Creating jobs, developing skills
A permanent job with a decent salary is an effective way out of poverty. How to boost growth in the labour market should thus be an essential question for policy makers.
Existing research and knowledge can help policy makers in choosing the best way forward. In close cooperation with our partners in Ethiopia, Nepal and Tanzania we have created this platform to highlight important policy challenges and show how research best can be utilised to solve them.
The network focuses on three types of policy issues:
Creation of new jobs
- What is the government and donor role in job creation?
- How can a country secure a balanced transition out of agriculture?
- Balanced growth in both urban and rural areas
- Inclusive growth with permanent jobs also for low-skilled workers
- Impact of labor market programs
- How getting a job changes a life
Well functioning labor markets
- Access to jobs for the poor and marginalized, including women
- Minimize the economic and social costs of labor migration
Improved competence for workers
- How can the productivity of subsistence farmers be improved?
- How can the incomes of labor migrants be improved?
- The effect of different types of skill training and supplemental programs
Active private sector development policies revisited: Impacts of the Ethiopian industrial cluster policy
Active private sector development policies revisited: Impacts of the Ethiopian industrial cluster policy
Human and financial capital for microenterprise development: Evidence from a field and lab experiment
There is an active research field into jobs in developing countries. Below are summaries and links to several interesting papers by researchers from other institutions that might be of interest.
Returns to On-the-job Soft Skills Training (2021)
RCT on effect of soft skills on productivity, type of work performed and retention. Soft skills training (80 hours) over 12 months, components on communication, time management communication, problem solving. (and control)
No effects on retention and attendance. Large (20% increase) effect on productivity and complexity of work assigned (7%). Wages constant
Suggested to be caused by increase in stock of soft skills: those with low leadership skills or low education improved most. Technical operators also saw big improvements.
Empirical evidence for soft skills stock increase: increase in saving and risk sharing (planning skills from training), extraversion and enrollment in govt programs.
Some effects might be because of increased social capital, but unlikely to be full effect as training was with very few direct coworkers.
Return on investment was 70% after 8 months and 260% after 20.
That productivity went up but wages remained constant shows that there is friction in the labor market: they could not leverage their increased marginal product to get higher wages/a new job. Therefore, there is scope for profit gains from soft skills training
Business training evaluations
What Are We Learning from Business Training and Entrepreneurship Evaluations around the Developing World? (2014)
The world bank research observer
Review of several business training papers. Mostly focused on really small businesses (microenterprises). There are several issues with existing papers:
- Takeup is relatively low, even with prior interest, N is usually low and there is high variability between firms. This all leads to very low power.
- Timing is on the short run (less than a year)
- Attrition might be selective. Unsuccesful firms learn to quit. Succesful firms move out. Firms unimpressed with training might refuse to participate
- Often survey attrition on profits (e.g. by large firms that don't want to disclose this). Training might also improve profit reporting or lead to Hawthorne effects. Profit generally hard to measure because of bad record-keeping and non-separation of private and business finances.
Impacts generally very low and non-sig (because of power) on key outcomes of survival, profits and employment generation (some effects on practices, but a bit mechanical).
- Heterogeneous effects: who does it work for most? (measure with stratification or look at correlates.)
- Spillovers/general equilibrium effects (measure with treatment intensity)
- Long-term effects
- Market failures: why do firms not buy training? Could beInformation failure, credit constraints or no supply.
Active Labor Market Policies
How Effective Are Active Labor Market Policies in Developing Countries? A Critical Review of Recent Evidence (2017)
The World Bank Research Observer
There are several types of Active Labor Market policies employed by developing countries:
1. Vocational training: increase skills of workers and make them more employable. Low effectiveness, but unsurprising if replacement of formal schooling. Results are similar when scaled for (relatively short) length of 3 months
2. Wage Subsidies. Overcome market failures of hiring/firing frictions if workers have unknown skills, and/or allow inexperienced workers to build experience. If it overcomes these frictions it does not necessarily displace unsubsidized workers. In reality, workers are fired immediately when the subsidy ends. Might be useful in times of crisis or to cause temporary employment.
