVIOUS MAIN JOB #2 WAS IN SELF-EMPLOYMENT OR AS A PAID OR UNPAID FAMILY WORKER IN A HOUSEHOLD BUSINESS), GO TO C24 FOR PREVIOUS MAIN JOB #2. C14a_contract3 Did you have a job contract or formal letter of agreement in this previous MAIN job? 1. Yes, permanent contract à Go to C15a 2. Yes, contract that is not permanent à Go to C14c 3. No 888. Don’t know 999. Refused C14b_verbal3 Did you have a verbal agreement about terms and conditions in this previous MAIN job? 1. Yes, verbal agreement that included permanent appointment à Go to C15a 2. Yes, verbal agreement that did not include permanent appointment à Go to C14c 3. No 888. Don’t know 999. Refused C14c_contractlength3 What was the length of your contract or verbal agreement? [If length not specified in contract or verbal agreement, put “Not specified” and go to C15a] [If don’t know then put “Don’t Know” and go to C15a] [If refused then put “Refused” and go to C15a] 3 C14d_contractunit3 [Specify unit in which length of contract or agreement is given] 1. days 2. weeks 3. months 4. years 888. Don’t know 999. Refused Go to à C15a C14e_temp3 What was the main reason that you do not have a permanent contract or permanent verbal agreement in this previous MAIN job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT ALL THAT APPLY.] 1. This employer did not offer me a job with a permanent contract 2. I did not want to have a job with a permanent contract 3. I was unable to do continuous work 4. I was not able to find any job that offered a permanent contract or verbal agreement 5. Other (specify) 888. Don’t know 999. Refused C15a_sick3 How many days of sick leave did your employer provide PER YEAR in this previous MAIN job? [If “As needed” then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C15b_casual3 How many days of PAID earned or casual leave did your employer provide PER YEAR in this previous MAIN job? [If “As needed” then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C15c_holiday3 How many days of holiday did your employer provide EACH WEEK in this previous MAIN job? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16a_maternityactual3 Did you take maternity leave from this previous MAIN job? 1. Yes 2. No à Go to C17 888. Don’t know à Go to C17 999. Refused à Go to C17 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16b_maternityactuallength3 How many weeks of maternity leave (paid or unpaid) did you take? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] 4 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C16c_maternityactualpaid3 How many weeks of those weeks were paid? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] à Go to C18 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17a_maternity3 Would your employer have allowed you to take maternity leave from this previous MAIN job if you needed it? 1. Yes 2. No à Go to C18 888. Don’t know à Go to C18 [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17b_maternitylength3 How many weeks of maternity leave (paid or unpaid) would your employer have allowed you to take if you needed it? [If as needed then put “As needed”] [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] [ONLY ASK IF RESPONDENT IS FEMALE; FOR MALES SKIP TO C18] C17c_maternitypaid3 For how many of those weeks would your employer have paid you? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C18_overtime3 Did your employer give you extra pay for overtime work in this previous MAIN job? 1. Yes 2. No 3. Do not work overtime 888. Don’t know 999. Refused C19_bonus3 Did your employer offer any bonus payments in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C20a_providentfund3 Did your employer offer a Provident Fund in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused 5 C20b_gratuity3 Did your employer offer gratuity (lump sum) in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C20c_pension3 Did your employer offer a pension in this previous MAIN job? 1. Yes 2. No 888. Don’t know 999. Refused C21a_actualtermination3 Were you terminated from this previous MAIN job? 1. Yes 2. No à Go to C21d 888. Don’t know à Go to C21d 999. Refused à Go to C21d C21b_actualtermlength2 How many days of notice were you given? [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] C21c_actualtermpay3 How many days of pay did you receive on termination? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] à Go to C23 C21d_termination3 If you had been terminated from this previous MAIN job, would you have been given notice? 1. Yes 2. No à Go to C23 888. Don’t know à Go to C23 999. Refused à Go to C23 C21e_termlength3 How many days of notice would you have been given? [If don’t know then put “Don’t Know”] [If refused then put “Refused”] C21f_termpay3 How many days, weeks or months of pay would you have received on termination? [If don’t know then put “Don’t Know” and go to C23] [If refused then put “Refused” and go to C23] 6 C23_reasonleave3 Why did you leave this job? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Terminated 2. Firm closed 3. Job completed 4. Found preferred job 5. Decided to start my own business 6. Decided to work for family business 7. Retired 8. Still work at this job but it is no longer main job 9. Other (specify) 888. Don’t know 999. Refused ENUMERATOR: IF C2_emptype3=8, 9, 10 OR 11 (THAT IS, IF THE RESONDENT INDICATED THAT PREVIOUS MAIN JOB #2 IS IN SELF-EMPLOYMENT OR AS A PAID OR UNPAID FAMILY WORKER IN A HOUSEHOLD BUSINESS), FILL OUT C24-C31 FOR PREVIOUS MAIN JOB #2. OTHERWISE GO TO SECTION D. C24_selfemptime3 How long did you run or work in this business? 