Understanding Female Labor Force Participation in Urban Bangladesh What Works for Working Women? Understanding Female Labor Force Participation in Urban Bangladesh APHICHOKE KOTIKULA RUTH HILL WAMEQ AZFAR RAZA 1 What Works for Working Women? © 2019 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org World Bank Office Dhaka Plot – E-32, Agargaon Sher-e-Bangla Nagar Dhaka-1207, Bangladesh Tel: 880 2 5566 7777 Fax: 880 2 5566 7778 www.worldbank.org.bd This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because the World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution Please cite the work as follows: Aphichoke Kotikula, Ruth Hill and Wameq Azfar Raza. 2019. What works for working women? Understanding female labor force participation in urban Bangladesh. Washington, DC: World Bank. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Cover photo credit K M Asad Design & Print Progressive Printers Pvt. Ltd. 2 Understanding Female Labor Force Participation in Urban Bangladesh TABLE OF CONTENTS Executive Summary........................................................................................................................................................ 9 1. Introduction............................................................................................................................................................... 11 2. Literature Review...................................................................................................................................................... 13 3. Data: Sampling and Summary of Statistics.......................................................................................................... 17 3.1 Sampling Methodology..........................................................................................................................................................................................17 3.2 Demographic Characteristics of Sample Households.........................................................................................................................20 3.3 Comparing DIGNITY to Other Data.................................................................................................................................................................22 3.4 Slums vs. Nonslums...................................................................................................................................................................................................22 4. Women’s Work: Drivers and Constraints.............................................................................................................. 25 4.1 Time-Use Data in the DIGNITY Survey...........................................................................................................................................................25 4.2 Work and the Nature of Work.............................................................................................................................................................................28 4.3 Skills.....................................................................................................................................................................................................................................30 4.4 Childcare..........................................................................................................................................................................................................................31 4.5 Transportation..............................................................................................................................................................................................................32 4.6 Perceptions, Gender Norms, and Attitudes................................................................................................................................................33 5. Exploring Constraints to Work............................................................................................................................... 37 6. Conclusion.................................................................................................................................................................. 47 Annex...............................................................................................................................................................................................................................................49 References....................................................................................................................................................................................................................................51 LIST OF FIGURES Figure 1 U-Shape Curves of Female and Male Employment Rates...........................................................................................................11 Figure 2 Locations of PSUs.................................................................................................................................................................................................19 Figure 3 Duration of Time in Current Community..............................................................................................................................................21 Figure 4 Time-Use among Women, In and Out of the Labor Force.........................................................................................................27 Figure 5 Time-Use among Men, In and Out of the Labor Force.................................................................................................................28 Figure 6 Cognitive and NonCognitive Skills among Women, In and Not In the Labor Force..................................................30 Figure 7 Cognitive and Noncognitive Skills among Men In and Not In the Labor Force...........................................................31 3 What Works for Working Women? LIST OF TABLES Table 1 Sample Allocation.................................................................................................................................................................................................18 Table 2 Demographic Characteristics of the Sample........................................................................................................................................20 Table 3 Migration History....................................................................................................................................................................................................21 Table 4 Comparison of DIGNITY Sample of National Data.............................................................................................................................22 Table 5 Comparison of Slum and Nonslum Areas...............................................................................................................................................24 Table 6 Typical Day of Men and Women in the DIGNITY Sample.............................................................................................................26 Table 7 Time Utilization by Status of Labor Market Participation...............................................................................................................27 Table 8 Who Works: Demographics and Educational Level of Individuals in the Labor Force...............................................28 Table 9 Nature of Work.........................................................................................................................................................................................................29 Table 10 Primary Reasons for Not Participating in the Labor Force.........................................................................................................29 Table 11 Need for Childcare..............................................................................................................................................................................................31 Table 12 Unmet Demand for Childcare.....................................................................................................................................................................32 Table 13 Employment and Transportation..............................................................................................................................................................33 Table 14 Frequency of Travel Outside the Home................................................................................................................................................34 Table 15 Perceptions, Norms and Attitudes about Safety and Freedom of Movement.............................................................35 Table 16 Correlates of Labor Force Participation..................................................................................................................................................40 Table 17 Correlates of Wage Earning versus Self-Employment..................................................................................................................41 Table 18 Correlates of Hours Spent at Work (log-linear outcome)...........................................................................................................42 Table 19 Correlation between Cognitive and Noncognitive Skills and Female Economic Activity.....................................43 Table 20 Correlation between Cognitive and Noncognitive Skills and Type of Employment................................................44 Table 21 Correlation between Cognitive and Noncognitive Skills on the Hours Spent at Work...........................................45 4 Understanding Female Labor Force Participation in Urban Bangladesh ACKNOWLEDGEMENTS The “What works for working women? Understanding female labor force participation in urban Bangladesh” Report is a joint production of the World Bank’s Gender Group and the Poverty Global Practice. It was prepared by a team led by Aphichoke Kotikula, Ruth Hill and Wameq Azfar Raza. The report is also a background paper for the Bangladesh Poverty Assessment 2019. The team thanks the leadership and support from management including Caren Grown (Senior Director, Gender Group), Stefano Mocci (Manager, Gender Group), Benu Bidani (Practice Manager, Poverty) and David Seth Warren (Practice Manager, Social Development). The peer reviewers were Ana Maria Munoz, Sonya Sultan, and Syud Amer Ahmed. Administrative support was provided by Nelly Obias, Shegufta Shahriar and Shirley Regina Bekoe. The team thanks Faizuddin Ahmed for his help on the sampling and Kathleen Beegle, Maria Eugenia Genoni, Benedicte de la Briere, Monica Yanez Pagans, Anna O’Donnell, Sabah Moyeen, Bakhtiar Sohag for their guidance and support throughout the process of this study. The team would also like to thank Data Analysis and Technical Assistance (DATA) for their input the questionnaire design and the DIGNITY survey field work. The report’s design was done by Enamul Shahriar and printing by Shudev Kumar Kar from Progressive Printers. This work has also been funded in part by the Umbrella Facility for Gender Equality (UFGE), a World Bank Group multi-donor trust fund expanding evidence, knowledge and data needed to identify and address key gaps between men and women to deliver better development solutions that boost prosperity and increase opportunity for all. The UFGE has received generous contributions from Australia, Canada, Denmark, Finland, the Bill & Melinda Gates Foundation, Germany, Iceland, Netherlands, Norway, Spain, Sweden, Switzerland, United Kingdom, and the United States. 5 What Works for Working Women? 6 Understanding Female Labor Force Participation in Urban Bangladesh ACRONYMS AND ABBREVIATIONS BDT Bangladesh Taka CESD Center for Epidemiologic Studies Depression Scale  Dhaka CC Dhaka City Corporation DIGNITY Dhaka low Income area GeNder, Inclusion, and povertY EA Enumeration Area FLFP Female Labor Force Participation GBV Gender-Based Violence HIES Household Income and Expenditures Survey IFPRI International Food Policy Research Institute LFP Labor Force Participation LFS Labor Force Survey PPS Probability Proportional to Size PSU Primary Sampling Unit PVC Polyvinyl Chloride RMG Ready-Made Garment SMA Statistical Metropolitan Area TVET Technical and Vocational Education and Training 7 What Works for Working Women? 8 Understanding Female Labor Force Participation in Urban Bangladesh EXECUTIVE SUMMARY Labor income growth has been the fundamental driver of poverty reduction in Bangladesh in recent years (Joliffe et al 2013, Hill and Genoni 2018). During the last five years urban poverty reduction has stagnated; this has also been a period in which urban female labor force participation (FLFP) rates have also fallen. Although the decline in FLFP may be driven in part by changes in the demand for female labor in the ready-made garment (RMG) sector in recent years, it is also possible that it is being driven by supply-side factors—for example, social norms around women’s work and travel outside the home, and other factors that affect the cognitive and noncognitive skills of women as they contemplate entering the workforce. As Bangladesh continues to urbanize, understanding the factors that constrain FLFP in urban areas is increasingly important in understanding how to ensure urban income growth and poverty reduction. This paper explores the factors that constrain women in slums and low-income neighborhoods in Dhaka from engaging in the labor market and supplying their labor to wage earning or self-employment. It uses unique individual-level data on labor market participation, time-use, norms, and skills, both cognitive and noncognitive. The data reconfirms well-known patterns associated with FLFP: that is, it is higher in low-income neighborhoods and among women with little education, and younger unmarried women. The paper also highlights the correlation between soft skills and type of work. The paper also quantifies the important correlation between the need for childcare, as well as safety in public spaces and in the workplace. Women report the lack of good, affordable childcare to be a constraint, and are often looking after children even while engaging in other activities (something that is impossible to do when working away from home). They are 15 percent less likely to work if they have children less than five years old. The survey revealed that public spaces in Dhaka are often male-dominated spaces, and that this is correlated with women’s decisions about employment. Women who do not feel that the environment outside of their home is safe are 10 percent less likely to participate in the labor market. Men spend four times more time outside of their communities than women do. A quarter of women only leave their neighborhood once a month, and one in ten never leave their community at all. This means that there are many women in Dhaka who live life as if they were in a remote village despite living in one of the biggest cities in the world. Finally, the evidence discussed in this report is consistent with the notion that for many women, working is not deemed desirable, and is something that is not associated with feeling happy about life. Making public spaces more female-friendly; travel safer for women; and female work more socially acceptable may be important in order to increase the ability of women to grow their families’ incomes happily, and free of fear. 9 What Works for Working Women? 