Research Data Center of IZA (IDSC)
Not a member yet
105 research outputs found
Sort by
G²LM|LIC - Female Entrepreneurship and Professional Networks
No full textDespite being the only region in the world where there are more female entrepreneurs than men, the vast majority of female-owned businesses in Sub-Saharan Africa are microenterprises and women’s businesses earn 34% lower profits than male-owned ones. Identifying the constraints faced by female entrepreneurs is vital for fostering economic growth. Recent studies have shown that interfirm relationships and access to professional networks can be important determinants of business success (Ashraf et al., 2019; Kanter, 1994; Cai and Szeidl, 2017). However, past interventions that aimed at expanding business networks mainly considered male entrepreneurs (Cai and Szeidl, 2017). At the same time, the literature on female entrepreneurs in the developing settings has primarily focused on informal microenterprises. Therefore, there is limited evidence on the importance of interfirm relationships for female entrepreneurs of potential high-growth firms.
In this project, we conduct an RCT in Ghana on a sample of 1,771 growth-oriented female entrepreneurs to investigate the effect of online networking groups combined with legal support on business collaborations and firm performance. We randomly assigned the female entrepreneurs into two treatment arms and a control group:
Treatment 1: Networking 40% (N=704)
Treatment 2: Networking and Legal Support 35% (N=608)
Control 25% (N=436)
In the first treatment arm, women are assigned into online networking groups of 8 entrepreneurs on the WhatsApp platform in two rounds. Each week, each member is assigned to meet virtually with another group member. We also provide a directory of all entrepreneurs in the treatment group with their contact information. The aim of this treatment is to expand the business networks of participants and increase their opportunities for business collaborations. In the second treatment arm, entrepreneurs also receive legal support in addition to the online networking groups. The legal support entails weekly video lessons by a local corporate lawyer that discusses risks of collaborations and ways of mitigating these risks through the use of written agreements and effective communication. Entrepreneurs can also consult the lawyer individually during the four-month intervention period. This treatment aims to reduce contracting frictions in addition to networking frictions, potentially increasing collaborations between entrepreneurs who meet on the platform.
The intervention was implemented between February and June of 2021. The first post-intervention follow-up survey was conducted between August and October 2021 and the second follow-up survey was conducted between April to July 2022, one year after the intervention.
From the analysis of the one-year follow-up survey, we find that access to online networking opportunities (i) does not increase total business collaborations, but (ii) shifts business collaborations from friends and family members to business network members in the intervention, and (iii) leads to greater innovation, better business practices and higher profits by 21%. The increase in profits is concentrated in the upper tail of the distribution. We find the largest effects for those in groups with more-educated, higher-quality, and more diverse entrepreneurs. While legal support leads to a decline in overall collaborations due to a greater reduction in collaborations within pre-existing networks (e.g., friends and family), it does not have additional impacts on business outcomes, suggesting that legal contracting barriers are unlikely to be the key barrier to growth for these firms. Our findings reveal that a low-cost, light-touch online intervention that increases networking opportunities can effectively improve outcomes of female-owned firms.
However, the current dataset does not allow us to disentangle the main drivers of our results which is crucial for understanding the policy implications of this intervention and its scale-up potential (even in different settings). For this reason, in March 2023, we conducted 17 in-person qualitative interviews to collect preliminary evidence on mechanisms. The qualitative interviews reveal that the WhatsApp networking groups allowed women in our sample to find new business partners, market their products, learn better business practices, and gain new ideas.
In this project we conduct a 3-year post-intervention quantitative survey on our entire sample of female entrepreneurs. The goal of this long-run data collection is to test potential mechanisms highlighted by the qualitative interviews and to understand the persistence of our results beyond the first year post-intervention. Testing these mechanisms in the long run is important because outcomes such as formation of collaborations and women’s empowerment may require multiple years to realize
G²LM|LIC - Whistleblowing and Worker Wellbeing: Evidence from Bangladesh’s Garments Sector
No full textIn many developing countries, the private sector lacks monitoring systems to provide firms with incentives for good behavior. In part, this problem is due to weak, sometimes corrupt state institutions (Dal Bó and Finan, 2016). In part, it may also be due to principal-agent problems within the firm and to limited organizational capacity (Bloom et al., 2014; Boudreau, 2019). In principle, external whistleblowing systems (e.g., implemented by regulatory agencies) could support employees to inform state or other entities about employer misconduct. But while theoretical literatures on principal-agent-monitor problems and on secure survey design generate predictions on how the design and implementation of whistleblowing systems affect information transmission and misbehavior (Chassang and Padró i Miquel, 2018; Chassang and Zehnder, 2019), Little is known about how these predictions perform in practice.