1. Up to 6 months 2. 7-12 months 3. 1-3 years 4. 3-5 years 5. 5-10 years 6. More than 10 years 888. Don’t know 999. Refused C25a_registered3 Was the business registered in any national or local agency? 1. Yes 2. In the process of being registered 3. No à Go to C26 888. Don’t know à Go to C26 999. Refused à Go to C26 7 C25b_registeredorg3 Which organization was the business registered with? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Joint Stock Company 2. Board of Investment (BOI) 3. Bangladesh Small and Cottage Industries Corporation (BSCIC) 4. Bangladesh Export Processing Zones Authority (BEPZA) 5. Department of Inspection for Factories and Establishments (DIFE) 6. Cooperatives 7. City Corporation / Municipalities / Union Parishad (UP) 8. NGO Affairs Bureau 9. Other (Specify) 888. Don’t know 999. Refused C26_whyselfemp3 Why did you start or work in this business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Because I was unable find other work 2. Because I wanted independence/ be my own boss 3. Because I wanted to have flexible working hours 4. Because I wanted higher income 5. Because I wanted to be close to home 6. Because parents, relatives or friends have their own business 7. Because I wanted do grow professionally 8. Because I wanted to have job security 9. Because family members wanted me to work in the business 10. Other (specify) 888. Don’t know 999. Refused C27_difficulty3 What was the main difficulty in running this enterprise/ business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] 1. Acquisition of capital/credit 2. Lack of knowledge 3. Registration 4. Legal requirements 5. Acquisition of new technologies 6. Finding qualified workers 7. Acquisition or rent of a place 8. Difficulty getting paid for good or services provided 9. Lack of electricity 10. Lack of fuel 11. Lack of other infrastructure (roads, buildings) 12. No difficulties 13. Other (specify) 888. Don’t know 999. Refused 8 C28_familyemp3 Please tell me the exact number of family members who worked in this business (not including yourself) [Do not leave blank. If 0 then put “0”. If don’t know then ask respondent to estimate. If still don’t know then put “Don’t know”; if refused then put “Refused”. Please check that anyone who reported that Previous Job #1 was self-employed with no employees (C2_emptype3=8) SHOULD NOT report anyone here. Anyone who reported that Previous Job #1 was self- employed and employing paid or unpaid family members (C2_emptype3=9) SHOULD report someone here.] C29_nonfamilyemp2 Please tell me the exact number of nonfamily members who worked in the business [Do not leave blank. If 0 then put “0”. If don’t know then ask respondent to estimate. If still don’t know then put “Don’t know”; if refused then put “Refused”. Please check that anyone who reported that Previous Job #1 was self-employed with no employees or only employing family members (C2_emptype3=8 or 9) SHOULD NOT report anyone here. Anyone who reported that Previous Job #1 was self-employed with some non-family members (C2_emptype3=10) SHOULD report someone here.] C31_reasonleave3 Why did you stop running/working in this business? [DO NOT READ CHOICES BELOW. LISTEN TO SPONTANEOUS RESPONSE AND SELECT APPROPRIATE CODE FROM LIST.] [MARK ALL THAT APPLY] 1. Firm did not make enough profit 2. Found preferred job 3. Decided to start another business 4. Decided to work for another family business 5. Retired 6. Still work in this business but it is no longer main job 7. Other (specify) 888. Don’t know 999. Refused 9 D. WORKING CONDITIONS ENUMERATOR READS: Now I would like to ask you some questions about your CURRENT MAIN job, that is, the job where you have spent most of your time over the past 30 days. For each statement, please say how much you agree or disagree with each of the following: Strongly disagree, disagree, agree, or strongly agree. I would like to remind you that we will only use your answers for research purposes, and we will never share your name or contact information with anybody, including your employer, without your permission. 1= 2= 3 = Agree 4= 5= Does not Strongly Disagree Strongly apply/Refused disagree Agree D1_hazard I am exposed to hazards at work (examples: dangerous machines, extreme heat or cold, chemicals) D2_hygiene The hygiene in my workplace is good (examples: clean toilet, good garbage management, available drinking water) D3_violence I have personally been exposed to threats of physical violence, bullying, harassment, unwanted attention or discrimination D4_benefits My employer gives me the benefits that I am supposed to get for my work D5_pay My employer always pays me on time the amount I am supposed to get paid D6_train I have the training opportunities I need to perform my job well D7_progress I have the opportunity to progress / get promoted D8_freedom I prefer jobs where I have independence or freedom to decide how I do my work D9_self I would rather be self- employed than choose any job working for an employer D10_satisfied All things considered, I am fully satisfied with my job E. CHOICE EXPERIMENT ENUMERATOR READS: In this section of the interview we want to try and understand what type of jobs and or employment opportunities you most prefer. I will be doing this by presenting you choice options for two different jobs/employments and then asking you to tell us which one you prefer. Not all jobs/employments come with ideal benefits and hence people