10 Understanding Female Labor Force Participation in Urban Bangladesh 1 INTRODUCTION Bangladesh has made great strides in female labor force participation (FLFP): it now outperforms neighbors such as India and Pakistan. But the national rate of 36 percent is still considered too low. This rate is still 46 percentage points behind the rate for male counterparts in Bangladesh (82 percent) (BBS 2018). Moreover, the FLFP rate is low compared to countries such as Cambodia (81 percent), Indonesia (51 percent), and Vietnam (73 percent) (World Bank WDI 2019). In fact, the trend for the women’s employment rate in Bangladesh is unusually low when comparing it with other countries with a similar income level (Figure 1). The men’s employment rate, however, is on par with other countries (World Bank 2018). © K M Asad Figure 1 U-Shape Curves of Female and Male Employment Rates Source: South Asia Economic Focus (Spring 2018). Recently, FLFP rates in Bangladesh have fallen as a result of lower rates in urban settings, and an urbanizing population. In 2017 FLFP was 31 percent in urban areas compared to 39 percent in rural areas (BBS 2018). However, the FLFP in urban areas has been declining in recent years, from 34.5 percent in 2010, to 33 percent in 2013, and most recently to 31 percent in 2016 and 2017 (BBS 2018). This is a troubling trend, as Bangladesh is urbanizing. And while 37 percent of the country’s population currently lives in urban areas, the urban population is projected to reach 112 million, or 58 percent of the total population, by 2050 (UNDESA 2018). In effect, this suggests that almost all future population growth in Bangladesh will be in urban areas. 11 What Works for Working Women? Lower rates of urban FLFP and gender differences in returns to labor are constraining urban income growth and poverty reduction. Although on average urban areas are less poor than rural areas, the urban poverty rates are still quite high, and there are some urban neighborhoods where the level of poverty is very high. The number of slums has increased substantially, from 2,991 in 1997 to 13,938 in 2014, and the population in slums has increased at an annual rate of 12 percent—from 0.7 million in 1997 to 2.2 million in 2014—nearly double the population growth in urban areas as a whole (BBS 2014). The main driver of poverty reduction in Bangladesh has been income from labor; as such there is a need to better understand the constraints to FLFP and income growth in urban areas, particularly in low-income areas. The recent literature on urban FLFP, which will be reviewed in Section 2 of this paper, suggests that FLFP in urban areas cannot be fully explained by the same determinants that have been important in the past, such as level of educational attainment and marital status. There is a need to better understand FLFP in urban areas, and to do so using data that better captures some of the other factors that may be constraining women’s participation in the workforce. Female employment in urban areas is central to the narrative of Bangladesh’s economic growth. Large-scale job creation in the manufacturing sector, mostly in urban areas, has been contributing to employment growth (2.4 percent annually from 2003 to 2016), with wage employment growing by 5.7 percent per year. This expansion is also a factor in the growth in female employment, bringing millions of women into the labor force (Farole and Cho 2017). The 2017 Bangladesh Jobs Diagnostics also suggests that declining urban FLFP could be caused by a lower demand for female labor associated with the rapid employment slowdown in the RMG and textile sectors—the largest source of urban female employment—since 2010 (Farole and Cho 2017). However, lower FLFP rates could also be driven by changes in supply-side factors such as changes in skills, demographics, or social norms around female work. This study attempts to fill in the data and knowledge gaps on women’s economic empowerment in urban areas. While an array of national-level datasets has collected a wide spectrum of information, they rarely comprise all of the information needed to study the drivers of FLFP. Quantitative data on urban labor market dynamics in Bangladesh, especially for women, is still quite limited. For instance, the time spent by women on household work and caregiving is an important deterrent to economic activity. This information, however, is collected only through a survey that is separate from the Household Income and Expenditure Survey (HIES) and the Labor Force Survey (LFS), which collect detailed employment-related information. Additional pertinent information that relates to labor force participation, such as information about skills, urban migration, employment patterns and history, perception of social capital, and safety, transportation, and service utilization are collected through a variety of instruments. The data gap is being filled by the primary data collection of a specialized survey called Dhaka low Income area GeNder, Inclusion, and povertY (or DIGNITY) survey; it is representative of poor urban areas and is specifically designed to address these limitations. DIGNITY collected information from 1,300 urban households living in poor areas of Dhaka in 2018 on a range of issues that affect FLFP as identified through the literature. These range from household composition and demographic characteristics to socioeconomic characteristics such as detailed employment history and income (including locational data and travel details); and from technical and educational attributes to issues of time use, migration history, and attitudes and perceptions. The existing literature on known constraints to FLFP in Bangladesh will be presented in Section 2, with subsections related to various subtopics. Section 3 discusses the main data used for this paper, the DIGNITY survey, and describes its sampling methodology and key summary statistics. Section 4 presents a descriptive analysis of FLFP, along with discussions of key constraints and enabling factors for women’s employment in urban areas. Finally, Section 5 presents our results, drawn from a multivariate analysis of women’s work in Dhaka. 12 Understanding Female Labor Force Participation in Urban Bangladesh 2 LITERATURE REVIEW © Sagarika Bose The goal of this paper is to present the factors that contribute to FLFP in low-income areas of urban centers in Bangladesh. Qualitative studies suggest that working women in urban areas tend to exhibit greater measures of empowerment, such as managing household expenditures, having access to savings, and freedom of movement (Kabeer 2017). As extreme poverty is an important factor in explaining some of the rise in female employment in Bangladesh, one may expect a higher level of FLFP in low-income areas than the rest of Dhaka city (Bridges, Lawson and Begum 2011).The literature on female employment in low-income areas of Dhaka tends to rely on qualitative studies. Salway, Jesmin, and Rahman (2005) compared married women in Dhaka who were engaged in wage employment with those who had never worked. Drawing on fieldwork in four Dhaka slums, Banks (2013) revealed a complex balance involved in meeting economic and social priorities. Her work suggests that female employment is seen by men as a short-term coping strategy to deal with the high cost of living in urban areas, while women, particularly women working in industries like the garment sector, recognize that their work can contribute to the household income. Still, women need to balance work and domestic responsibilities. Reviewing the existing literature about constraints to—and enabling factors for—women’s work in Bangladesh can help guide analysis. The new World Bank publication Voices to Choices: Bangladesh’s Journey in Women’s Economic Empowerment analyzes multiple data sources 13 What Works for Working Women? to understand recent trends and patterns in women’s employment. This book identifies the relationships between FLFP and key drivers such as level of educational attainment, marriage, presence of children, and use of time. Marriage has been a known correlate of FLFP in Bangladesh (Joliffe et al 2013, Mahmud and Bidisha 2018). A recent analysis of the Labor Force Survey (LFS) data in Voices to Choices reveals that marriage penalizes only urban women’s participation in the labor markets, whereas it is associated with higher labor force participation (LFP) probabilities for urban men, and for both women and men in rural areas. In 2016, the probability of LFP was 6.4 percentage points lower for urban married women than for urban unmarried women, whereas marriage was associated with a 14-percentage-point premium for urban men. Unlike rural women, who often work in the agricultural sector, urban women’s work may more often require leaving the home, which is more likely to result in familial disapproval. Arranging childcare may also be easier in rural settings due to a network of family members who can help care for the children. Voices to Choices also show that married women have increasingly been entering the labor markets. While FLFP of married women has shifted upward across all age groups, the highest increase has been among married women in their 30s and 40s. “The LFP rate of married women in the cohort born between 1968 and 1977 (who were thus 26–35 years old in 2003, and 39–48 years old in 2016) increased from 26 percent in 2003 to 42 percent in 2016” (Solotaroff et al 2019, p. 32). This shift may suggest an improvement in attitudes of husbands and mothers-in-law toward women’s LFP as well as a reduction in childcare duties due to fertility decline. Moreover, having older family members living in the household appears to be positively correlated with FLFP. The presence of children in the household is significantly related to the LFP probabilities of both men and women, though the relationship varies by the child’s age. A cross-country study suggests that having a child can reduce LFP by about two years during a woman’s reproductive life (Bloom and Finely 2009). Women with children ages 0–5 in the household are significantly less likely to (re)join the labor force than women without young children. (For men the relationship is not significant). However, both women and men are more likely to be in labor markets if there are children ages 6–10 in the household (compared to women and men without children this age). The negative association between the likelihood that women will be working when children ages 0–5 are present is almost three times greater for urban than for rural women. This finding may suggest that access to childcare services is a crucial factor in women’s LFP (Solotaroff et al 2019). The relationship between educational level and LFP in Bangladesh appears to follow a U-shaped curve pattern, according to the analysis of the 2016 LFS in Voices to Choices. LFP is high among those without education and those with only a primary education; and it falls with an increasing educational level, with a trough observed at the level of secondary school certificate. On the other hand, the LFP of women with higher education degrees are highest compared to the other groups. Rahman and Islam (2013) also finds that while the relationship between casual and daily employment and education is negative and significant, the opposite is true for salaried, self and family employment. Time-use data analysis shows that traditional gender roles still impose a disproportionate burden of care work on Bangladeshi women. According to the analysis of the Bangladesh Time Use Survey data in Voices to Choices, urban women spent about six hours per day on household domestic tasks and unpaid care work, while urban men spent only one hour per day on these activities. Even among working women in urban areas, whose number of hours of work per day is only slightly less than men’s (7 vs. 8.6 hours), they still spend about 3 hours a day on domestic tasks and unpaid care work. This data shows that care work may present an even greater obstacle for urban women than for rural women, and that access to childcare services may be most urgently needed in order to bring more urban women into the workforce. 