In the proposed research, the research team study how the design and the introduction of a whistleblowing system affects information transmission by employees, misconduct by firm owners or by managers, and ultimately workers’ wellbeing and relations with management. Our setting is Bangladesh’s garments sector, where weak state institutions offer little to no legal protections to whistleblowers. In response, multinational apparel buyers introduced their own whistleblowing system in the form of an anonymous, toll-free helpline managed by a reliable third-party. They collaborate with the helpline, named Amader Kotha (AK), or “our voice” in Bangla, to implement a field experiment. Theytest how varying the resolution protocol affects workers’ incentives to report, the actual incidence of labor issues, and workers’ wellbeing and relations with management. They hypothesize that lack of plausible deniability and coordination problems lead employees to underreport certain types of employer misbehavior. Further, the study hypothesizes that women, who comprise the majority of workers in this setting, are both more subject to mistreatment by their largely male managers and face higher costs of reporting.
The project team will conduct a field experiment with 158 garments factories that participate in the AK Helpline. They will randomly assign half of these factories to treatment and half to control and compare their outcomesat baseline and 9 months after the intervention. Factories that are treated will experience a change in the AK Helpline’s resolution system, namely, an increase plausible deniability for callers and/or a reduction in coordination problems among workers by providing an information escrow2. The control condition is the status quo AK Helpline resolution system. They will test for effects on four main outcomes: (1) worker incentives to report sensitive issues, (2) actual occurrence of labor issues, (3) worker wellbeing, and (4) worker-manager relations. In addition, The team will explore potential effects on collective action by workers3. They will test for heterogeneous treatment effects on these outcomes by gender. To measure these outcomes, they will use a combination of individual survey data, helpline call data, as well as an anonymous voting system to get at the actual incidence of employer misbehavior.
Our research contributes to three main strands of literature: A growing literature on labor standards and economic development, and in particular, their interaction with global trade; an extensive theoretical literature on contract theory and collusion in organizations, and specifically in relation to the design of whistleblowing mechanisms; and to literature advancing the design of survey instruments to elicit sensitive behaviors. This project provides, to our knowledge, with the first field-based experimental evidence on the design of whistleblowing mechanisms under fear of retaliation. The project team is also the first to study the incidence of harassment at the workplace in a randomized controlled trial.
This research is highly policy-relevant. There is a great deal of interest among policymakers and multinational buyers in how to design whistleblowing and grievance resolution systems to provide employers and their managers with incentives for good behavior. This research also responds to interest in gender equality, highlighting the role of internal reporting and grievance mechanism to improve conditions for women at the workplace. This research is part of a research agenda on labor conditions and productivity in developing countries (Boudreau et al, 2019, Boudreau 2019) and information mechanisms in affecting development outcomes (González-Torres, 2019). Our research team’s ability to partner with critical stakeholders in the global apparel supply chain, such as the Alliance for Bangladesh Worker Safety (a coalition of multinational buyers that formed in the aftermath of the 2013 Rana Plaza collapse in Bangladesh to improve their suppliers’ safety), is possible, demonstrates stakeholders’ interest in this research agenda.</br
G²LM|LIC - Together to Work? Role of Travel Buddies on Women’s Employment and Mobility
No full textIntroduction to the Dataset
This dataset is derived from a clustered randomized controlled trial (RCT) conducted in urban neighborhoods across the Delhi NCR region in India. The experiment evaluated whether enabling women to coordinate travel with peers for job interviews would reduce mobility-related constraints and increase employment uptake. The study design, survey instruments, and sampling methodology were developed in collaboration with local partners and refined over multiple pilot rounds.