3. Search and Matching assistance. Might have use if formal schooling is not a good signal of quality, and/or it might not teach the skills employers need. Rarely find an effect on employment. In fact, only rarely do they actually lead to matches.
Perhaps workers are not interested in the jobs offered: people randomized into industrial or textile jobs quit at a high rate. Furthermore, firms are not willing to raise wages or spend more time searching: it appears that vacancies can be easily filled by them.
Managers and Work Quality
Managerial Quality and Productivity Dynamics (2021)
Paper examines correlates of managerial characteristics with productivity from highly detailed production data. And develops and simulates a model to examine effect of (achieveable) shocks to these characteristics.
Using Exploratory Factor Analysis they determine what matters for productivity, leads to seven main characteristics: 1. Tenure (industry-specific) 2. Demographics (similarity to workers and discrimination attitudes) 3. Cognitive skills (arithmetic, memory) 4. Control (expected, over life) 5. Personality (perseverance, conscientiousness etc) 6. Autonomy (taking initiative) 7. Attention (effort expended to reach targets, hard-working)
Of the big five, only conscientiousness matters (found in pilot), like other studies have found
Examine relation of these chars with efficiency/productivity on a production line over time. Two main ways: higher initial productivity or stronger learning over time.
Strongest impacts for initial productivity: control, attention and tenure. For learning: Attention, Tenure and autonomy.
Training can improve both autonomy and attention, with attention having strongest effects. An increase in autonomy will likely also increase wages, as these can be observed more easily and can thus be used to negotiate for higher wages.
Low Productivity in Developing Countries
Why Do Firms in Developing Countries Have Low Productivity? (2010)
American Economic Review: Papers & Proceedings
Short paper that examines some causes of low productivity of firms in developing countries.
1. Management Practices. Knowledge of these practices is very low (information constraints). Quality was not checked.
2. Delegation of decision making. Owners fear that managers will steal from them if given spending powers (caused by weak rule of law and bad inventory systems). This means that owner must take all decisions, which becomes unfeasible at large firm sizes and places an upper limit on firm size. This also slows down problem-fixing within the firm (and reduces productivity)
3. Financial Constraints. Many firms report low access. Loans for management training might not be given because as better organizational practices cannot be used as collateral. Furthermore, if access to finance is bad new entrants with good management practices will not be able to join the market to compete with existing badly managed firms.
Improved Management and Productivity
Does management matter? Evidence from India (2013)
The Quarterly Journal of Economics
Paper examines the impact of management practices on large textile firms in India.
Treatment with three phases:
1. Diagnostic phase. Evaluated current practices, started tracking database and gave recommendations (1 month)
2. Implementation phase. Regular visits by consultants to help and guide them implement the recommendations (5 months)
3. Measurement. Collected detailed data using the database system set up. (throughout)
2 was randomly assigned, control also received 1 and 3. Therefore control also had some improvements (which they show some evidence for). Low N (11 treat 6 control), but high T helped with significance. Examined training through 38 practices.
Adoption of new practices took time but was effective: increase of 38%. Non-treated treatment plants (e.g. plant within the same firm) also increased by 18% (within-firm spillovers). Control increased by 12%. Big reduction of defects and (useless) inventory. TFP increased by 17%. Long-run effects show increased number of plants opened. Labor was constant: less for fixes (through quality improvements) but more output.
Why do these firms persist? Many practices were unknown, and if known management had misperceptions on how effective they would be. Also, the owner sometimes had little time to implement them. Better managed firms did not compete these out of the market because barriers to entry are high and better managed firms cannot increase even further because the owners do not have enough control and are unwilling to delegate
Differences in Management
What Drives Differences in Management Practices? (2019)
American Economic Review
Paper examines the correlation between management practices and firm succes, and also what drives differences between these practices.
Uses panel on 35,000 firms in the US (2010 and 2015) and links this with other datasets. Asks questions on 16 management practices which are scored from 0-1. This is aggregated into one Structured Management score (range 0-1). Included things like continuous monitoring, evaluation and improvement (from lean-style manufacturing), but also decentralization. Dispersion of this measure is huge. Also, they find that almost 50% of variation comes from measurement error.