14 Understanding Female Labor Force Participation in Urban Bangladesh Occupational sex-segregation is another constraint on women’s access to jobs. In labor markets across the world, a widespread phenomenon is an imbalance in the types of employment in which men and women engage, and consequently the occurrence of male-dominated and female-dominated sectors or jobs. The concept of “women’s work” in Bangladesh is often dictated by gender norms that affect how people form their perceptions about what is appropriate for men and for women. Voices to Choices suggests that traditionally low-paid workers such as housecleaners and maids are predominantly female. However, the 2016 LFS data shows an encouraging trend that many manufacturing and higher-skill service occupations showed a high share of female workers1. The qualitative data from low-income areas in urban centers provides additional nuance. Women are not only employed in garment factories, but also in PVC tape, small manufacturing, and dyeing factories, and bakeries. Another positive trend is the dwindling of occupational segregation by gender, and the expansion of “mixed gender” industries2 in the country. As women began to enter male-dominated occupations, the number of mixed- gender industries expanded substantially—from 16 percent in 2003 to 75 percent in 2016. The percentage of women working in female-dominated industries has also fallen, from 15 percent in 2003 to only 4 percent in 2016. The expanded presence of mixed-gender employment could be a sign that Bangladesh’s labor market is now malleable enough to accept women into a wide variety of occupations, and that there is an increased social acceptability of women working with men. In this regard, a study of female engineers in Bangladesh suggests that organizations can create their own organizational culture and practices that are welcoming for female staff, despite the prevailing social norms (Hossain and Kusakabe 2005). Other constraints to FLFP identified through the literature and qualitative research include the risks of gender- based violence (GBV), restricted mobility, gender norms, and socioemotional skills gaps. However, quantitative data are rarely available. Restricted mobility and the practice of purdah3 is conjectured to prevent women from seeking lucrative jobs outside the home (Kabeer 2013, Ahmed and Sen 2018, Asadullah and Wahhaj 2019). While household surveys such as the LFS and the HIES do not contain such information, the location of work data is telling. In 2016, 62 percent of working women worked from home, whereas almost all men were working outside their homes. Although home-based work may seem like a solution to narrowing gender gaps in LFP, it hurts women financially, as it is associated with low-paid (or often unpaid) and low-skill work. However, the LFS does not have data about mobility restriction. GBV also continues to inhibit women’s employment outside the home, as women are likely to face a double risk of harassment—that is, harassment while traveling to and from work as well as in the workplace (Solotaroff et al 2019). The DIGNITY survey was designed specifically to collect data about constraints to women’s employment; and this paper will attempt to fill in the knowledge gaps by analyzing the data DIGNITY gathered. 1 For example, women account for 39 percent of health professionals and 45 percent of manufacturing workers in food processing, wood working, and garment. 2 In this context, mixed industries are defined as those whose share of female workers are between 40 to 60 percent. 3 The concept of purdah, or female seclusion practice, will be explained in detail in section 4.6. 15 What Works for Working Women? 16 Understanding Female Labor Force Participation in Urban Bangladesh 3 DATA: SAMPLING AND SUMMARY OF STATISTICS © Lucy Milmo 3.1 Sampling Methodology The DIGNITY (Dhaka low Income area GeNder, Inclusion and poverTY) survey was designed to shed light on poverty, economic empowerment, and livelihood in urban areas of Bangladesh. A broad array of information was collected on issues related to women’s economic empowerment, ranging from demographic and socioeconomic characteristics to detailed work history, time use, and attitudes about, and perceptions of, work. The key feature of the DIGNITY survey is that it collected economic data directly from the main household members, generally the main couples, unlike traditional surveys which only interviewed the heads of households (who tend to be men in most cases); thus, failed to gather valuable information from the female population. The survey instrument has two main modules: the traditional household module (in which the head of household is interviewed on basic information about the household); and the individual module, in which two respondents from each household are interviewed individually. In the second module, two persons—one male and one female—from each household, usually the main couple, are selected for the interview. The survey team deployed one male and one female interviewer for each household, so that the gender of the interviewers matched that of the respondents. The DIGNITY survey is representative of low-income areas and slums of the Dhaka City Corporations (North and South, from here on referred to as Dhaka CCs), and an additional low-income site from the Greater Dhaka Statistical 17 What Works for Working Women? Metropolitan Area (SMA), following a two-stage stratification design. In the first stage primary sampling units (PSUs) were selected using probability proportional to size (PPS). In the second stage, all of the households in the PSU were listed, and 20 households were selected for interviewing. Strata were used in both stages. First Stage: Selection of the PSUs Low-income PSUs were defined as nonslum census enumeration areas (EAs), in which the small-sample area estimate of the poverty rate is higher than 8 percent (using the 2011 Bangladesh Poverty Map). The sampling frame for these low-income areas in the Dhaka CCs and Greater Dhaka is based on the population census of 2011. For the Dhaka CCs, all low-income census EAs formed the sampling frame. In the Greater Dhaka area, the frame was formed by all low-income census EAs in specific thanas4 where World Bank project locations were located. Three strata were used for sampling the low-income EAs. These strata were defined based on the poverty head- count ratios. The first stratum encompasses EAs with a poverty headcount ratio between 8 and 10 percent; the second stratum between 11 and 14 percent; and the third stratum, those exceeding 15 percent. The number of EAs selected from each stratum in the Dhaka CCs and the greater Dhaka areas is shown in Table 1. Slums were defined as informal settlements that were listed in the Bangladesh Bureau of Statistics’ slum census from 2013/14. This census was used as sampling frame of the slum areas. Only slums in the Dhaka City Corporations are included. Again, three strata were used to sample the slums. This time the strata were based on the size of the slums. The first stratum comprises slums of 50 to 75 households; the second 76 to 99 households; and the third, more than 100 households. Small slums with fewer than 50 households were not included in the sampling frame. Very small slums were included in the low-income neighborhood selection if they are in a low-income area. The number of slums selected from each stratum is shown in the bottom panel of Table 1. Table 1 Sample Allocation Selection criteria Criteria 1 Criteria 2 Criteria 3 (poverty (male and Number of (only female (only male  Strata headcount female EAs working- age working- age ratio or working-age member) member) number of member) households) Low-income areas in the Dhaka CCs 1 20 8%-10% 328 57 15 2 18 11%-14% 181 33 7 3 5 15% + 82 15 4 Low-income area in greater Dhaka 3 5 15% + 180 31 9 Slums in the Dhaka CCs 1 3 50 to 75 52 6 2 2 8 76 to 99 137 22 1 3 7 100+ 116 21 3 Altogether, the DIGNITY survey collected data from 67 PSUs. Locations of the PSUs are shown in Figure 2. 4 Thana, or Upazila, is an administrative unit in Bangladesh, functioning as a subunit of a district. 18 Understanding Female Labor Force Participation in Urban Bangladesh Second Stage: Selection of the Households In each sampled PSU a complete listing of households was done to form the frame for the second stage of sampling: the selection of households. When the number of households in a PSU was very large, smaller sections of the neighborhood were identified, and one section was randomly selected to be listed. The listing data collected information on the demographics of the household to determine whether a household fell into one of the three categories that were used to stratify the household sample: i) households with both working-age male and female members; ii) households with only a working-age female; iii) households with only a working-age male. Households were selected from each stratum with the predetermined ratio of 16:3:1. In some cases there were not enough households in categories (ii) and (iii) to stick to this ratio; in this case all of the households in the category were sampled, and additional households were selected from the first category to bring the total number of households sampled in the PSU to 20. The total sample consists of 1,300 households (2,378 individuals). Table 1 provides the breakdown of households into each stratum. Figure 2 Locations of PSUs Source: Authors’ rendition, based on the DIGNITY data. 19 What Works for Working Women? 3.2 Demographic Characteristics of Sample Households Summary statistics of the demographic characteristics of the sample are presented in Table 2. The respondents are equally distributed across genders as a result of the fact that most households have both men and women of working age, and in these cases one man and one woman were interviewed. The average age of women in our sample is 32.8 years old, while the men on average are 37.1 years. The majority of respondents (91 percent) report being married, although the proportion is marginally higher among women (92 percent) than among men (90 percent). The average age at marriage is 18.9 years, but this masks a wide variation across genders: the average age of marriage for men is 22.3 years, while for women it is only 15.8 years. The average household size is four people. In terms of education, men and women have both completed an average of 3.5 years of schooling. Ten percent of men had attended technical and vocational education and training (TVET) compared to 7 percent of women. Table 2 Demographic Characteristics of the Sample Full Sample Male Female Test (Male –    Mean  SD  Mean  SD  Mean  SD  Female) Gender (1=male; 0=female)  0.49 0.50 Age of respondent (completed 34.95 11.63 37.15 11.83 32.84 11.04 *** years)  Respondent is head of household 0.50 0.50 0.90 0.30 0.12 0.33 *** (1/0) Marital status Never married (1/0) 0.04 0.20 0.08 0.27 0.01 0.09 *** Married (1/0) 0.91 0.29 0.92 0.28 0.90 0.31 *** Widow/widower (1/0) 0.03 0.18 0.00 0.05 0.06 0.24 * Divorced (1/0) 0.00 0.06 0.00 0.03 0.01 0.08 *** Separated/deserted (1/0) 0.01 0.12 0.00 0.02 0.03 0.17 ** Age at marriage (continuous 18.85 4.68 22.29 4.03 15.78 2.65 *** years)  Household size 3.94 1.55 3.97 1.54 3.91 1.55 Years of completed education  3.55 3.70 3.58 3.84 3.52 3.55 Respondent received TVET (1/0)  0.08 0.28 0.10 0.30 0.07 0.26 ** N  2378     1117    1259       Notes: Table shows individual level weighted means of the sample in the study. * p<0.1; ** p<0.05; *** p<0.01. TVET refers to technical and vocational education and training. Because migration–particularly rural to urban migration—is synonymous with urbanization in Bangladesh, understanding the migration profile of people in low-income communities is crucial. The DIGNITY data confirm the hypothesis that the majority of the people in the survey are migrants. About 74 percent are from other places, while only 26 percent are from Dhaka zilla (or district, including those born in the community) (see Table 3). Migration to urban areas is mostly economically motivated; about 7 in 10 migrants moved to the city to find work. The migrant ratio among women is slightly higher than among men, reflecting the general trend that women tend to move due to marriage. About 24 percent of female respondents moved to their current community because of marriage, compared to about 1 percent of men. On the other hand, about 75 percent of men reported moving to the community to find work, compared to about 50 percent of the women. In any case, the ratio of adult women who have moved because of work is much higher than the urban average of only 14 percent according to the 2016 LFS. 20 Understanding Female Labor Force Participation in Urban Bangladesh Table 3 Migration History Male Female Test (Male -    Standard Standard Mean  Mean  Female) Deviation  Deviation  Origin  From Dhaka Zilla (1/0) 0.28 0.45 0.25 0.43 From elsewhere (1/0) 0.72 0.45 0.75 0.43 Primary reason for moving  Work (1/0) 0.73 0.44 0.47 0.50 *** Marriage (1/0) 0.01 0.09 0.24 0.43 *** With or for family (1/0) 0.09 0.29 0.16 0.37 *** N  1117   1261      Note: Table shows weighted means at the household level. * p<0.1; ** p<0.05; *** p<0.01. Figure 3 Duration of Time in Current Community The respondents’ duration of stay in the community is distributed across the time spectrum (Figure 3). While about 30 percent have lived the community for five years or less, an equal amount have lived in the community for more than 20 years. Moreover, about 16 percent of the sample were born in the community. The data suggests that these communities have been established with dynamic population movement. 21 What Works for Working Women? 3.3 Comparing DIGNITY to Other Data An initial comparison suggests that the population interviewed for the DIGNITY survey is largely comparable to those found in the HIES for household composition and other demographic characteristics (household size, gender ratio, and the adult-to-child ratio) (see Table 4). Understandably we begin to note divergent trends between the two samples when it comes to housing and socioeconomic characteristics. This is likely driven by the fact that the DIGNITY sample is exclusively focused on poorer areas of Dhaka, and therefore is able to capture a focused range of representative households. For instance, the households from the DIGNITY survey are markedly worse off when it comes to the earner-to-nonearner ratio (the dependency ratio) and completed years of education. Table 4 Comparison of DIGNITY Sample of National Data     HIES 2016   DIGNITY Survey Dhaka Dhaka Dhaka Dhaka       CC SMA CC SMA Demographic characteristics             Household size   3.7 3.7   3.9 4.1 Gender Ratio (Male: Female)   49% 49%   49% 49% Proportion of adults in the household (>15 years)   72% 72%   74% 72% Proportion of married adults   93% 92%   91% 90% Housing characteristics             Roof material (brick/cement)   47% 21%   24% 38% Electricity connection   100% 96%   100% 100% Use of improved source of water   100% 97%   100% 68% Socioeconomic status and job           Dependency ratio   39% 36%   53% 47% Years of education completed   9.3 8.3   3.5 4.2 Note: Table shows weighted means. 