Experimental Design
The unit of randomization was the neighborhood, defined based on geographic and infrastructural characteristics. The sampling frame was limited to areas located within a 12-kilometer radius of a sample garment factory. A detailed mapping exercise was carried out to identify and define neighborhoods using prominent physical features such as main roads, highways, parks, and open areas. In instances where natural boundaries were insufficient, paved roads or non-residential buildings were used to delineate cluster boundaries. To minimize information spillovers, buffer zones were created between neighborhoods, and clusters in close proximity were excluded if spillover risk could not be mitigated.
A total of 106 neighborhoods were finalized for inclusion. For each neighborhood, we computed the distance from its geographic centroid to the factory and binned neighborhoods into nine distance-based strata. These bins showed a high correlation with travel costs to the factory, measured using prevailing rates for shared and private auto-rickshaws. Within each city-distance stratum, neighborhoods were randomly assigned to one of three groups: (i) Matching and Coordinated Travel, (ii) Matching without Coordination, and (iii) Control.
Sample Recruitment and sampling Procedures
Following randomization, household screening began in February 2024. Within each cluster, enumerators selected a random entry point to begin door-to-door screening. They adhered to a ”right-hand rule” to navigate lanes and avoid bias toward households located near the main access roads. The screening questionnaire assessed eligibility based on the following six criteria:
Age between 18 and 40 years (initially capped at 35, later relaxed to 40)
Possession of a valid government-issued ID (Aadhaar)
Ability to operate a home or factory sewing machine
Not currently engaged in paid work outside the home
Not employed by the partner factory (Shahi Exports) within the last 3 months
Expressed interest in working at the factory, regardless of household approval
These criteria were designed to align with the factory’s hiring requirements while focusing on women who may face barriers to labor force entry. A total of 693 women meeting these criteria were enrolled across 106 neighborhoods, with an average of 7 women per neighborhood.
The original enrollment target was 750 women (10 per neighborhood), but the target
was revised after it became evident that some clusters had fewer eligible women. To ensure statistical power, we increased the number of clusters from 75 to 106 and recalculated sample requirements accordingly.
Survey Instruments and Data Collection
Data collection occurred over three major phases—two pilots and one main study—and included four core instruments: baseline survey, neighborhood meeting attendance survey, factory-site survey, and endline survey.
Baseline Survey: Conducted between March 27 and June 19, 2024, the baseline survey was administered in-person at the respondent’s home immediately following enrollment. Trained female enumerators administered structured questionnaires using tablets. Modules included demographics, employment history, safety perceptions during commuting, mobility and trip histories, aspirations, gender attitudes, and social connections with other women in the neighborhood. We also recorded detailed information about household composition and
resource control.
Meeting Attendance Survey: This survey was administered to women assigned to the two treatment groups, particularly those invited to participate in travel coordination meetings. Early implementation revealed low attendance at these meetings. In response, field protocols were adjusted to include home visits and escorting participants to meetings in smaller groups, located closer to their residences. The meeting survey collected information on attendance, group composition, preferences for travel companions, and perceived utility of the intervention.
Factory-Site Survey: Enumerators were stationed at partner factory gates during scheduled interview days. For all study participants who appeared, we administered a short survey that recorded their mode of travel, companions, time of departure, cost of travel, and reasons for attending. This data forms the basis for the primary outcome: interview attendance.
Endline Survey: Conducted between May 31 and August 11, 2024, the endline survey covered all enrolled women approximately 5–6 weeks after the interview invitation window closed. Many respondents had relocated or were temporarily unavailable, requiring multiple follow-up visits and a shift to phone surveys in several cases. Phone surveys employed a shortened version of the endline instrument, focusing on key outcomes such as employment status, travel behavior, and mobility confidence. Attrition between baseline and endline was 19.8%, resulting in an endline sample of 560 women.
Pilots
Pilot 1: Conducted starting November 26, 2023, Pilot 1 served to test the feasibility of the screening and enrollment protocol and covered 31 respondents. The team assessed the clarity of survey questions and the operational logistics of mapping and randomization.
Pilot 2: Launched on February 6, 2024, Pilot 2 involved 30 women and incorporated revisions from Pilot 1. Updates included streamlining the survey instrument, clarifying consent procedures, and refining respondent tracking protocols.
The endline surveys for these 61 pilot participants were conducted concurrently with the main study endline in June–August 2024, using the same survey instruments and protocols to ensure comparability.