If management is seen as a type of technology modifier in a Cobb-Douglas style production function it strongly correlates with productivity (log(output/employment)). It explains 20% of the 90-10 percentile dispersion in productivity, similar to R&D and more than ICT and high-skill workers. It also leads to lower rates of exit and higher employment growth. The effect for exit is more pronounced for younger firms. Consistent with standard market selection models: Firms with weak management capability are rapidly competed out of the market by better managed firms.
60% of variation in practices is between firms. So 40% is caused by differences in practices between plants *within* a firm (after correcting for measurement error).
To examine what drives changes use two approaches.
1. Right-to-Work programs. These made union membership non-obligatory and allowed firms to be more free in their payment practices (e.g. adding incentive schemes). This increases incentive management, but also employment (not TFP). DiD and RD show similar results. Lack of TFP results could be because of 'congestion effects' driving up local labor and land prices.
2. Million dollar plants. These are large, multinational plants (with likely effective management strategies) that write a tender for a new location. Compared 'winners' with runner-ups. The opening of an MDP plant increases management scores and employment. If manager flow is high it also increases TFP.
This shows that the business environment (1) and learning (2) drive changes in management practices. Implies that informational constraints play a role in non-universal adoption of these (seemingly effective) management practices. However, this does not explain the large intra-firm variation in practices.
Trainings for small businesses
Small Business Training to Improve Management Practices in Developing Countries: Reassessing the Evidence for "Training Doesn't Work" (2020)
World Bank Policy Research Working Paper
Paper provides meta-analysis of existing Micro/Small businesses training programs. And, examines promising new approaches to standard training programs.
SME training is politically popular, but initial evidence was not very positive. Furthermore, training programs might redistribute sales among business, not necessarily increasing total revenues.
Existing programs usually taught record-keeping, accounting, marketing, human resources, hiring workers, stock control, inventory management and operations management. Usually classroom based with costs mostly determined by trainer, venue and transportation costs. Length is usually short, so when benchmarked against returns to schooling a 5-day training should not lead to much more than a 5% increase in profits.
Businesses generally adopt few of the taught practices: 1 in 20. Random effects meta-analysis model of standard trainings finds a 10% increase in profits, and 5% increase in sales.
1. Gender-based. Focusing on women. Uncertain if the gender component adds value
2. Kaizen Continuous improvement. Some potential, but little evidence yet.
3. Local customization, peers or mentors. Does not appear very effective, but might be for businesses expanding into new markets.
4. Simpler training. Might be useful for smaller firms/owners with lower educational background. Evidence not very convincing (short run, small effect sizes)
5. Personal initiative (soft skills). Teaching owners to having higher autonomy. Especially effective when taught by business owners.
Scaling this up can be done through creating a market (and ensuring access for smaller firms through vouchers). Business training is an experience good. Alternatively: use technology. TV shows, SMS based coaching. Appears not very effective so far. Or: funnel firms. Give everyone a general training but go in-depth for those that appear more receptive.
Larger samples allow better insight into what characterizes firms which might be better recipients
Long-run effects of management training
Do Management Interventions Last? Evidence from India (2020)
American Economic Journal: Applied Economics
Paper looks at the long-term (10 year) impact of the Bloom et al (2013) paper. Revisit the same sample for detailed measure of management practices, and less so for productivity.
Find that management practices dropped over time for treated plants. Still remained substantially above control. But also important spillovers to other plants within the same firm (untreated). These eventually matched the treatment. So important knowledge spillovers. This also held for control firms that had no treatment (remember, control firms did get monitoring and a small intervention.) Practices were dropped because of new managers, little time of managers or because they didn't work. See long-term productivity improvements, mostly because it allowed them to update to higher-producing looms. This was impossible with worse management practices.
What happened? You could expect the Toyota way: systematic approach leads to continuous improvements. Or the 'inappropriate technology' view: these are imposed externally, are not appropriate and will all be dropped. Neither holds.
Meta-analysis of small business trainings
The impact of business support services for small and medium enterprises on firm performance in low- and middle-income countries (2016
Campbell Systematic reviews
Paper is a meta-analysis of 'support services for SMEs' (40 studies).
