Labor markets in LMICs (what we know about growth, part 3)

This is part 3 of a series where I summarize the best available research on the causes and consequences of economic growth in developing countries. Part 1 focused on the sectoral transformation, where countries transition from most of their workforce being in agriculture to working in manufacturing or services. Part 2 took an even more granular lens and focused on the growth of firms. In this part, I will take the flip perspective: rather than looking at firms, I will look at workers.

Labor markets are how people earn their living. When labor markets fail to match workers to jobs, productive workers miss out on income that could be transformative for their lives, and the economy loses productive capacity that is essential for growth. So the functioning of labor markets in developing countries is a top priority for anyone who wants to promote economic growth.

1. Unemployment is much higher than official statistics suggest; self-employment is akin to unemployment.

How high is unemployment in developing countries? This question is surprisingly hard to answer. The typical approach to measuring unemployment is to ask whether people want to work but are unable to find work. By this approach, which is taken by national agencies, the unemployment rate in developing countries is similar to developed countries at 5-6%. (ILO)

However, this approach underestimates unemployment, by ignoring self-employed people. 55% of employment in LMICs is self-employment, rising to a staggering 80% in low-income countries. (ILO) “Self-employment” in developed countries conjures up the images of freelancers, contractors, entrepreneurs—people who choose self-employment because it has benefits over wage employment. However, in developing countries, it’s much closer to unemployment. Donovan et al (2023) observe that unemployed workers and self-employed workers switch into wage employment at similar rates, and they earn similar wages after switching. Moreover, self-employment does not seem to be a first step into stable employment; when workers in developing countries move from unemployment to self-employment, they are 24 percentage points more likely to return to unemployment in the next three months, compared to workers in developed countries to make the same move. Finally, a large share of labor market flows (transitions between unemployment and employment) in developing countries is just the movement between self-employment and unemployment, suggesting that people move between these two states at a very high rate. All of these facts suggest self-employment is a close substitute to unemployment. It’s obviously not as bad—some income is better than no income—but it suggests that self-employed people want wage jobs just as much as unemployed people, but cannot get them. Thus, there is an unemployment crisis in developing countries much beyond what the official statistics suggest.

How large is this problem? Breza et al (2021) find that at minimum, 24% of rural self-employment during India’s agricultural lean season occurs solely because workers cannot find jobs. If we heroically extrapolate this to a whole country, then the true unemployment rate in low-income countries jumps from 6% to 25%.[1] 25% unemployment is a catastrophic failure, and a crisis that cuts against both poverty alleviation for individuals and aggregate growth. The main takeaway is that unemployment is a much bigger issue than it seems in developing countries, and should be a high priority for development interventions.[2]

2. Wages are too high, plausibly because of social pressure.

Why is real unemployment so high? In a well-functioning market, wages would fall so that firms were willing to hire more of these unemployed workers. More generally, wages should rise when workers are more demanded, and then fall when workers are less demanded. Kaur (2018) shows that in village labor markets, the first part is true, but not the second. High rainfall in Indian villages increases yields and thus increases the demand for harvest labor, while low rainfall decreases yields and thus reduces labor demand. Yet wages increase in response to high rainfall, but they do not decrease in response to low rainfall. This failure of wages to move downwards is a key cause of unemployment, because it results in wages that are too high for employers to want to hire workers. While this evidence is specific to agriculture, Gelb et al (2017) show that manufacturing wages in Africa are much higher than those in countries with similar incomes; it stands to reason that this would also cause unemployment.

Why don’t wages fall? Workers surveyed in Kaur (2018) say that wage cuts would be unfair. Furthermore, at labor stands in Kenya, 73% of workers say that their peers would stop talking to them if they worked for below the market wage; 53% say their peers would stop lending money to them; 67% say they would be informally barred from working at the labor stands. (Breza et al, 2023) Moreover, randomly offering workers jobs in private (rather than in public) makes workers 25 percentage points more likely to accept a wage 10% below the prevailing wage. This points to significant social pressure keeping wages too high—a discouraging conclusion, since circumventing social pressure seems quite hard. But it’s at least the right question to ask, and there could be scope for a social norms intervention here.