Dataset Composition and Documentation
The final dataset includes three sets of anonymized files corresponding to the Pilot 1, Pilot 2, and Main Study phases. Each phase includes:
Individual-level baseline survey data
Meeting attendance survey (for treatment groups)
Factory-site tracking survey
Endline survey data
All data are provided in Stata (.dta) format with clearly labeled variables. Additional documentation includes:
Survey instruments (baseline, endline, factory site, and meeting)
Variable-question mapping
</ul
G²LM|LIC - Depression Treatment and Female Performance in the Labor and Marriage Markets in India
No full textThis study investigates the impact of depression treatment on women’s outcomes in both the labor market and the marriage market in India. Depression is a prevalent barrier to women’s productivity, household bargaining power, and subjective well-being. The intervention provides structured depression treatment and examines its causal effects on women’s economic participation, marriage attitudes, and interpersonal relations.
The dataset includes measures of mental health, cognition, risk attitudes, household consumption, intimate partner violence, and subjective well-being. Productivity is captured through experimental tasks such as the lentil sorting exercise.
Confidentiality / Anonymization:
All data have been anonymized according to Harvard Dataverse and G²LM|LIC standards. Sensitive variables (e.g. intimate partner violence) are included only in de-identified form and require controlled access.
Methods:
Randomized Controlled Trial (RCT) of depression treatment
Surveys on labor market, marriage, and household outcomes
Cognitive and risk-preference tests
Productivity measures (lentil sorting exercise)</li
G²LM|LIC - Female Wage Labor and Fertility - Evidence from the Cut-Flower Industry in Kenya
No full textThis study examines how the emergence and expansion of the cut-flower export industry in Kenya—particularly rose-growing greenhouses employing predominantly low-skilled women—affects female labour market participation and fertility outcomes. The project constructs a new geospatial database documenting the timing, location, and growth of flower-processing plants from 1999–2022 using historic satellite imagery, phone verification, and customs records. These data are linked to multiple rounds of household surveys, DHS birth histories, and population censuses to estimate how new wage opportunities for women shape employment patterns and fertility decisions. The empirical strategy follows a staggered difference-in-differences design comparing treated villages to geographically suitable control locations. The resulting database provides the first systematic mapping of Kenya’s rose-greenhouse expansion and supports research on structural transformation, gender, and demographic change in sub-Saharan Africa.
Data Description
The dataset contains a time series documenting the expansion of the Kenyan cut-flower industry at the county level. For each year from 1999 to 2022, it records the number of identifiable greenhouses and total greenhouse area based on satellite imagery and auxiliary sources. The dataset captures greenhouse-based export production, predominantly roses. Open-field production (e.g. summer flowers) is not systematically measurable from satellite imagery and is therefore outside the main scope.
Scope Note
This dataset records the expansion of greenhouse-based rose farms only. Summer-flower production or non-greenhouse fields cannot be reliably detected and are not included. The time series begins in 1999 because earlier satellite images are of insufficient quality to identify structures consistently.
Variables
county — county name
year — calendar year
gh_numn_greenhouses — number of flower growing greenhouses
gh_area — total greenhouse area</br
G²LM|LIC - Commuting Constraints and Labor Productivity: A Field Experiment on Women’s Mobility in India
No full textThis dataset is part of the project Commuting Constraints and Labor Productivity: A Field Experiment on Women’s Mobility in India, which investigates how transportation access shapes women’s work productivity and career progression in gig work.
The study is designed as a randomized controlled trial (RCT). Female gig workers were randomly assigned to different commuting support conditions to measure the causal impact of mobility constraints and safety considerations on labor supply, earnings, and job choices.
Data collection consists of:
Baseline survey: Demographics, household composition, transportation patterns, financial status, motivations for joining gig work.
Follow-up surveys (two rounds): Daily job activity, commuting modes, safety perceptions, time use, well-being, and household updates.
Administrative platform data: Earnings, job assignments, customer ratings, cancellations, and platform logs (not publicly available due to NDA).
The dataset allows for longitudinal and experimental analysis of how commuting constraints and safety considerations affect women’s participation and productivity in the gig economy.