They find a positive effect of these programs on business performance (0.15SD), but note that this is not very large. Also for employment creation (0.15SD, barely sig). No effect on productivity.
In Western countries SMEs contribute 60-70% of employment. This is lower in say africa where it is 20%. There is also some evidence that they contribute to GDP growth. This explains why it makes sense to focus on supporting SMEs.
This all is based on the assumption that SMEs need this support: either their informality prevents them from accessing credit markets, necessary for growth. Solutions can be tax breaks to make formalization more attractive, allowing them to get more credit. Or they face specific constraints (limited pool of labour, coordination failure etc). Solutions can be training programs, innovation support or clustering.
Regarding bias, 65% of studies fell under high risk of bias (at least 4 out of 5 issues present). An dummy for these reduces treatment effects to 0.9SD (still sig)
Also some indications of publication bias: high number of papers on the edge of sig.
Randomly giving people a job
The Impacts of Industrial and Entrepreneurial Work on Income and Health: Experimental Evidence from Ethiopia (2018)
American Economic Journal: Applied Economics
Paper compares the impacts of industrial work and the opportunity of starting a business on incomes, and health. Also provides insights into the hiring strategies of industrial companies in Ethiopia.
Worked with 5 industrial firms in Ethiopia that offered low-skill entry level positions. Very little interview process. People (mostly women) interested in such a job were given:
RCT (+ control):
1. Given a job offer at an industrial plant
2. Given a 5-day entrepreneurial training and a 300 USD startup grant.
For 1, 30% quit within a month. After 1 year only 23% was left. Especially those with observable human capital (education, experience) quit: likely their outside options were stronger. Wages in industry were similar/lower than informal jobs.
Entrepreneurship increased incomes, industrial same as informal market. Industrial jobs reduced health, mostly driven by a few with serious disabilities. Entrepreneurship led to small increase in subjective well-being.
Mechanisms: the job market for industry jobs is very fluid and competitive: and workers can easily switch between jobs. Industry jobs are used as temporary employment to smooth income shocks. They churn through jobs until they find something good or move to white-collar job/child caring. Many quit industrial jobs altogether, indicating that they learned the jobs were not for them.
Labor Turnover in Ethiopia
Understanding the challenges of high labor turnover in the Ethiopian manufacturing industry (2020)
Paper explores what drives turnover and quitting decisions in Ethiopian manufacturing.
Paper uses both 7-month panel dataset and qualitative interviews (n=24 women, n=11 husbands) to examine why women with partners quit manufacturing jobs.
Turnover is high: 50% of women who were offered a job quit within 7 months. Manufacturing jobs did appear to pay about 25% more and had more stable hours compared to informal employment.
Big drivers of quitting were unrealistic expectation of wages, and generally the type of work. Work was described as tiring and hard on their health. Many supervisors were incivil/abusive through shouting. Including salary cuts for late-coming, breaks or mistakes.
Women with more schooling and higher earning potentials were more likely to stay in a job. Women with higher-earning husbands were also more likely to quit. Women with more domestic responsibilities (children, household head) also more likely to quit. Few had a next job lined up.
Husbands were generally supportive and played a small role in the decision-making process.
Jobs and Political Participation
Jobs and political participation - Evidence from a Field Experiment in Ethiopia (2021)
Paper examines how a randomized job offer affects political participation in the short (6 months) and long (3 year) run.
Women with partners and an interest in a factory job were randomly assigned a job at said factory. 5 total survey waves to examine effects over time. Jobs are largely disempowering: women have low mobility, must work late to reach targets and are verbally abused. Labor laws are not enforced for fear of investors leaving country.
Treatment led to wage job and higher earnings. However, also a reduction in local government participation (Kebele meetings) and likelihood of raising political issues. In short run reduction in time available (but not long run). Also reduction in internal/external political efficacy (how they individually can affect policy).
Mechanism: expected to be resource hypothesis: work increases resources (education, income) and thus have the capacity to be more politically active. No evidence for that, instead: the disempowering aspects of the job reduce perception of political efficacy and thus participation.