3. Workers see significantly lower wage growth over their life compared to workers in developed countries.

Each country has a life-cycle wage profile; a relationship between how old someone is, and how much they earn. In rich countries, a person can expect their wages to double over the course of their working life; in poor countries, they can only expect a 50% increase. (Lagakos et al, 2018) In other words, life-cycle wage growth is twice as high in rich countries compared to poor countries.

The main reason this disparity matters is because wage growth over the life-cycle is a path to upward mobility for poor people. If you make more money as you get older, that makes it possible to enter the middle class, build up savings/​assets, invest in your children’s education, and other things that we understand as necessary to build up a broad middle class in any country. But a more subtle reason we care about wage growth is that it signals growth in skills. In a world where businesses pay people more when they are more productive, wage growth reflects the fact that older and more experienced workers are simply better at doing a job/​doing harder jobs than younger and less experienced workers.[3] Interpreted this way, stagnant wage growth means that workers may not be gaining much experience and skill over their life cycle. This limits their productivity growth, and thus economic growth as a whole.

So why is wage growth over the life cycle so different across countries?

  1. Education. Across countries, educated workers have steeper life-cycle wage growth than uneducated workers. This could reflect their ability to accumulate more skills from a job than uneducated workers, if education makes them able to pick up skills more easily while on the job. The differences in education rates between developed and developing countries alone accounts for one-third of the difference in wage growth.

  2. Slippery job ladder. The flat wage growth on average masks important heterogeneity; conditional on staying in the same job, wages grow faster in developing countries than in developed countries. (Donovan, Schoellman and Yu 2023) So the weak average growth reflects the fact that people are much less likely to stay in the same job, and that when they switch jobs, they often see wage decreases: workers who move into a high-wage job are 39 percentage points more likely to leave it within three months compared to similar workers in developed countries, moving into a lower-wage job or unemployment. This is a slippery job ladder; rather than climbing the rungs from unemployment to stable jobs that pay more and more, people “fall off the ladder” and move to lower-paying jobs, which prevents them from seeing sustained wage growth.

  3. On-the-job training. Firms in developing countries provide less on-the-job training to workers (Ma, Nakab and Vidart 2024) which hinders them from learning over their life cycle. This behavior is a logical response to the slippery job ladder described above; if workers stay in their jobs for much less time in developing countries than in developed countries, firms have no incentive to invest in training them, since they won’t be able to reap the benefits of training them. This difference accounts for another third of the difference in wage growth.

How can we target life-cycle wage profiles so that people see more wage growth? For one thing, education interventions look much better because of their potential impacts on life cycle wage growth being much larger than their effects on wages at some point in time (what people normally look at). Moreover, subsidizing firm-provided training could incentivize firms to train their workers more even in the face of turnover.

4. Village labor markets are efficient during peak seasons, but inefficient during lean seasons.

Rural labor markets employ the majority of workers in developing countries, and it’s important to understand how they function in particular. What does it mean for a labor market to work efficiently? The core criterion is that it clears; every worker who would like to work at the market wage gets a job, and every employer who would like to hire at the market wage gets a worker. Labor market clearing is the bedrock of an efficient labor market. Do village labor markets in developing countries clear? The answer depends on whether we are talking about the peak season (agricultural activity is high) or the lean season (agricultural activity is low).

During peak seasons, the evidence is unanimous that labor markets clear. Jayachandran (2006) finds that when agricultural productivity increases due to quasi-random rainfall shocks, agricultural wages increase. Moreover, Breza et al (2021) randomly hire a large number of workers across Indian villages, and when they hire during the peak season, local wages increase (as the number of workers decreases). The responsiveness of wages indicates a healthy labor market that is responsive to local conditions and worker productivity, a reassuring finding.

