Observing labor flows help economists analyze the health of an economy, from unemployment levels to the impact of sudden shocks on the workforce. However, getting a close look at labor flows can be difficult; it requires broad data across all employment sectors that follow the same people over time, so that the researcher can trace workers’ paths between employment states. To analyze labor flows in one country, economists can use data gathered by government agencies and other official surveys.
But what if researchers want to study the relationship between labor flows and other macroeconomic variables, like GDP? To do that, they need data across several countries.
Kevin Donovan, an assistant professor of economics at Yale SOM and an affiliate of the Yale Economic Growth Center, and co-authors Will Jianyu Lu of the Central Bank of Chile and Todd Schoellman of Federal Reserve Bank of Minneapolis constructed a novel dataset that consists of labor information for 80 million people across 49 countries, including many developing nations. Creating such a dataset is a daunting task: even if a given country records the needed data, they likely use different methods, styles, or definitions. Because this dataset is so massive, it required a lot of manipulation to be usable. The Economic Growth Center supported the project by providing a research assistant to the team for a year.
“It’s a difficult question,” Donovan explained in an EGC interview. “Researchers have observed certain patterns within rich countries, focusing on a few countries at a time. But can you extrapolate those results to developing countries? And what drives those cross-country differences that are observed with highly aggregated data? Those types of questions require more detailed data that haven’t been available to researchers.” After constructing the dataset, the team leveraged their data to make striking discoveries about the relationship between labor flows and economic development. Their paper “Labor Markets and Development” is forthcoming in the Quarterly Journal of Economics.
Donovan and coauthors’ first key result is that all labor flows are negatively correlated with GDP per capita. In other words, they show that people in the poorest countries move into employment, out of employment, and change jobs the most.
But what do these flows represent? There are several possible interpretations. One is that, because many developing countries are growing quickly, the observed flows could indicate that workers are moving to more productive jobs or that new technologies are pulling people into the labor market from unemployment. A more negative hypothesis is that the structure of developing countries makes it difficult to climb the job ladder into these more productive jobs, so these flows aren’t greatly contributing to growth.
The authors find more evidence in support of the latter explanation. In poorer countries, people aren’t consistently switching to higher paying jobs. Instead, these high flows are entirely accounted for by movement into and out of marginal jobs: informal wage work, self-employment, and low-earning formal jobs. At the heart of this issue is a “slippery” bottom of the job ladder. Non-employed people who start a job are substantially more likely to move back into non-employment in developing countries. Those who are lucky enough to move from a low to high paying job are likewise more likely to see their wages decline again or lose their job entirely. These results imply that the high labor flows observed in poorer countries are concentrated among the least well off within those countries, shuffling between various states and similar jobs rather than moving to better positions over time.
If labor flows in developing countries are concentrated among people who cannot consistently move to jobs where they are more productive, why can’t they find better jobs? Understanding the answer to this question is essential to designing policy solutions that address underlying causes rather than the symptoms of an inefficient labor market.
What they observe is that developing countries have more initial low-quality matches between firms and workers, but also more rapid exit from them. Such a pattern is consistent with a growing body of microeconomic experiments in the labor markets of developing countries which focus on information asymmetries between firms and workers. In rich countries, workers and firms have a variety of tools to uncover whether a match is likely to be successful before hiring. Firms can look at workers’ résumés and workers can learn information about whether the company meets their preferences, all before hiring. This information tends not to be as easily accessible in poorer countries. Instead, workers in developing countries become “experience goods,” meaning that workers and firms learn how well they suit each other only after the worker gets hired. That is, the firm and worker have to “experience” the match before realizing whether they are a good fit. This forces them to create more low quality matches to weed out the bad ones from the good.
This research is unique in its scope: by analyzing such a wide range of countries, the researchers find results that are both highly unexpected and empirically rigorous. While the authors admit that it is difficult to derive policy implications directly from their cross-country data, they are hopeful that it can spur additional research that will provide more specific solutions in more localized contexts.
Dorothea Schmidt-Klau, senior economist at the International Labour Organization’s Employment Policy Department, expressed optimism about the kind of work the dataset will enable. “This research is exactly what we’ve been waiting for,” she said. “The data set is unique and innovative, and the analysis provides important insights for employment policy development to help the world reach Sustainable Development Goal 8, which aims to provide full, productive employment and decent work for all. We are looking forward to working with the data as well as the researchers, and we’ve already identified questions for future research.”
The dataset, which will be publicly available on the research team’s project homepage, sets the stage of future work to determine the correct intervention for specific contexts. “One of the things our dataset can offer is context,” Donovan explained. “If a researcher finds some result during a labor market intervention in Uganda or South Africa, our data is something they can use to evaluate whether the results seem consistent with broader cross-country patterns that relate poverty and labor market outcomes. This type of complementarity is useful as we start to develop a suite of labor market policies for the developing world.”