Social skills in the labor market
The Growing Importance of Social Skills in the Labor Market (2017)
The Quarterly Journal of Economics
Paper examines how the importance of social skills has changed over time in the US. Based on key observation that more cognitive jobs have not become better paying, despite technological improvements (which should favor these). Hypothesis: the growth is there, but only for social skills which do not compete with computers.
Paper first develops a model to examine this. Based on the idea of working within teams where people specialize in tasks. People with higher social skills are better able to work together (To be precise: reduce transactions costs in trading tasks). Jobs that have higher variance in productivity have more scope for gains from task trade. As performance in a task is individually determined. Jobs with low variance in productivity have more tasks that can be classified as routine.
Model leads to testable predictions which are tested using data on cognitive/social/non-cognitive content of a range of jobs and two panel datasets (starting in 1979 and 1997). Social skills is measured using youth sociability, clubs and sports participation. For the 1997 data it is measured using Big Five Extraversion.
First, find that both cognitive and social skills matter strongly for wages, and complement each other. Workers with high social skills are also more likely to work in jobs where social skills matter more (e.g. they sort into these jobs). When they do, they are also rewarded with higher wages. Next, it examines whether jobs that favor social skills have grown more over time. They have increased more than analytical (e.g. math skills) which saw moderate increases, and way more than routine jobs which declined strongly. Holds for both employment and wages. Social skills are also rewarded more now.
Soft skills in Jamaican small businesses
The Impact of Soft-Skills Training for Entrepreneurs in Jamaica (2022)
Paper analyses the effect on small business outcomes of a intensive soft-skills training versus a combined soft-skills and business practices training (and control).
Implemented in Jamaica with small business (max 5 employees). Main constraints for businesses are contract enforcement, admin things and electricity. Both groups got 5 half-days training on personal initiative which tries to create self-starting, future-oriented and persevering mindset. Soft skills treatment then got 5 more half days that go more in-depth, especially focusing on dealing with setbacks (perseverance). The combined treatment got 5 half days on standard business practices.
High attrition (73% for 3 months, 59% for 12 months). Attendance rates at trainings were low as well: 60% attended at least five classes. They could not really find predictors for this.
Soft skills training leads to a 0.28SD increase in sales and profits index in the short run. Also 11% more likely to have positive profits. No effect after 12 months. Appears to be driven by the adoption of more business practices. Soft skills are somewhat increased by the treatment, but the effect is small. Also very small effect using incentivized games. Effect is driven by male entrepreneurs, but there are important heterogeneities at baseline: male-led businesses are substantially larger. Female-led businesses did adopt more practices but this did not lead to higher profits.
Making transformational entrepreneurs
Making Entrepreneurs: Returns to Training Youth in Hard Versus Soft Business Skills (2021)
Paper examines the 3.5 year impacts of mostly soft-skills and a mostly hard skills business training on youth (early 20s).
Early years (20s) the brain has high plasticity and determines long-term skills/grit etc. This is a crucial age to create transformational entrepreneurs (as opposed to subsistence entrepreneurs). These lead more stable businesses that generate more jobs.
RCT on nationally representative sample of students at the end of high school in Uganda. A 'majority' subscribed (this was a positively selected group with 70% interest in going to university). These students were randomly assigned to hard skills training, soft skills training or control. The hard skills training was 75% hard skills and 25% soft skills, and soft was reversed. Trainers were recruited and trained by the program, requiring a fee from trainees. Use machine learning for control variable selection and FWER for multiple hypothesis testing.
Training was effective: participants had higher hard (soft) skills after training concluded. 3.5 year there was no longer a gap between the two trained groups, but still with control. Soft skills group performed better at behavioral experiment focusing on negotiation and persuasion. Economic outcomes very similar between groups, but highly positive. More likely to be economically active, higher earnings, more (successful) businesses.
Highly cost effective with costs per participant at 118 US$. Treated arms earned around 70 US$ extra per month.
Labor Rationing (2021)
American Economic Review
Paper experimentally examines the surplus labor hypothesis by randomly hiring large proportions of laborers in Indian villages.