3.4 Slums vs. Nonslums Slum spaces are typically characterized by dense and often poorly-constructed living spaces; poor access to basic facilities; and are predominantly occupied by the poor. Approximately 93 percent of households in this survey are located within Dhaka City Corporation (Dhaka CC). Overall, households in slum communities make up about 16 percent of the city’s population. Comparison between the slum and nonslum areas reveal few differences in terms of demographic characteristics. Women represent half of the population in both areas, while their average age is approximately 35 years (see Table 5). The average age when individuals marry is fairly young – approximately 19 years. The number of household members is marginally higher in slum areas (4.4) compared to nonslum areas (3.85). Approximately 89 percent of respondents live in male-headed households in both locations. Individuals living in nonslum areas are generally more well educated; on average they have 0.86 more years of education. Similarly, a greater proportion of household heads are able to read and write (52 percent vs. 43 percent for slum and nonslum areas, respectively). Housing in low-income areas of Dhaka is characterized by limited space, but with decent access to services. On average, the houses in slum areas are smaller than those in nonslum areas (121.4 versus 127.5 square feet), and with fewer rooms (1.2 versus 1.3 rooms). Though the number of rooms per person is the same in both locations (0.3 rooms), the space per capita is smaller in slums (32.1 versus 35.1 square feet). Approximately 81 percent of households in 22 Understanding Female Labor Force Participation in Urban Bangladesh nonslum areas have separate kitchens, compared to 59 percent in slum areas. Households in nonslum areas are generally built with better building materials. Roofs made of bricks and cement are 7 percentage points higher among nonslum households, while the opposite is true for those made with more temporary materials such as tin. Similarly, the proportion of respondents who can access improved toilets in slums (79 percent) is 10 percentage points below that in nonslum areas (89 percent). While © Aphichoke Kotikula access to water from an improved source is universal for slum households, only 94 percent of nonslum households have this: this is likely a positive externality of the comparatively higher levels of population congestion in slums, where a single improved source of water is accessed by everyone in the community. Access to electricity is universal in the sample. As with housing conditions, access to services varies by household location. Schools for instance, both public and private, are accessed 4 percentage points more by slum households. The proportion of slum households with access to public healthcare facilities is 4 percentage points higher than the 24 percent in nonslum areas. The opposite is true for private healthcare facilities. Perhaps this is driven by the fact that, given the concentration of population in slum areas, many service providers, such as education and healthcare providers, congregate around these areas. Lastly, in terms of transportation, for both short and long-distance modes of travel, nonslum households have more frequent access than slum households (by 10 and 19 percentage points respectively). This could reflect the fact that slums typically form near either potential or existing job sites, thereby precluding the need for vehicular transport. Modes of work-related transport are further discussed in Section 4.5. 23 What Works for Working Women? Table 5 Comparison of Slum and Nonslum Areas Slum  Nonslum  Test (Slum - Mean  SD  Mean  SD  Nonslum) Demographics Gender (1=male; 0=female) 0.49 0.50 0.49 0.50 Age of respondent (completed years)  35.56 12.20 34.83 11.52 Age of marriage (completed years)  18.70 4.64 18.87 4.69 Marital status: never married (1/0)  0.03 0.18 0.04 0.21 Marital status: married (1/0)  0.92 0.28 0.90 0.29 Household size (continuous number) 4.40 1.82 3.85 1.47 *** Male-headed household (1/0)  0.89 0.31 0.89 0.31 Education Years of completed education (completed years)  2.89 3.52 3.68 3.72 *** Household head can read/write (1/0)  0.43 0.50 0.52 0.50 *** Housing Number of rooms in household (continuous numbers) 1.31 0.59 1.16 0.41 *** Number of rooms per capita in household (continuous 0.33 0.15 0.33 0.14 numbers) Space per household (sq ft) 127.53 54.74 121.44 48.47 ** Per capita space in household (sq ft) 32.14 17.73 35.08 18.32 *** Has a separate kitchen (1/0)  0.59 0.49 0.81 0.39 *** Roof material: tin (1/0) 0.79 0.41 0.70 0.46 *** Roof material: brick/cement (1/0) 0.20 0.40 0.27 0.45 *** Has access to improved toilets (1/0)  0.79 0.41 0.89 0.31 *** Gets water from an improved source (1/0)  1.00 0.05 0.94 0.23 *** Has electricity (1/0)  0.99 0.07 1.00 0.05 Service utilization Public primary/secondary school (1/0) 0.23 0.42 0.19 0.40 Private primary/secondary school (1/0) 0.37 0.48 0.33 0.47 Public health facility (1/0) 0.28 0.45 0.24 0.43 * Private health facility (1/0) 0.51 0.50 0.59 0.49 *** Local transport (bus) (1/0) 0.62 0.48 0.72 0.45 *** Long distance transport (1/0) 0.37 0.48 0.56 0.50 *** N  663    1713       Note: Table shows weighted means of individuals living in slum and nonslum areas. * p<0.1; ** p<0.05; *** p<0.01. 24 Understanding Female Labor Force Participation in Urban Bangladesh 4 WOMEN’S WORK: DRIVERS AND CONSTRAINTS This section presents key patterns related to women’s work drawn from the DIGNITY data. We begin with the use of time in activities that individuals perform during the course of the day, then drill down to statistics about work and the nature of work. Data about expected constraints and drivers of female labor are also discussed in this section. 4.1 Time-Use Data in the DIGNITY Survey Time-use data report the amount of time individuals spend on activities such as paid work, chores, childcare, leisure, and self- care activities. The data allow us to better understand how people allocate their time across a variety of activities. This data also reveals how men and women allocate their time differently, partly due to gender roles; and how this may affect women’s economic empowerment. In most countries where data are available, women bear a disproportionately higher responsibility for unpaid work, and spend proportionately less time in paid work than men. Unfortunately, generally speaking, time-use data is not widely available (Rubiano-Matulevich 2018). In the case of Bangladesh, the only time-use survey that has been conducted is the 2012 Pilot Survey for Time Use. Attempting to fill the data gap and test new methodology on time use, the DIGNITY survey includes a module on time use, building on earlier survey work, such as the 2012 pilot survey, and the Women’s Empowerment in Agriculture Index surveys developed by the International Food Policy Research Institute (IFPRI). The DIGNITY survey used the one-day recall method to collect time-use data: the data reveal sharp gender differences in the use of time (Table 6). Men spend much more time in paid employment, while women spend more time on unpaid domestic work, childcare, and elderly care. Men tend to work long hours, about 10 hours on average, while women spend about 4 hours per day in employment-related activities. Comparing the DIGNITY results with the 2012 Time Use Pilot suggests that women in low-income communities and slums work longer hours, on average, than women in urban areas on the whole (4 vs. 2 hours per day), but allocate fewer hours 25 What Works for Working Women? to domestic and care work. As with the urban average, we observe that time spent on personal maintenance, leisure, and other activities is slightly higher among women than among men. Finally, the burden of childcare and domestic work disproportionately falls on women; women’s time spent on these activities is eight times greater than that of men (0.34 hours a day for men vs. 5.16 for women). Table 6 Typical Day of Men and Women in the DIGNITY Sample Male Female Test (Male - Female) Mean SD Mean SD Main Activities (continuous hours) Personal activities 11.45 2.32 12.12 2.61 *** Other activities 2.24 2.85 2.81 2.23 *** Working 9.97 4.14 3.91 4.72 *** Domestic work 0.15 0.6 4.34 2.22 *** Providing care 0.19 0.64 0.82 1.36 *** Total 24.00   24.00      Secondary activity Providing childcare as a secondary activity 0.41 1.42 1.82 2.84 *** N 1117   1259      Note: Table shows weighted means. * p<0.1; ** p<0.05; *** p<0.01. Time spent on childcare is prone to omission, as caregivers may combine it with other tasks: for example, they may be preparing meals or watching TV while they are also watching their children (Rubiano-Matulevich 2018). The DIGNITY survey attempts to address this problem by specifically asking respondents whether they also care for children while they are undertaking other activities. The results confirm our hypothesis: time spent on childcare is greater when accounting for the time that people spend caring for children while also performing other activities. The DIGNITY data show that women spend an additional 1.8 hours looking after children while performing other activities, more than twice the amount of time reported for direct childcare. On the other hand, men spend about 0.4 hours on secondary childcare, bringing men’s total time spend on childcare per day to about 36 minutes per day. This new data collection effort underscores the risks of not collecting data about secondary childcare activity, which could lead to grossly underestimating the need for childcare. Comparing the use of time for women who are in and out of the labor force reveals constraints to female labor force participation (FLFP) (see Table 7). Within the economically active population, the number of hours that men work per day is greater than that of women: on average, working women spend about 6.5 hours per day while men spend 3.7 hours longer (10.2 hours per day). On the other hand, working women spend more time on domestic work and providing care. The combined time that working women spend on these activities is comparable to the surplus working hours that men perform. While women who do not work report little time spent in direct childcare and elderly care, the time is complemented with secondary child care because they provide on average 2.6 hours of childcare as a secondary activity (that is, caring for children while they are also doing other activities). 26 Understanding Female Labor Force Participation in Urban Bangladesh Table 7 Time Utilization by Status of Labor Market Participation Male Female Test (Male - Female) Mean SD Mean SD Panel A: Individuals in the labor market (continuous hours) Personal activities 11.37 2.27 11.28 2.48 Other activities 2.11 2.76 2.06 1.94 Working 10.20 3.91 6.49 4.69 *** Domestic work 0.13 0.52 3.61 2.08 *** Providing care 0.19 0.64 0.56 1.09 *** Total hours per day 24.00 24.00 Providing childcare as a secondary activity 0.42 1.44 1.28 2.36 *** N 1086   753      Panel B: Individuals not in the labor market (continuous hours) Personal activities 14.04 2.36 13.28 2.32 * Other activities 6.19 2.88 3.83 2.20 *** Working 2.80 4.46 0.38 1.00 *** Domestic work 0.74 1.76 5.33 2.02 *** Providing care 0.24 0.60 1.18 1.61 *** Total hours per day 24.00 24.00 Providing childcare as a secondary activity 0.17 0.69 2.57 3.24 *** N 32   506      Note: Table shows weighted means. * p<0.1; ** p<0.05; *** p<0.01. The variation in women’s working hours is also high compared to that of men. Figure 4 shows a large variation in women’s working hours per day. Among the middle 50 percent of women, the number of working hours ranges from 2 to 10.5 hours per day. The variation is much greater than that for men; men’s working hours in the middle 50 percent ranges from 9 to 13 hours per day (Figure 5). Figure 4 Time-Use among Women, In and Out of the Labor Force 27 What Works for Working Women? Figure 5 Time-Use among Men, In and Out of the Labor Force 4.2 Work and the Nature of Work The labor force participation (LFP) rates are 58 percent and 97 percent among women and men respectively (see Table 8). The FLFP rate is remarkably higher than the norm found in urban Bangladesh: the 2016/17 Labor Force Survey (LFS) reports the labor force participation rate (LFPR) as 33 percent. However, the FLFP is higher in low- income areas in Dhaka than in more well-off neighborhoods. While the LFP of adult men in urban areas of Dhaka is at the same level as in low-income areas (around 94 percent), the participation rate of adult women is 50 percent higher in low-income areas of Dhaka than in urban Dhaka as whole (31 percent) (BBS 2018). Table 8 Who Works: Demographics and Educational Level of Individuals in the Labor Force Male Female   Mean SD Mean SD Test (Male - Female) Individual is economically active (1/0) - full sample 0.97 0.17 0.58 0.49 Age of respondent (completed years)  36.86 11.65 33.00 10.06 *** Age at marriage (continuous years)  22.26 4.03 15.72 2.73 *** Marital status: Never married (1/0)  0.08 0.27 0.01 0.12 *** Household size (continuous numbers)  3.98 1.53 3.80 1.61 ** Male-headed household (1/0) 0.92 0.27 0.81 0.39 *** Years of completed education (completed years)   3.62 3.85 3.07 3.52 *** Respondent received TVET (1/0)  0.10 0.30 0.10 0.30 N 1085   753     Note: Table shows weighted means. * p<0.1; ** p<0.05; *** p<0.01. TVET refers to technical and vocational education and training. As shown in Table 8, the average working woman is 33 years, 3.9 years younger than her male peers (36.9 years). Only 1 percent of the women in the labor force have never been married, while the figure is eight times higher among men. Women are marginally less well educated than their male peers. While the average woman has 28 Understanding Female Labor Force Participation in Urban Bangladesh completed 3.1 years of schooling, the corresponding figure for men is 3.6 years. Surprisingly, the proportion of individuals who have received TVET is 10 percent for both genders. Table 9 Nature of Work Male Female   Test (Male - Female) Mean  SD  Mean  SD  Individual engaged in wage-employment (1/0)  0.75 0.43 0.75 0.44 Individual engaged