choose from the alternative jobs/ employment opportunities they find, whereby each job comes with different levels of certain benefits such as written contract, duration of employment, notice of termination, leave facility, provident fund facility, working hours and of course monthly salary. I will present you with two alternative jobs/employment opportunities that differ in the levels of benefit types that I just mentioned [ENUMERATOR: REPEAT THE BENEFIT TYPES]. Here we only list some main types /characteristics of an job/employment for you to consider and ask you to assume that any other characteristics that are not listed are similar for both alternative jobs that we present. I will present you with 6 choice occasions and each time you will choose your preferred job. Your choice and opinion are very important and can inform government policy makers to help them design appropriate benefit types in line with workers’ preferences. When choosing a job, please compare all benefit types and differences in their levels for the two alternatives and choose carefully. Before showing the actual 6 choice sets, I will now go through with you one example choice and help you to understand better and provide us your reasoned/thoughtful choice in the actual questions [ENUMERATOR: READ OUT EACH JOB ATTRIBUTE AND INDICATE WHAT LEVEL OF ATTRIBUTE EACH JOB HAS] [EXAMPLE CHOICE] JOB A JOB B Written Contract 6 months 1 year Termination Notice 15 days 15 days Working hours 30-45 hours per week 45-60 hours per week Amount of paid leave (not 10 days 10 days including major government holidays / festival leave) Provident Fund No No Monthly salary 20% higher than your current monthly 10% higher than your current monthly income from main economic activity income from main economic activity E0_choicetest If you are given the opportunity to choose from these two different jobs that differ in the levels of some or all benefit types, which job would you choose? 1. Job A 2. Job B ENUMERATOR: PLEASE EXPLAIN CLEARLY AND PROCEED ONCE YOU ARE CONVINCED THAT THE PERSON UNDERSTANDS THAT THEY ARE PRESENTED WITH TWO ALTERNATIVE JOBS THAT DIFFER IN THE LEVEL OF BENEFITS AND IF THEY WERE GIVEN A CHANCE TO CHOOSE FROM THESE, WHICH JOB THEY WILL CHOOSE FOR THEMSELVES. Enumerator reads: Now I will present you six different choice options and each time I will ask you to choose one job from two alternatives which will differ in levels of some or all benefit types. There is no right or wrong answer. In each case we would like to know what (which job) you prefer most [E1-E6: CHOICES FOR 6 MORE SETS OF ALTERNATIVES] FULL SET OF ATTRIBUTES AND LEVELS. SPECIFIC COMBINATIONS CHOSEN USING A FRACTIONAL FACTORIAL DESIGN. EACH RESPONDENT WILL BE ASKED A MAXIMUM OF 6 CHOICES.] Written Contract None 6 months 1 year Long term Termination Notice None 15 days 30 days 60 days Working hours 30-45 hours per week 45-60 hours per week 60-75 hours per week Amount of paid leave None (excluding 5 days government holidays 10 days and festival leave) 15 days Provident Fund Yes No Monthly income Same as now 10% increase over current income 20% increase over current income 30% increase over current income 40% increase over current income 50% increase over current income 1 B. HOUSEHOLD INFORMATION ENUMERATOR READS: FOR THE PURPOSE OF RESEARCH ONLY, I WOULD LIKE TO ASK YOU SOME BASIC INFORMATION ABOUT YOUR HOUSEHOLD. B1_housetype Type of dwelling of the household [Enumerator to record] 1. Building – Flat up to 4 stories 2. Building- Flat multiple storied ( more than 4 storied) 3. Buiding- one storied single house 4. Building- multi-storied single house 5. Tin-shed single unit house 6. Tin-shed Terrace house 7. Katcha one unit house 8. Katcha terrace house 9. Other (specify) B2_houseown Considering the unit where you live, do you: 1. Own it 2. Live in it without paying rent 3. Pay rent to live in it 4. Pay a subsidized amount of rent to live in it 5. Other (specify) 888. Don’t know 999. Refused B3_land Other than your homestead, how much land does your household own [If don’t know then put “Don’t Know”=888] [If refused then put “Refused”=999] B4_mainincome What is the main source of the household’s income? 1. Self-employed / family business (agricultural work) 2. Self-employed / family business (non-agricultural work) 3. Paid employment 4. Pension, dividends, interest, house, rent, property income 5. Other (specify) 888. Don’t know 999. Refused B5_avgincome_range In what range does the household’s monthly average income fall? 1. Less than 5,000 Taka 2. 5,000-10,000 Taka 3. 10,000-20,000 Taka 4. 20,000-30,000 Taka 5. 30,000 -40000 Taka 6. 40000-50000 Taka 7. 50000-60000 Taka 8. 60000- 70000 Taka 9. 70000- 80000 Taka 10. Above 80000 Taka 888. Don’t know 999. Refused F. FOLLOW-UP QUESTIONS ENUMERATOR READS: This section includes a few questions to help us understand how to improve this survey. I will first read a few statements about the choices you just made. For each statement, please say how much you agree or disagree with each of the following: Strongly disagree, disagree, agree, or strongly agree. 1= 2= 3= 4= 5= Strongly Disagree Agree Strongly Refused/Don’t disagree Agree know F1_clear The choices I just had to make were clear F2_relevant The attributes and their levels were relevant to my decisions F3 When you chose a job, which are the first three attributes/benefit types that F3_attribute1 influenced the choice you considered? F3_attribute2 [please rank from the first to third] F3_attribute3 1