During lean seasons, however, labor markets likely do not clear, and workers are under-hired relative to an efficient market. Breza et al find that even hiring 24% of a village’s labor force outside of the village during the lean season has no effect on village wages, suggesting that at least 24% of workers wanted to work at the prevailing wage but were unable to. Muralidharan et al (2022) randomize an intensification to India’s rural employment guarantee and find that in treated areas, private wages and employment increased. These two results together are only consistent with village employers acting as monopsonists when hiring workers, and thus hiring workers less than is efficient.[4] This is a source of inefficiency that is especially likely to bind during the lean season when fewer employers are hiring workers.

The fact that village labor markets under-utilize workers during the lean season is strong evidence in favor of governments providing rural employment programs during the lean season; such active labor market policies are common in developing countries, and it’s reassuring to know that they might be effective (subject to implementation issues).

Conclusion

Labor markets are crucial for people to earn incomes, and for economies to grow by matching workers with firms to produce more. Developing countries have labor market challenges that significantly limit their ability to use productive workers, as well as the ability of those workers to earn income and grow their skills.

This post is much lighter on evidence-based solutions than the previous posts. This is partly a function of labor markets being a much harder object to grapple with, compared to firms or households—and it’s partly a function of the fact that research into labor market failures and their connection to aggregate economic growth is scarce. We are still in the phase of establishing the problems, and it remains to be seen what the right solutions are.

But there’s an optimistic side to thinking about labor markets; because the core problem is matching workers and firms who would like to be matched, and this matching is fundamentally decentralized, I see this as one area where private approaches to solving problems are unusually effective development interventions. Consider a few examples:

  1. An entrepreneur sets up a job board in a big city, going to many employers in the city and asking them to send her information whenever they want to hire workers. She collects a lot of job postings and takes them to slums where lots of workers live, including new migrants who don’t know where to look for jobs. If they apply for these jobs and the employer hires them, the employer pays her a commission.

  2. An entrepreneur offers a certification program for workers to demonstrate literacy and numeracy skills, knowing that the education system has spotty quality, so many high school graduates will lack foundational literacy and numeracy. He offers employers insurance payments in case their worker fails to meet job requirements and needs to be fired. If workers get hired for their certification, they pay a share of their wages to the entrepreneur up to some limit.

  3. A mobile payments company knows that seasonal migration is profitable for many workers, so it offers a loan for migration during the lean season. The loan has a high-enough interest rate to be profitable even if many borrowers default, but low-enough that migrants can repay it while still having enough income to make migration worthwhile.

None of these are bulletproof ideas, only demonstrations of an underlying fact; fixing labor market failures generates a positive surplus, and whoever fixes those failures can claim some of that surplus as their reward. The second example is the real model of Harambee, a nonprofit that has been evaluated successfully (Carranza et al 2020). So while the evidence is not as clear on effective solutions as it is on problems, perhaps people who are motivated to solve these issues can blaze the trail.


  1. ↩︎

    The official unemployment rate is 6%; of the remaining 94%, 80% are self-employed. If 24% of those are functionally unemployed, then 24% * 80% * 94% = 19% of the labor force is unemployed but not recorded as such. 19% + 6% = 25% unemployment.

  2. ↩︎

    This argument is contradicted by Feng et al (2023) who measure unemployment from household surveys and find that in developing countries, unemployment is lower than in developed countries. The difference comes down to the definition of unemployment, as Feng et al require someone to do no work at all to count as unemployed.

  3. ↩︎

    Skeptical readers might point out that wage growth could just reflect long-term contracts prevalent in developed countries, wherein wage growth is built into the contract, regardless of a worker’s skill. Lagakos et al look for evidence that long-term contracts drive the difference between developed and developing countries, and conclude that it does not. So it’s likely that the true explanation involves real differences in skills across workers.

  4. ↩︎

    This is a standard but counterintuitive economic argument; a monopsonistic employer will actually hire more workers when they have to pay higher wages. See here for an intuitive explanation of why this is the case.