There is low (formal-ish) employment in rural labor markets. This might be because labor markets are dysfunctional and distorted. Or it might be because there are not enough jobs and workers go for self-employment as a last resort. There is oversupply if workers are rationing: they would prefer wage work at market wage over what they are normally doing.
RCT where in some villages a large portion of workers was hired (up to 24%). In control villages only 1-5 workers. Done both in periods with little other work (lean periods) and just before/after main harvest season (semi-peak).
In semi-peak months wage increases by 5%, and local employment goes down by 22%. This happens instantly, indicating that markets function well. In lean months the wage and aggregate employment do not move. But non-hirees are more likely to be employed: they take the jobs that are now in the work program. These would otherwise have been self-employed, so they do gain. In lean months at least 25% of labor is rationed out of the market. This is especially the case for small landholders, where perhaps more family labor is used than necessary.
To approach this normally they propose a survey question: If you had received a job offer on day X for the prevailing market price, would you have taken it? Can detect rationing somewhat, though it does have problems.
Efficient Corruption for jobs
Jobs for Sale: Corruption and Misallocation in Hiring (2021)
American Economic Review
Paper examines how corrupt hiring in a public service job works and its effects.
While corruption is generally thought of as a bad thing, there might be positive effects: those willing to pay for it might value it more as well.
Paper examines what Community Health Workers (CHWs) in a (undisclosed) developing country get upgraded to a supervisor position with substantial (40%) higher pay. All hirees paid bribes: average 17 months of wages. Performance data available for all CHWs under supervisors. Constructs a measure (SPI) that asseses how much supervised CHWs improved under supervisors (to see how well supervisors did).
Finds that applicants (historically) performed higher than non-applicants, and actual hires performed even better. Actual hires have higher values for the SPI. It appears that the hiring process is partially merit-based, and partially corrupt. This is driven by higher wealth individuals (who could bribe more) having higher education levels. So depending on the type of job corruption is less of a problem.
What constrains the growth of small firms?
Experiments and Entrepreneurship in Developing Countries (2019)
Annual Review of Economics
Paper gives an overview of experimental papers that aims to find what constrains the growth of firms in developing countries.
Important insight is that firms face different circumstances (e.g. formal vs informal) and this affects what constraints they face. This is hard to determine a priori (e.g. through a model), so experiments are extremely useful for this. Especially when experiments are test predictions from an economic theory, which aids in assessing how generally useful results are.
Urban labor markets appear to work reasonably well in many cases. Perhaps for higher skilled workers less so (but the evidence is weak).
Baseline data on firms/entrepreneurs is more predictive of firm performance than expert assessments of business plans. Peers might also do better.
The reason that many training programs have little effect might be because firms don't implement practices. Could driven by a lack of one of the four 'tions', in order: 1. Perception - do not know they are behind. 2. Inspiration - don't know how to fix the problems. 3. Motivation - No incentive to implement the changes. 4. Implementation - they cannot get it done within the organization. This constraint might differ per firm, and might explain why generic training programs are less effective.
Giving firms an extra worker
Labor Drops: Experimental Evidence on the Return to Additional Labor in Microenterprises (2019)
American Economic Journal: Applied Economics
Paper examines what the impact of a wage subsidy is on employment, firm survival rate and profits for small enterprises in Sri Lanka.
Many firms in developing countries are small, with no employees. This might be because adding additional workers does not increase the firm's profits. For example because the owner is not a good manager. However if there are labor market/learning constraints, a short-term subsidy could increase employment in the long run. E.g. to compensate for initial low productivity of workers, or to subsidize search costs.
RCT where a random subset of firms is given a 6-month subsidy equal to the minimum wage IF firms employ a worker for that period. Take-up of the subsidy was 24%. Subsidy-hired workers were similar to regular hires, though somewhat more likely to be related to the owner.
Find a small effect in the survival rate for wage-subsidized firms: about 5% more likely to still be in businesses: even after a long time (4 years). This is likely caused by the subsidy acting as a temporary capital increase, allowing them to weather more shocks. Employment increases during the subsidy but dropped immediately after, consistent with firms operating with an efficient number of employees. For manufacturing firms there appears to be more scope for increasing employment. No effect on profits. Return on labor is often below subsidy and minimum wage amount.