in self-employment (1/0)  0.23 0.42 0.11 0.31 *** Employment industry  RMG worker (1/0) 0.03 0.16 0.19 0.39 *** Self-employed (1/0) 0.26 0.44 0.11 0.31 *** Service (1/0) 0.27 0.44 0.10 0.30 *** Porter/day labor (1/0) 0.11 0.31 0.08 0.28 * Maid/servant (1/0) 0.00 0.00 0.31 0.46 *** Payment and terms  Remuneration received monthly (1/0)  0.44 0.50 0.84 0.36 *** Total monthly income (wage + business) (BDT)  13,076 7,235 5,546 4166 *** Contract type: Permanent (1/0)  0.33 0.47 0.41 0.49 *** Contract type:  Temporary (1/0)  0.27 0.45 0.35 0.48 *** Contract type:  Casual/daily work (1/0)  0.13 0.34 0.04 0.19 *** N  1085   753      Note: Table shows weighted means. * p<0.1; ** p<0.05; *** p<0.01. Among those who work, three-quarters of both men and women work as wage earners, as shown in Table 9. Engagement in self-employment shows some stark differences across gender lines. Only 11 percent of women engage in self-employment, compared to 23 percent of men. The proportion of women working in the ready-made garment (RMG) sector overwhelms the number of men (19 percent vs. 3 percent). The opposite is true in the service sector, where 27 percent of men are engaged, compared to 10 percent of women. The proportion of individuals working as porters or day laborers is more evenly split along gender lines: 11 percent of men, and 8 percent of women. Lastly, being a rickshaw puller is entirely a male profession (about 18 percent of men), while all maids are women (31 percent of women who work). As with the reported national statistics from the LFS, a vast majority (about 80 percent) of women who do not work indicate household work as the reason for not working (Table 10). Table 10 Primary Reasons for Not Participating in the Labor Force   Male Female Test (Male - Female)   Mean SD Mean SD Household work (1/0) 0.00 0.00 0.79 0.41 *** Illness (1/0) 0.67 0.47 0.11 0.31 *** Other (1/0) 0.33 0.47 0.10 0.30 *** N 51   556      Note: Table shows weighted means. * p<0.1; ** p<0.05; *** p<0.01. While the number of women who report being paid on a monthly basis is nearly double that of men (84 vs. 44 percent), men on average report earning nearly three times as much as women – 13,076 taka compared to 5,546 taka. In terms of contract type, 41 percent of women report having permanent jobs, in contrast to only 33 percent 29 What Works for Working Women? of men, while temporary employment is reported by a higher proportion of women than men (35 percent vs. 27 percent). Finally, casual or daily employment is more than three times as prevalent among men (13 percent) than among women (4 percent). 4.3 Skills This section explores the standardized cognitive and noncognitive skills of the urban poor. The importance of collecting information about personal attributes in order to account for unobservable innate skills that may play significant roles in the level of educational and/or income attainment has long been discussed. These skills, however, are difficult to quantify, due to the difficulty that respondents in low-capacity development settings often have in comprehending such questions, and/or the difficulty of developing valid measures for them. For instance, the well-used modules that are used to study cognitive capacities, such as the Raven’s test5, are difficult to compare across different contexts. Data on these skills were recently collected from 1,300 individuals in low- income neighborhoods in urban Bangladesh in partnership with Knack. This company uses a tablet-based game with predictive analytics to quantify the traits of respondents in a globally standardized way. These traits include “grit” and “thinking critically,” among others. Analysis of this data, if predictive, can provide insight into the skills and hidden talent that can lead toward potential types of careers; and the outcomes can be compared to those of respondents from across the globe. This may also allow the identification of traits that may be holding workers back from excelling and provide concrete policy guidance for interventions. Overall, the data reveal that men show higher Knack scores in areas such as “acting with self-confidence,” “growth mindset,” and “thinking creatively,” while women tend to have a higher score in the “engaging with people” dimension. The data reveal two general observations about gender and labor force participation. First, working women tend to have different traits than nonworking women. Figure 6 shows that working women show a higher score in “being self-disciplined,” but a lower score in “seeking knowledge” than nonworking women. Second, working men show an even higher score in “being self-disciplined” than working women (Figure 7), while working women score higher than working men in “gritting things out.” Figure 6 Cognitive and NonCognitive Skills among Women, In and Not In the Labor Force 5 Raven’s test is a nonverbal group test typically used to measure abstract reasoning and fluid intelligence. 30 Understanding Female Labor Force Participation in Urban Bangladesh Figure 7 Cognitive and Noncognitive Skills among Men In and Not In the Labor Force 4.4 Childcare The presence of children, particularly children younger than 5 years old, has been shown to be a constraint to FLFP in Bangladesh and in other countries (Solotaroff et al 2019). The DIGNITY data show that while most women have children, there are differences in the characteristics of the children and in access to childcare between those women who are in the labor force, and those who are not. Table 11 compares these characteristics. Nonworking women are more likely to have children than working women (87 vs. 84 percent), and their children tend to be younger (7.2 vs. 8.4 years old). Finally, when looking at the figures for young children who need care, the difference becomes more noticeable. While 32 percent of working women have a child that needs care, this ratio is 15 percentage points higher among women who do not work. The DIGNITY data suggest therefore that lack of access to childcare appears to be related to women not being in the labor force— or the decision to work at home instead of outside of the home—since the majority of women report that they cannot find childcare when it is needed.6 Table 11 Need for Childcare In the labor Not in the force labor force Test (In - out of the labor force) Mean SD Mean SD Has a child (1/0) 0.84 0.37 0.87 0.34 Age of youngest child (continuous years) 8.42 5.86 7.22 6.78 *** Has child below 5 years age (1/0) 0.32 0.47 0.47 0.50 *** Childcare is available when required (1/0) 0.15 0.36 0.12 0.33 N 753   508     Note: Table shows weighted means of labor force participant and non-participant women. * p<0.1; ** p<0.05; *** p<0.01. 6 The question is “If you wanted to do something (livelihood-related, training-related, self-care) and could not take your child with you, is there someone who could care for your child in your absence?” and the responses are “yes/no.” 31 What Works for Working Women? The unmet demand for childcare appears to be large, particularly among working women. The DIGNITY survey also asked respondents with young children about their childcare arrangements while at work. While 63 percent of men rely on their spouses for childcare, only about 13 percent of working women can do so (see Table 12). However, the time that husbands spend in taking care of children (even when he is the caregiver) is also relatively limited. In such a situation, working women use other alternatives, such as help from other household members (16 percent) or from relatives outside of the household (14 percent). This apparently leaves a large number of children being cared for by no one. It should be noted that some children may be old enough to take care of themselves,7 while in other cases, the parents might be working close by. Interestingly, very few people use formal arrangements such as a school or nursery, which may suggest something about the affordability or availability of formal childcare. Table 12 Unmet Demand for Childcare Male Female Test (Male - Female) Mean SD Mean SD Spouse (1/0) 0.63 0.48 0.13 0.34 *** Other household member (1/0) 0.09 0.29 0.15 0.36 *** Relative not living in the house (1/0) 0.05 0.23 0.14 0.34 *** Other individual (1/0) 0.01 0.08 0.02 0.13 ** School/nursery/other facilities (1/0) 0.00 0.07 0.01 0.10 Noone (1/0) 0.21 0.41 0.55 0.50 *** N 772   516      Note: Table shows weighted means of sample that participate in the labor force. The question about caretakers is administered to adults with a youngest child younger than 15; the average age of children discussed is 5.6 years. * p<0.1; ** p<0.05; *** p<0.01. 4.5 Transportation Access to transportation is of key importance in accessing job opportunities. However, men and women tend to have different kinds of experience with transportation in terms of affordability, reliability, and safety. Recent research has shown high levels of gender-based violence (GBV) in public transport and the adjacent public spaces in many cities around the world (Dominguez Gonzalez, Arango, McCleary-Sills and Bianchi 2015). In Bangladesh, a recent study in the Dhaka area by BRAC reveals that 94 percent of female users of public transportation have experienced verbal, physical, or other forms of sexual harassment. Moreover, 20.5 percent of women have stopped using public transportation due to sexual harassment (Daily Star 2018). Such constraints may cause women to modify their travel patterns and thus their choice of employment. Analysis of daily commuting patterns in Dhaka shows that members of the poorest quintile predominantly commute by foot, with a mean travel time of 40 minutes, and the average distances travelled by women are even shorter. The physical limits of commuting on foot means nearly all workers in the poorest quintile live within 4-5 kilometers from their place of employment, thereby curtailing their employment opportunities (Hill and Rahman, Forthcoming). Analysis of the DIGNITY data in Table 13 suggests that unequal access to transportation appears to be a constraint for women’s work. Only 35 percent of working women travel outside their community to work, in sharp contrast to men, where working outside the community is the norm for the majority (56 percent). Data about modes of transportation used by workers underscore the gender gaps. Women rely heavily on walking to get to work, about 13 percentage points higher than men. Moreover, very few women use public buses for commuting; only about 7 Even when looking at the results for children younger than five years old, many parents (38 percent of women and 6 percent of men) reported that no one was taking care of their children while they are at work. 32 Understanding Female Labor Force Participation in Urban Bangladesh 3 percent take buses to go to work, while 8 percent of men take the bus. Finally, the data also reveal that women spend roughly the same amount of time as men commuting to work. For example, female garment workers and maids spend roughly 18 minutes in their commute to work. This pattern suggests that women face a constraint finding jobs close to their dwellings. As indicated in studies on GBV, the risk of GBV inside buses is a strong constraint that prevents women from using buses and taking employment opportunities that are located further away from their homes. The data on travel time shows that travel costs for women are much lower than that of male workers (14 taka vs. 25 taka); but their commute time is longer than men’s (19 minutes each way, compared to 17 minutes for men). The risk of sexual harassment and assault may explain the low rate of female uptake in public transportation. Table 13 Employment and Transportation Male Female Test (Male - Female) Mean  SD  Mean  SD  Works at home (1/0) 0.10 0.30 0.25 0.43 *** Works within the community (1/0) 0.34 0.47 0.40 0.49 ** Works outside the community (1/0)  0.56 0.50 0.35 0.48 *** Transport used for work *** Walk (1/0)  0.74 0.44 0.87 0.34 *** Bicycle (1/0) 0.04 0.00 0.00 0.00 *** Rickshaw/cycle van (1/0) 0.09 0.18 0.02 0.12 *** Bus (1/0) 0.08 0.27 0.03 0.18 Cost of one-way travel to a job (BDT)  24.76 21.56 13.85 10.58 *** Time for one-way travel to non-home work (min- 16.66 13.40 19.17 12.89 *** utes)  N  1085   653      Note: Table shows weighted means of transport utilization and characteristics among individuals who participate in the labor force. * p<0.1; ** p<0.05; *** p<0.01. 4.6 Perceptions, Gender Norms, and Attitudes Gender norms affect women’s ability to participate in the workforce. These norms can be unpacked into several areas that directly constrain women’s ability to work, including mobility restriction and seclusion, among other things. While women’s roles in the economic sphere have expanded, conservative gender norms may still be restricting their physical mobility, and imposing various social practices on them in relation to the observance of purdah. These social practices are seen in conventions or rules about veiling, ranging from the wearing of simple head scarves to full-body burqas (Ahmed and Sen 2018). Moreover, these norms promote the seclusion of women; enforce their exclusion from public places; and specify gender-specific labor (Amin 1997). In addition, as elsewhere, GBV affects Bangladeshi women: in 2015, more than half of married women (54.7 percent) reported experiencing violence at home during the past 12 months. In addition, more than 25 percent of women have experienced physical violence from nonpartners during their lifetime, and 6.2 percent reported experiencing it in the past 12 months. The prevalence rate is lowest in city corporation areas (16 percent), which is lower than the urban average (24 percent), as well as that in rural areas (29 percent) (BBS 2016). 