Access and Embargo
The dataset is under embargo for three years to allow the research team to complete analysis and publication. A de-identified version will be made available upon release, in line with ethical and confidentiality standards
G²LM|LIC - Overcoming Constraints to Female Labor Force Entry
No full textWe conducted a baseline survey with 2,499 female final year undergraduate students between October 2018 and February 2019.1 Of them, 1,224 (49%) were randomly assigned to the treatment group. The intervention was reinforced between February-May 2019 (intervention reinforcement). The respondents were interviewed again between, August-September 2019 (follow-up 1 ), December-January 2020 (followup 2 ) and then finally between May to June 2020 (follow-up 3 ).
Experiment
We conducted a randomized experiment with an expected sample of 2,500 female students in the final year of their undergraduate degree. We exclusively focussed on students with liberal arts majors, across 28 public colleges in urban areas of Lahore, Pakistan. We collected baseline data through face-to-face interviews, carried out by a team of experienced enumerators, in the students’ respective college.
Intervention
Our intervention consists of a documentary video on actual educated women from public colleges in Lahore who have been successful in the labor market, that is, have secured a job and are satisfied in their current jobs. The documentary is intended to emphasize that setbacks are an opportunity to learn; that the process of learning is enjoyable in itself; and that economic empowerment can help both their standing in the household and household welfare. Students in the placebo group were exposed to a documentary on a subject unrelated to the treatment. In addition to the treatment or placebo videos documentary, all students in our sample were given information about ‘Job Talash’; an existing job search portal that connects job seekers with employers in metropolitan Lahore. That is, all students in our sample had access to information on existing jobs in Lahore. All randomisation - providing the treatment or the placebo intervention, as well as the order in which information on Job Asaan or the documentary was shown was randomised at the individual (student) level.
Sample and treatment assignment
Sample selection and treatment assignment at the individual level was done as follows:
Step 1: We requested the college administration for a list of students enrolled in the final year of the bachelors program.
Step 2: We identified the proportion of the total working sample to be drawn from each college (on the basis of enrolment data from step 1).
Step 3: We randomly select 70% of this sample to be the actual sample and keep 30% as a replacement sample to be contacted if a sample student is not located or if she refuses to participate in the survey.
Step 4: We collect all data on tablets using SurveyCTO (www.surveycto.com). At the time of the data collection, our software assigns the student to either the treatment or placebo group, with equal probability. </li
G²LM|LIC - Can Temporary Financial Incentives for Female Industrial Workers Lead to Long-Term Retention and a Better Allocation of Talent?
No full textA G²LM|LIC Research Project conducted by the University of Oxford.
In this project, the research team studies whether worker turnover contributes to the misallocation of talent in low-income countries. To this end, they will experimentally evaluate the impacts of offering financial incentives for worker retention in the context of a female-dominated occupation in the nascent garment manufacturing industry in Ethiopia.
Our main hypothesis is that many of the young female workers who quit the position have not spent enough time on the job to learn the true quality of the match, which, for some of them, may be high. Standard economic theory argues that worker turnover is an essential part of the process through which economies achieve an efficient allocation of talent: new hires are uncertain about the quality of the match with their employer and they quit their job when they become convinced that it is a poor match (Jovanovics, 1979). However, there are several reasons to expect that quitting decisions may be made before the quality of the match has been fully revealed. First, workers do not internalize the costs that high levels of turnover impose on the firm: more resources have to be devoted to recruitment, planning production and meeting deadlines becomes more difficult, and the investment in training the worker is lost. In markets where access to informal work opportunities is relatively easy, this can create an asymmetry between the cost of turnover to the worker and the cost of turnover to the firm. Second, early signals about the quality of the match may be misleading. Workers may eventually adjust to many features of the work environment that are initially perceived as disamenities, but often underestimate the extent of this adjustment (a phenomenon called “projection bias”, e.g., see Loewenstein et al., 2003). Workers that leave their job early – which represent the bulk of the turnover problem faced by firms (Jovanovics, 1979; Donovan et al., 2019) – may thus be quitting sooner than optimal. This can be particularly detrimental for young, female workers that have recently migrated and are only weakly attached to the labor market and thus have noisy priors about match quality in different jobs.