33 What Works for Working Women? The DIGNITY data reveal large gender gaps in mobility; women, on average, have more limited mobility than men and tend to be homebound, with little freedom to travel outside their communities. While 84 percent of men go outside of their community every day, only 40 percent of women do so (see Table 14). What’s more, a quarter of women only leave their community once a month, and one in ten never leave their community. This means that there are many women in Dhaka that live life as if they were in a remote village, even though they live in one of the biggest cities in the world. Going outside the community is often not a decision made unilaterally: 35 percent of women said that they have to ask permission from someone in their household before going outside their community. Table 14 Frequency of Travel Outside the Home Male Female   Test (Male - Female) Mean SD Mean SD Frequency of travel outside the home Every day 0.84 0.37 0.40 0.49 *** At least once a week 0.09 0.28 0.12 0.32 ** At least once every two weeks 0.01 0.12 0.04 0.20 *** At least once a month 0.02 0.14 0.10 0.30 *** Less than once a month 0.03 0.17 0.25 0.43 *** Never 0.01 0.12 0.09 0.29 *** Time spent (by space - in hours) At home  13.85 5.09 20.45 4.63 *** Outside home but within the community  4.50 5.44 2.17 3.80 *** Outside the community  5.66 6.18 1.39 3.53 *** N 1117   1259      Note: Table shows weighted means of how frequently individuals travel outside the home. * p<0.1; ** p<0.05; *** p<0.01. Women’s freedom to leave the house, enter public spaces, and participate fully as citizens are important elements in their economic empowerment. Collecting location of activity data is another innovation of the time-use module in the DIGNITY survey. In addition to asking about activities that respondents have performed over the past 24 hours, the survey also asks where the activities take place—at home; outside the home, but inside the community; or outside the community. The results reveal a stark contrast in mobility between genders: men spend four times the amount of time that women do outside the community. The time-use module of the survey reveals that women spend more time at home than men (20 vs. 14 hours per day), while men spend more time than women outside the home, but in the community (4.5 vs. 2.2 hours), and outside the community (5.6 vs. 1.4 hours per day). In short, women spend more than four-fifths of their day homebound. The risk of GBV also greatly impacts women’s decision making. Concerns about safety and security play a large role in determining whether a woman will participate in the labor force. Violence and abuse against women in public spaces and in the workplace are not uncommon. This issue is particularly acute in urban settings, where women, largely poor, are heavily reliant on walking as their primary mode of commuting. The DIGNITY data reveal a big gender gap in perception of safety, especially outside the home. Risk of sexual harassment while being outside the home may therefore be one of the main constraints for women in seeking employment. 34 Understanding Female Labor Force Participation in Urban Bangladesh Most women feel safe at home during the day (98 percent), which is the same rate as for men.8 But there is a large gender gap for perceptions of night-time safety outside the home. The percentage of women who reported that they feel safe outside the home at all times (either on their own street or outside their community) is only 69 percent, compared to 96 percent of men. This data shows that the perception of safety has a disproportionate effect on women. Such gender differences may affect their decision to travel outside the home and to participate in economic activities. Social norms are also reflected in veiling practices. Women reported modifying their veiling practices when they travel further away from their home. About 50 percent of female respondents reported that they wear a burqa when they go outside of their neighborhood, while about 20 percent said they only cover their heads with headscarves while in the neighborhood, but that they wear a burqa when going outside of their neighborhoods. Table 15 Perceptions, Norms and Attitudes about Safety and Freedom of Movement Not in the In the labor All women Test (In - out of the labor Perceptions, norms and attitudes labor force force force) Mean SD Mean SD Mean SD Needs permission to go outside (1/0) 0.35 0.48 0.37 0.48 0.33 0.47 Feels safe outside home (1/0) 0.69 0.46 0.63 0.48 0.73 0.44 *** Wears veil (burqa) when walking outside own 0.49 0.5 0.59 0.49 0.42 0.49 *** neighborhood (1/0) N 1261   508   753     Note: Table shows weighted means for women. * p<0.1; ** p<0.05; *** p<0.01. Finally, Table 15 reveals a sharp contrast in the indicators related to agency9 between working and nonworking women. While 37 percent of nonworking women need to seek permission to go outside the community, only 33 percent of working women do. Similarly, working women tend to have a greater perception of their safety; they report feeling safe outside home (either inside the neighborhood or outside the neighborhood) much more than women who are not in the labor force (73 percent vs, 63 percent). Veiling practices are also associated with lower employment rates. These relationships will be explored in a multivariate setting in Section 5. 8 Data not shown in the table. 9 Agency means an individual’s (or group’s) ability to make effective choices and to transform those choices into desired outcomes. Agency can be understood as the process through which women and men use their endowments and take advantage of economic opportunities to achieve desired outcomes (World Bank 2011). 35 What Works for Working Women? 36 Understanding Female Labor Force Participation in Urban Bangladesh 5 EXPLORING CONSTRAINTS TO WORK This section explores how the factors in the previous section correlate to three key labor market outcomes: FLFP; employment choice (whether one works as a wage earner or is self-employed); and labor supply (the amount of time a worker spends working per day). The conditional correlations are explored using a linear model, with neighborhood fixed effects and robust standard errors. The factors considered here are factors that the narrative on female labor market outcomes would suggest are important, but the relationships presented in this section are purely linear correlations: they do not indicate a causal or deterministic relationship. Selection into labor market participation has not been modeled for the regressions presented on employment choice and labor supply. The right-hand side variables are similar across the models. These variables include age, age at marriage, marital status, household size, years of education completed, years of completed education squared, whether the respondent received technical and vocational education and training (TVET), has a child that requires care, is in a male-headed household, number of years spent in the community, whether they were born in the community, and whether the respondent is depressed (CESD Score > 10). Additional variables related to gender norms are added in order to model specification of regressions for female labor market outcomes—for example, whether the respondent needs permission to go outside the community, feels safe outside the home, and/or wears a burqa whenever outside the home—and also data about the educational level of the spouse. Table 16 presents the correlates of female labor force participation (FLFP). The model is restricted to women, since most men (97 percent) participate in the labor market. Younger women are marginally more likely to participate in the labor force; and they are 0.5 percentage points more likely to work for each year younger that they are. The probability of being in the labor force increases by 35 percentage points among never-married women. Women with 37 What Works for Working Women? more education are also less likely to participate in the labor force: each additional year of education reduces labor force participation (LFP) by 4.2 percentage points. While this result may seem surprising, it fits with the U-shaped curve relationship between education level and FLFP in Bangladesh (mentioned in section 2). Specifically, they are located in the left part of the U-shaped curve, as DIGNITY is collected from the low-income areas (the average years of completed education for women is only 3.5 years). However, those with TVET are significantly more likely to work (by 23.1 percentage points), suggesting that TVET is undertaken with specific employment opportunities in mind. In line with the existing literature, women with children five years of age or less that require care are 15 percentage points less likely to participate in the labor force. The duration of residence in a community does not appear to be correlated with the decision to enter the labor market. Perceptions and attitudes are strongly associated with FLFP. Women who perceive the environment outside the home to be safe and who are comfortable traveling outside their homes by themselves are 9.6 percentage points more likely to participate in the labor market. On the flip side, women with conservative outlook10 are 8.1 percentage points less likely to engage in the labor market. Lastly, depression is strongly associated with LFP: women who report being depressed are 10.5 percentage points more likely to be working. More research may be needed to understand this correlation. However, during fieldwork a few women reported that the ideal is to not be working, and that they worked only because they had to for financial reasons rather than because they wanted to. This dissatisfaction with life outcomes may account for what is being expressed in this variable. Table 17 presents the coefficients of a model that explores the correlates of employment type: that is, whether an individual chooses to engage in wage earning or self-employment. (Note that wage earning takes the value of 1 and self-employment takes the value of 0). Column 1 shows the coefficients for women: younger women11 and women with fewer years of education are more likely to be self-employed (0.5 and 3.5 percentage points respectively). Having gone through TVET is positively associated with being self-employed (27.9 percentage points). Women with children of an age requiring care are similarly more likely to be engaged with self-employment than wage employment. Finally, the probability of being engaged in self-employment rather than wage employment is higher among women who require permission to go outside the home, or who wear burqas when they do. By comparison, younger men are more likely to be self-employed. Using time-use data with a 24-hour recall period from 4 am the previous day to 3:59 am on the day of the survey, we explored the correlates of the number of hours an individual decided to work (Table 18). Among women, while the number of years of education is negatively associated with labor supply (15.8 percentage points), based on the positive sign and significance (1 percent< p) of the quadratic term (1.4 percentage points), this correlation may reverse as the amount of education increases (see Column 1). Understandably, a woman with a child requiring care is similarly likely to spend less time at work (by 49.6 percentage points). In line with previous models, age is the sole significant correlate of the length of time a male individual spends at work that is negatively correlated; that is, younger men are likely to spend less time at work (by 1.1 percentage points). The analysis also explores whether there are any differences between slum and nonslum areas in terms of constraints to women’s labor market participation. Table 1 in the Annex presents the results of whether being in slums changes the magnitude of correlations between respondent characteristics and labor market outcomes. Women who have TVET and who live in slum areas are 20.9 percentage points more likely to engage in the labor 10 Conservative outlook is proxied by burka wearing practice, which is also used by Ahmed and Sen (2018) and Asadullah and Wahhaj (2019). 11 It should be noted that DIGNITY sample mostly comprises principle couple of the households, thus may not capture young women working in the garment industry. 38 Understanding Female Labor Force Participation in Urban Bangladesh force than those in nonslum areas. But in slum areas, women who work tend to have received TVET more than women in other areas. Finally, women who live in slums and who have children who require care are 18.9 percentage points less likely to engage in the labor market than nonslum women, suggesting a greater need for childcare in slum areas. This may also suggest that access to childcare (either through formal or informal sources) is more limited for women in slum areas than for women living not living in slums. A household with an additional member is associated with an increase in the likelihood (by 5.2 percentage points) of labor force participation among women living in slums than otherwise. The correlation between other factors, such as duration of living in the community, perceptions about social norms, and LFP do not vary between slum and nonslum areas. Finally, in terms of perception of safety and FLFP, being in a slum is no different from being in a nonslum area. In the same table, Column 3 examines whether the correlation between respondent characteristics and the choice of whether to engage in wage earning or self-employment differs by type of location. We find that women in slum areas are 0.9 percentage points more likely to work as wage employees as their age increases by one year than women from nonslum settings. With regard to the amount of time spent at work (Column 5), each additional year of education is correlated with 30.5 percentage points more time spent at work by women living in slum areas than nonslum ones. The negative and significant coefficient of the quadratic education term, however, suggests that the effects may reverse for slum-dwelling women faster than for their nonslum counterparts as the number of years of education increases. For women with TVET, the women from slums are likely to spend 96.2 percentage points more time at work than women living in nonslum areas. Lastly, cultural norms such as those that expect women to have permission to go outside appear to play a larger role in slum areas when it comes to the time spent at work: women from slums are likely to spend 79.2 percentage points less time working than their nonslum counterparts. Next we examine the correlation of cognitive and noncognitive skills and labor market outcomes using a specification similar to the one used earlier; the results are presented in Tables 19, 20, and 21. This is assessed separately, since the information on cognitive and noncognitive skills was collected from a subset of participants (593 women and 507 men). The results suggest that while traits such as being able to adapt to change, persistence and grit, self- discipline, and/or seeking knowledge are not strongly correlated to whether an individual chooses to engage in the labor market, they are significantly correlated to whether the woman chooses wage earning or self-employment (see Table 20, Column 1). Women who are better at adapting to change are 6.3 percentage points—and those who are self-disciplined are 5.7 percentage points—more likely to engage in wage employment. Alternately, women who are persistent and willing to “grit things out” are 10.2 percentage points more likely to be engaged in self- employment. None of these factors are significantly correlated with the type of employment that men undertake. Overall, these regressions highlight the fact that the correlates of FLFP are distinct from those of men. First of all, LFP is ubiquitous among the male population in our sample. Delving deeper into the characteristics of men that are correlated with the choice to engage in wage earning or self-employment, in most instances, men seem to be affected very little by demographic, socioeconomic, or societal norms and attitudes. However, these factors are significantly correlated with female LFP. While unmarried women who have no children and who do have TVET are the most likely to be engaged in the labor market, the opposite holds true for married women living in a conservative male-headed household. Businesses in urban spaces are often run from home: this allows for more flexibility, especially for women with young children who require care. Similarly, women from conservative households12are more engaged with self-employment, most likely from within their homes. 12 A conservative household is defined as a household in which women are required to ask permission to go outside the community, or are required to wear a burqa when stepping outside the home. 39 What Works for Working Women? Table 16 Correlates of Labor Force Participation Economic Activity Coefficient Standard Error   (1) (2) Age of respondent (completed years) -0.005** 0.002 Age of marriage 0.002 0.007 Marital status: never married (1/0) 0.350** 0.148 Household size 0.000 0.012 Years of completed education -0.042** 0.017 Years of completed education (squared) 0.002 0.001 Respondent received TVET (1/0) 0.231*** 0.069 Has child (<=5yrs) requiring care -0.150*** 0.042 Lives in male-headed household (1/0) -0.208*** 0.055 Duration in community: continuous years -0.002 0.003 Born in the community -0.039 0.070 Needs permission to go outside -0.005 0.042 Feels safe outside home 0.096** 0.041 Wears burqa whenever outside the home (1/0) -0.081** 0.040 Education of spouse (for women) 0.008 0.011 Respondent is depressed (CESD Score > 10) (1/0) 0.105** 0.049 R2 0.21   N 1,247   Note: Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. As 97 percent of men participate in the labor force, they are excluded from this model. * p<0.1; ** p<0.05; *** p<0.01. 40 Understanding Female Labor Force Participation in Urban Bangladesh Table 17 Correlates of Wage Earning versus Self-Employment Female Male Standard Standard   Coefficient Coefficient Error Error (1) (2) (3) (4) Age of respondent (completed years) -0.005* 0.003 -0.008*** 0.002 Age of marriage -0.004 0.008 -0.003 0.005 Mar status member: never married (1/0) 0.177 0.229 -0.391 0.316 Household size 0.003 0.012 -0.018 0.012 Years of completed education -0.035* 0.020 -0.021 0.013 Years of completed education (squared) 0.003 0.002 0.001 0.001 Respondent received TVET (1/0) -0.279*** 0.086 0.034 0.065 Has child (<=5yrs) requiring care -0.165*** 0.050 -0.056 0.040 Live in male headed household (1/0) 0.018 0.053 -0.019 0.073 Duration in community: continuous years -0.003 0.003 0.000 0.003 Born in the community 0.036 0.078 -0.092 0.059 Needs permission to go outside -0.084* 0.051 Feels safe outside home 0.013 0.048 Wears burqa whenever outside the home (1/0) -0.173*** 0.046 Education of spouse (for women) -0.005 0.013 Respondent is depressed (CESD Score > 10) (1/0) 0.058 0.053 -0.033 0.062 R2 0.31   0.22 N 742   1,001 Note: Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. The dependent variable takes on the value of 1 if the person engages in wage employment, and the value of 0 for the self-employed. Sample is restricted to individuals who report being in the labor force in the past 30 days. Questions such as the need to ask for permission to go outside, perceptions of safety, and burqa wearing are asked of female respondents only, and as such are left out of the male- specific model. * p<0.1; ** p<0.05; *** p<0.01. 41 What Works for Working Women? Table 18 Correlates of Hours Spent at Work (log-linear outcome) Female Male Standard Standard Coefficient Coefficient Error Error   (1) (2) (3) (4) Age of respondent (completed years) -0.010 0.008 -0.011** 0.005 Age of marriage 0.006 0.025 -0.002 0.010 Marital status: never married (1/0) 0.040 0.627 -0.341 0.653 Household size -0.018 0.046 0.034* 0.021 Years of completed education -0.158** 0.063 -0.003 0.025 Years of completed education (squared) 0.014*** 0.005 0.001 0.002 Respondent received TVET (1/0) -0.696** 0.271 -0.027 0.127 Has child (<=5yrs) requiring care -0.496*** 0.158 -0.066 0.077 Lives in male-headed household (1/0) -0.131 0.169 -0.127 0.096 Duration in community: continuous years 0.005 0.010 -0.009 0.006 Born in the community 0.262 0.298 -0.244* 0.137 Needs permission to go outside -0.152 0.160 Feels safe outside home 0.151 0.150 Wears burqa whenever outside the home (1/0) -0.219 0.173 Education of spouse (for women) -0.002 0.042 Respondent is depressed (CESD Score > 10) (1/0) -0.227 0.175 -0.199 0.145 R2 0.26   0.12 N 742   1,001 Note: Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. Sample restricted to individuals who report being in the labor force in the past 30 days. Questions concerning the need to ask for permission to go outside, perceptions of safety and, burqa wearing are asked only of female respondents, and as such are left out of the male-specific model. * p<0.1; ** p<0.05; *** p<0.01. 42 Understanding Female Labor Force Participation in Urban Bangladesh Table 19 Correlation between Cognitive and Noncognitive Skills and Female Economic Activity   Female   Coefficient Standard Error   (1) (2) (Ln) Knack 1: Adapting to change 0.007 0.030 (Ln) Knack 2: Gritting things out 0.038 0.031 (Ln) Knack 9: Being self-disciplined 0.026 0.022 (Ln) Knack 17: Seeking knowledge -0.044 0.033 Age of respondent (completed years) -0.003 0.003 Age of marriage -0.002 0.009 Household size 0.003 0.016 Years of completed education -0.036 0.023 Years of completed education (squared) 0.001 0.002 Respondent received TVET (1/0) 0.293*** 0.087 Has child (<=5yrs) requiring care -0.193*** 0.051 Lives in male-headed household (1/0) -0.169** 0.074 Duration in community: continuous years -0.004 0.004 Born in the community -0.058 0.094 Needs permission to go outside -0.022 0.053 Feels safe outside home 0.099* 0.053 Wears burqa whenever outside the home (1/0) -0.082 0.053 Education of spouse (for women) 0.007 0.014 Respondent is depressed (CESD Score > 10) (1/0) 0.073 0.063 R2 0.25   N 593   Note: Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. As 97 percent of men participate in the labor force, they are excluded from this mode. * p<0.1; ** p<0.05; *** p<0.01. 43 What Works for Working Women? Table 20 Correlation between Cognitive and Noncognitive Skills and Type of Employment (Wage or Self- Employment) Female Male Standard Standard   Coefficient Coefficient Error Error (1) (2) (3) (4) (Ln) Knack 1: Adapting to change 0.063* 0.035 0.005 0.031 (Ln) Knack 2: Gritting things out -0.102*** 0.039 -0.026 0.027 (Ln) Knack 9: Being self-disciplined 0.057** 0.025 -0.032 0.020 (Ln) Knack 17: Seeking knowledge -0.012 0.042 -0.032 0.032 Age of respondent (completed years) -0.003 0.004 -0.006** 0.002 Age of marriage -0.001 0.009 -0.003 0.006 Household size 0.011 0.016 -0.021 0.015 Years of completed education -0.017 0.026 -0.005 0.017 Years of completed education (squared) 0.002 0.002 0.000 0.002 Respondent received TVET (1/0) -0.356*** 0.088 0.021 0.083 Has child (<=5yrs) requiring care -0.151** 0.064 -0.038 0.052 Live in male-headed household (1/0) -0.014 0.072 -0.067 0.077 Duration in community: continuous years -0.006 0.004 0.001 0.003 Born in the community 0.059 0.109 -0.108 0.076 Needs permission to go outside -0.042 0.063 Feels safe outside home 0.003 0.061 Wears burqa whenever outside the home (1/0) -0.170*** 0.058 Education of spouse (for women) -0.015 0.018 Respondent is depressed (CESD Score > 10) (1/0) 0.071 0.062 0.012 0.076 R2 0.36   0.23 N 351   507 Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. The dependent variable takes on the value of 1 if the person engages in wage employment and the value of 0 for the self-employed. Sample restricted to individuals who report being in the labor force in the past 30 days. Questions concerning the need to ask for permission to go outside, perceptions of safety, and burqa wearing are asked of female respondents only, and as such, are left out of the male- specific model. * p<0.1; ** p<0.05; *** p<0.01. 44 Understanding Female Labor Force Participation in Urban Bangladesh Table 21 Correlation between Cognitive and Noncognitive Skills on the Hours Spent at Work Female Male Standard Standard   Coefficient Coefficient Error Error (1) (2) (3) (4) (Ln) Knack 1: Adapting to change 0.149 0.125 0.033 0.057 (Ln) Knack 2: Gritting things out 0.039 0.108 -0.001 0.060 (Ln) Knack 9: Being self-disciplined 0.121 0.087 0.010 0.055 (Ln) Knack 17: Seeking knowledge -0.051 0.126 -0.046 0.057 Age of respondent (completed years) -0.010 0.011 -0.010** 0.005 Age of marriage 0.007 0.031 0.007 0.012 Household size 0.037 0.058 0.039 0.028 Years of completed education -0.237*** 0.077 -0.002 0.029 Years of completed education (squared) 0.019*** 0.006 0.001 0.003 Respondent received TVET (1/0) -0.801*** 0.305 -0.084 0.167 Has child (<=5yrs) requiring care -0.504** 0.205 -0.042 0.093 Lives in male-headed household (1/0) -0.335 0.229 -0.156 0.097 Duration in community: continuous years -0.005 0.014 -0.008 0.008 Born in the community 0.039 0.426 -0.192 0.187 Needs permission to go outside -0.124 0.209 Feels safe outside home 0.129 0.194 Wears burqa whenever outside the home (1/0) -0.282 0.206 Education of spouse (for women) 0.065 0.054 Respondent is depressed (CESD Score > 10) (1/0) -0.303 0.219 -0.219 0.189 R2 0.34   0.16 N 351   507 Note: Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. Sample restricted to individuals who report being in the labor force in the past 30 days. Questions concerning the need to ask for permission to go outside, perceptions of safety, and burqa wearing are asked of female respondents only, and as such are left out of the male-specific model. * p<0.1; ** p<0.05; *** p<0.01. 45 What Works for Working Women? 46 Understanding Female Labor Force Participation in Urban Bangladesh 6 CONCLUSION Labor income growth has been the fundamental driver of poverty reduction in Bangladesh in recent years (Joliffe et al 2013, Hill and Genoni 2018). However, during the past five years urban poverty reduction has stagnated and, perhaps without coincidence, this has also been a period in which FLFP rates have fallen. As Bangladesh continues to urbanize, understanding the factors that constrain FLFP in urban areas is increasingly important in order to understand how urban income growth and poverty reduction can be ensured. Although the decline in FLFP may be driven in part by changes in the demand for female labor in the RMG sector in recent years, it is also possible that it is driven by supply-side factors—for example, norms around women’s work and travel outside the home, and factors that determine the cognitive and noncognitive skills of women as they contemplate entering the workforce. This paper has explored the factors that constrain women from engaging in the labor market. It uses a unique new dataset to provide individual-level data on labor market participation, time use, norms, and both cognitive and noncognitive skills. The paper highlights the fact that FLFP is higher in low-income neighborhoods and slums than in the city as a whole, showing again that FLFP tends to be lower at higher levels of income. At the individual level we also find that women with more education are less likely to work, another indicator of the same relationship. The results also show the well-known finding that although nearly all men work, FLFP is strongly determined by demographics— for example, by whether a woman is married or not, and by her age. However, the unique dataset used in this paper also allows us to quantify other important correlates of the labor force participation (LFP) decision. Access to childcare was shown to be strongly correlated with FLFP, and is perhaps one of the main drivers of demographic patterns of female employment found in the data. Women with children less than 5 years old are 15 percentage points less likely to be engaged in work than women with no children, or with older children. In addition, women reported that access to childcare was a main constraint, and detailed time-use data show that women are often looking after children even when they are engaged in other activities, something that would be impossible to do if working away from home. This points to the need to consider childcare when thinking about how to increase labor income for women in urban areas. This is a challenge that does not exist to the same extent in rural areas, where work is more often in and around the house or farm. 47 What Works for Working Women? Soft skills were also shown to be an important correlate of the nature of work that women engage in. Women who showed stronger skills in “being disciplined” were more likely to be employees; those with less discipline but more determination were more likely to be self-employed. This could point to the value of soft-skill training, and/ or mentorship programs to help women find jobs that they are better suited to, or to help them learn the soft skills that emerging employment opportunities require. However, perhaps the strongest finding from the data