The team plans to make three contributions to the literature. First, they will contribute to the literature on the allocation of talent by testing whether labor turnover – a process that is typically associated with improved allocative efficiency – may be partly driven by suboptimal decisions that destroy good matches. This is particularly important in the context of the allocation of talent of young female workers from low-income backgrounds, who often face the largest barriers to finding well-matched positions (Hsieh et al., 2019). Second, they will contribute to the literature on worker retention in developing countries, which has recently documented that turnover rates are higher than in richer economies (Donovan et al., 2019), especially in blue-collar occupations (Blattman and Dercon, 2018). Finally, they will contribute to the literature on structural behavioral economics (Della Vigna, 2019), by experimentally identifying and quantifying in the field a bias in belief formation that is currently supported by evidence from the lab and by observational studies (Loewenstein et al., 2003).
The research team works in a particularly interesting context: the Hawassa industrial park, the flagship economic project of the current government, where a number of foreign investors have recently started producing garment-manufacturing products. This is a new activity in this area, and most workers are new to the industry. Further, standard factory-floor positions, which make up the bulk of the workforce of the firms in the park, are virtually only taken up by young women. They have identified a firm in the park that is willing to randomize certain features of their payment scheme. They have already collected some data on the workers of the firm and have run a small pilot.
In the experiment, the team will randomize the timing of a retention bonus offered to workers. The first group of workers will be offered a bonus that is paid after three months on the job, which incentivizes them to complete the initial training (which lasts for 2 months). A second group will be offered a bonus that is paid after seven months on the job, which incentivizes workers to spend a significant amount of time on the production line. Our key empirical test will be whether the late bonus generates higher retention in the long term (i.e., after the seventh month of tenure) compared to the early bonus. They will collect detailed belief data to support our interpretation and to rule out alternative explanations (e.g., search costs that rise with tenure or gains in job-specific skills). Second, they will study whether workers in the late retention bonus group have higher levels of earnings and work satisfaction, one year after the offer of the bonus. This will test the misallocation hypothesis: turnover destroys jobs that are actually a good match for the worker. Third, they will study how the late bonus intervention changes the propensity to leave of workers at different levels of the productivity distribution.
These results will be of interest to firms investing in high-turnover environments and to the governments of Ethiopia and other low-income countries. The research team has developed a detailed impact plan to make sure results will be disseminated effectively and will influence policy decisions. They will also make significant investments in capacity building
Replication Data for: Mincerian Returns to Education
No full textThis database combines two large datasets. Psacharopoulos and Patrinos (2018) includes 1,120 estimates from 139 countries. Montenegro and Patrinos (2023) presents a comparable set of returns to education, with estimates for 142 economies from 1970 to 2014, based on 853 harmonized household surveys. The new combined dataset includes estimates of the returns to schooling from 1,973 surveys in 165 countries. This compilation provides estimates of private rates of return to education for 97 percent of the world’s population
Longitudinal Surveys of Australian Youth, 2015 cohort (Version 7.0)
No full textIn 2015 a nationally representative sample of about 14,500 15 year-old students was selected to participate in the OECD’s Programme for International Student Assessment (PISA). This group of young people became the sixth cohort of the LSAY program (LSAY Y15). As part of PISA, assessments in mathematical literacy, reading literacy and scientific literacy were administered to schools to provide information on student achievement. Students also completed a background questionnaire about themselves, their families, their views on a range of science-related issues, the environment, educational and vocational expectations, attitudes to school and learning, attitudes towards information and communications technology, and their habits and life in and outside of school. In 2016, members of the Y15 cohort were contacted for the first of their annual LSAY interviews. The questionnaire for their 2016 interviews included questions about their activities during the last 12 months, school, post-school study, engagement in education, career aspirations and decision-making, employment, job satisfaction, job search experiences, income, soft skills, personality, and wellbeing. From 2016, respondents were given the option to complete their interviews by telephone or online. Response rates for the 2016 interviews were lower than expected due to a high rate of missing or unusable contact details provided at the time of PISA. To address this, recruitment of a top-up sample of Year 12 students was conducted in 2017 to ensure that future waves of the survey remain representative of the 15-year-old population in 2015. See Notes (in the Social Science and Humanities section of Metadata) for version details about this dataset