was the importance of mobility, the fear of safety in public spaces, and the implications of these factors for FLFP. The survey revealed that public spaces in Dhaka are often male-dominated spaces, and that this is correlated with women’s decisions about employment. Men spend four times more time outside of their communities than women do. A quarter of women only leave their community once a month, and one in ten never leave their community at all. This means that there are many women in Dhaka that live life as if they were in a remote village, even though they live in one of the biggest cities in the world. The notion that social norms play an important role in determining FLFP is not new, but this survey has provided a unique ability to quantify the strength of the correlation between these norms, perceptions, and labor participation. About 50 percent of the women interviewed wear a burqa outside of their community, and 30 percent report not feeling safe in their communities. Not surprisingly, women who do not feel that the environment outside of their house is safe are 10 percentage points less likely to participate in the labor market. Also, those who wear burqas are 8 percentage points less likely to engage in the labor market. The data showed that many women in Dhaka are in both categories. The analysis does not point to a causal relationship between these norms and participation in the labor market, but it does highlight the fact that norms are very different between women who work and those who do not. Finally, the evidence is consistent with the notion that for many women working is not deemed desirable, and is something that is not associated with feeling happy about life. New data and findings from this report could provide evidence for novel interventions to boost female labor force participation in urban areas. The report shows that many families do not have access to childcare, while very few use formal childcare centers. In this regard, community-based child care, like the Guatemala’s Community Day Care Centers Program, could be an option. Such programs usually organize parents in poor urban areas in to a small group and choose a woman from the group to be the care provider. The provider will care for the children, while the cost will be covered by the parents and the government (Ruel et al. 2002). An impact evaluation by Hallman et al. (2005) suggests that interventions to make formal day care in poor urban areas more affordable may increase the labor hours of mothers living in the neighborhoods. Public childcare programs, like the Rio de Janeiro’s public day-care program in Brazil, could be another option. The program provides childhood development program for infants and children from 0-3 years living in low-income neighborhoods. Barros et al. (2013) finds that access to the program led to a very high utilization rate and a substantial increase in mothers’ employment (from 36 to 46 percent). In terms of safety in the community, programs like UN-HABITAT’s global Safer Cities Programme could be a useful starting point in reassessing the earlier initiative to address crime and violence activities in low-income communities (UN-HABITAT 2011). In future research, it will be important to continue to explore the relationship between social norms and participation in the labor market in order to identify interventions that could help encourage FLFP. The evidence indicates that making public spaces more female-friendly, travel safer for women, and female work more socially acceptable may be important in order to increase the ability of women to grow their families’ incomes happily, and free of fear. Policy goals and actions should support women who choose to work in urban areas of Bangladesh, as opposed to simply being compelled to work due to economic necessity. Otherwise, the country’s gain in FLFP may start to erode as urban households move out of extreme poverty. Ideally, women’s work should be rooted in economic empowerment, and viewed as a means for gaining prosperity and progression to a middle-class society. 48 Understanding Female Labor Force Participation in Urban Bangladesh ANNEX Sampling Weights and Probability of Selection The probability of selection of the sample was calculated separately for each stage and enumeration areas (EA) using the following specification: where the probability of the ith EA being selected from the jth stratum is presented by . represents the number of selected EAs within each stratum. The number of households in the i EA in the j stratum is th th represented by while represents the total number of households in the stratum. The probability of the second stage of the selection process is estimated using the following specification: where the probability of a household being selected for the sample is represented by is the number of households of type t in EA i in stratum j selected to be surveyed from a total number of households ( ) within each EA and the particular category. The overall likelihood of a particular household being selected from a particular stratum is therefore represented as the product of the two aforementioned probabilities, calculated as follows: The weight is subsequently constructed as the inverse of the likelihood of a particular household being selected ( ). 49 What Works for Working Women? Annex Table 1: Heterogeneity of Employment Dynamics across Slum and Nonslum Areas Labor force Wage vs self- Time spent at work participation employment Standard Standard Standard Coefficient Coefficient Coefficient Error Error Error   (1) (2) (3) (4) (5) (6) Slum X Age of respondent -0.002 0.004 0.009* 0.005 -0.006 0.018 Slum X Age of marriage -0.007 0.012 0.016 0.013 0.041 0.043 Slum X Never married 0.324 0.301 -0.134 0.380 -0.379 1.067 Slum X Household size 0.052** 0.020 -0.035 0.025 0.014 0.102 Slum X Years of completed education 0.037 0.037 0.022 0.045 0.305** 0.142 Slum X Years of completed education (sq) -0.007** 0.003 0.000 0.004 -0.022* 0.012 Slum X Respondent received TVET 0.209* 0.126 0.046 0.151 0.962** 0.479 Slum X Has child (<=5yrs) requiring care -0.189** 0.077 0.065 0.109 0.135 0.341 Slum X Male-headed household -0.138 0.102 -0.030 0.108 0.115 0.398 Slum X Duration in community (continuous years) -0.002 0.005 -0.011 0.007 -0.002 0.020 Slum X Born in the community 0.060 0.121 -0.173 0.171 -0.883 0.545 Slum X Needs permission to go outside -0.071 0.077 -0.047 0.106 -0.792** 0.310 Slum X Feels safe outside home -0.025 0.073 -0.133 0.094 0.095 0.291 Slum X Wears burqa whenever outside the home -0.037 0.076 0.113 0.103 0.168 0.323 Slum X Education of spouse (for women) 0.012 0.024 -0.013 0.025 -0.117 0.082 Slum X Respondent is depressed 0.018 0.085 -0.168 0.117 0.175 0.352 Slum (1/0) 0.569* 0.322 -0.190 0.383 -0.351 1.068 Age of respondent (completed years) -0.005** 0.002 -0.006** 0.003 -0.007 0.009 Age of marriage (continuous years) 0.004 0.008 -0.006 0.009 -0.002 0.030 Marital. status: never married (1/0) 0.337** 0.171 0.147 0.248 0.187 0.733 Household size -0.013 0.015 0.012 0.015 -0.008 0.056 Years of completed education -0.046** 0.020 -0.040* 0.022 -0.185** 0.074 Years of completed education (squared) 0.003* 0.002 0.003 0.002 0.016*** 0.006 Respondent received TVET (1/0) 0.206** 0.080 -0.290*** 0.099 -0.794** 0.317 Has child (<=5yrs) requiring care -0.114** 0.049 -0.174*** 0.057 -0.503*** 0.178 Live in male-headed household (1/0) -0.191*** 0.064 0.037 0.062 -0.165 0.196 Duration in community: continuous years -0.002 0.003 -0.001 0.003 0.003 0.012 Born in the community -0.063 0.084 0.056 0.097 0.420 0.375 Needs permission to go outside 0.009 0.049 -0.080 0.059 -0.040 0.184 Feels safe outside home 0.102** 0.049 0.034 0.057 0.126 0.182 Wears burqa whenever outside the home (1/0) -0.075 0.046 -0.185*** 0.050 -0.238 0.199 Education of spouse (for women) 0.006 0.012 -0.004 0.015 0.007 0.048 Respondent is depressed (CESD Score > 10) (1/0) 0.105* 0.055 0.080 0.060 -0.262 0.194 R2 0.23   0.32   0.28   N 1,247   742   742   Notes: Table shows coefficients of a linear model with neighborhood fixed effects and robust standard errors. Sample is restricted to women. * p<0.1; ** p<0.05; *** p<0.01. 50 Understanding Female Labor Force Participation in Urban Bangladesh REFERENCES Ahmed, T. and B. Sen. 2018. “Conservative Outlook, Gender Norms and Female Wellbeing: Evidence from Rural Bangladesh.” World Development, Volume 111: 41-58. Amin, S. (1997). “The Poverty-Purdah Trap in Rural Bangladesh: Implications for Women’s Roles in the Family.” Development and Change, 28 (2): 213-33. Asadullah, M. Niaz and Z. Wahhaj. 2019. “Female Seclusion from Paid Work: A Social Norm or Cultural Preference?” ECARES Working Paper 2019-10. March 2019. Attanasio, O, Paes de Barros, R, Carneiro, P, Evans, D, Lima, L., Olinto, P and Schady, N. 2017. Impact of free availability of public childcare on labour supply and child development in Brazil, 3ie Impact Evaluation Report 58. New Delhi: International Initiative for Impact Evaluation (3ie). Bangladesh Bureau of Statistics. 2016. Census of Slum Areas and Floating Population 2014. Ministry of Planning, Government of the People’s Republic of Bangladesh, 2016. Dhaka, Bangladesh. ____. 2016. Report on Violence Against Women Survey 2015. Ministry of Planning, Government of the People’s Republic of Bangladesh. Dhaka, Bangladesh. ____. 2018. Report on Labour Force Survey (LFS) 2016-17. Ministry of Planning, Government of the People’s Republic of Bangladesh. 2018. Dhaka, Bangladesh. Banks, Nicola. 2013.”Female Employment in Dhaka, Bangladesh: Participation, Perceptions and Pressures.” Environment and Urbanization 25 (1): 95-109. Barros, Ricardo Paes de; Olinto, Pedro; Lunde, Trine; Caralho, Mirela. 2013. The impact of access to free childcare on women’s labor market outcomes: evidence from a randomized trial in low-income neighborhoods of Rio de Janeiro (English). enGender Impact: the World Bank’s Gender Impact Evaluation Database. Washington DC; World Bank. Bloom, David E., and Jocelyn E. Finlay. 2009. “Demographic Change and Economic Growth in Asia.” Asian Economic Policy Review 4 (1): 45-64. Bridges, S., Lawson, D. and Begum, S. 2011. Labour Market Outcomes in Bangladesh: The Role of Poverty and Gender Norms. European Journal of Development Research (2011) 23: 459. Daily Star. 2018. Public Transport: 94 Percent of Women Harassed. March 07, 2018. Dominguez Gonzalez, K., D. Arango, J. McCleary-Sills and B. Bianchi. 2015. Violence Against Women and Girls (VAWG) Resource Guide | Transport Brief. Retrieved from http://www.vawgresourceguide.org/sector-briefs on March 21, 2019. Farole, T., and Y. Cho. 2017. Jobs Diagnostic Bangladesh. Job Series Issue 9, Washington, DC: World Bank. Hallman, Kelly, Agnes R. Quisumbing, Marie Ruel and Bénédicte de la Brière. “Mothers’ Work and Child Care: Findings from the Urban Slums of Guatemala City.” Economic Development and Cultural Change Vol. 53, No. 4 (July 2005), pp. 855-885. 51 What Works for Working Women? Heintz, J., N. Kabeer and S. Mahmud. 2018. “Cultural norms, economic incentives and women’s labour market behaviour: empirical insights from Bangladesh.” Oxford Development Studies, 46:2, 266-289, Hill, R. and M. E. Genoni. 2018. Bangladesh Poverty Trends 2010/11–2016/17. Washington, DC: World Bank. Hill, R. and Z. Rahman. Forthcoming. Poverty in Urban Bangladesh. Background Paper prepared for the Bangladesh Poverty Assessment, World Bank Group. April 2018. Hossain, Julaikha B. and Kyoko Kusakabe. 2005. “Sex segregation in construction organizations in Bangladesh and Thailand.” Construction Management and Economics, 23:6, 609- 619. IMF. 2018. Bangladesh: Staff Report for the 2018 Article IV Consultation. IMF Country Report No. 18/158. JICA (Japan International Cooperation Agency). 2017. Urban Transport Survey. Jolliffe, Dean, Iffath Sharif, Lea Gimenez, and Faizuddin Ahmed. 2013. Bangladesh Poverty Assessment: Assessing a Decade of Progress in Reducing Poverty, 2000-2010 (English). Bangladesh Development Series; Paper No. 31. Washington, DC: World Bank. Kabeer, Naila and Luisa Natali. 2013. Gender Equality and Economic Growth: Is there a Win-Win? IDS Working Paper 417. Institute of Development Studies, February 2013. Kabeer, Naila. 2013. Paid work, women’s empowerment and inclusive growth. UN Women Policy Paper. UN Women. Klasen, S., and J. Pieters. 2015. “What Explains the Stagnation of Female Labor Force Participation in Urban India?” The World Bank Economic Review 29 (3): 449–78. Mahmud, S. and S. H. Bidisha. 2018. “Female Labor Market Participation in Bangladesh: Structural Changes and Determinants of Labor Supply.” In S. Raihan (ed.), Structural Change and Dynamics of Labor Markets in Bangladesh (pp. 51-64). Singapore: Springer. Naila Kabeer. 2017. “Economic Pathways to Women’s Empowerment and Active Citizenship: What Does The Evidence from Bangladesh Tell Us?” The Journal of Development Studies, 53 (5): 649-63. Rahman, Rushidan I; Islam, Rizwanul. 2013. Female labour force participation in Bangladesh: trends, drivers and barriers. ILO Asia-Pacific working paper series. International Labour Organization. Ruel, M. T., B. de la Briere, K. Hallman, A. Quisumbing, and N. Coj. 2002. “Does Subsidized Childcare Help Poor Working Women in Urban Areas? An Evaluation of a Government-Sponsored Program in Guatemala City.” FCND Discussion Paper 131, IFPRI: Washington, DC. Salway, Sarah, Sonia Jesmin, and Shahana Rahman. 2005. “Women’s Employment in Urban Bangladesh: A Challenge to Gender Identity?” Development and Change 36 (2): 317-49. Solotaroff, Jennifer L., Aphichoke Kotikula, Tara Lonnberg, Snigdha Ali, Rohini P. Pande, and Ferdous Jahan. 2019. Voices to Choices: Bangladesh’s Journey in Women’s Economic Empowerment. International Development in Focus. Washington, DC: World Bank. UN. 2018. 2018 Revision of World Urbanization Prospects.  Population Division of the United Nations Department of Economic and Social Affairs (UN DESA). UN-HABITAT. 2011. Building Urban Safety through Slum Upgrading. United Nations Human Settlements (UN- HABITAT). World Bank. 2011. World Development Report 2012: Gender Equality and Development. World Bank ___ . 2019. World Development Indicators. Washington, DC: World Bank. ___. 2018. South Asia Economic Focus, Spring 2018: Jobless Growth? Washington, DC: World Bank. 52