This is the 11th in this year’s series of posts by PhD students on the job market.
The process of economic development entails the reallocation of labor and capital from agriculture into more productive sectors, namely manufacturing and services. In many countries that remain poor today, this process has been occurring slowly. In sub-Saharan Africa, large gaps persist between rural, predominantly agricultural, and urban incomes, and most workers are self-employed. This slow transition is typically attributed to frictions that prevent workers from optimally allocating across space and into jobs. But a job is more than a wage; it is a location, an environment, tasks, and a part of the worker’s identity. This can be a source of labor market power and can lead to an allocation of labor that is inefficient for production.
In my job market paper, I measure the geographic distribution of wage markdowns in Tanzania. To do so, I build a spatial general equilibrium model of monopsony that accounts for both local and spatial frictions. The extent to which a firm can markdown its wage depends upon its share of employment—measured at several levels—and how elastic workers are between types of employment. These elasticities are captured by three choices that a worker faces: between firms in their local market, between wage and self-employment, and between staying or migrating to another labor market.
Estimating Labor Market Power
The wage markdown is the wage share of the marginal product of labor that a worker is paid; it is a latent quantity that we need a model to estimate. Doing so in a low-income country context is difficult, it requires data on wages and employment and estimates of three key elasticities.
Comprehensive data on wages and employment in sub-Saharan Africa is sparse. One exception is Tanzania’s annual census of firms, the Employment and Earnings Survey, which I use to measure firm wages and employment. However, this misses a large share of total employment—85% of workers are self-employed. I measure total self-employment in each labor market using the 2012 census and I calculate self-employment income using consumption equivalents in the Tanzania National Panel Survey.
To estimate the labor supply elasticities, I use Tanzania’s 2010 sectoral minimum wage law which specified a specific level for 20 industries and a national floor for all others. Non-compliance was low; just 10% of workers were paid below the minimum wage in the first year of enactment. To estimate the between-firm elasticity, I instrument for wages using the bite of the minimum wage measured as the gap between the firm’s current wage-bill and the minimum wage compliant wage-bill. I estimate a between-firm elasticity of 2.5, comparable to estimates for Ethiopia (3.36).
Next, I estimate the sectoral elasticity (between wage and self-employment) using the model generated moment that relates the ratio of firm to self-employment to the relative wages in the two sectors. I estimate a sectoral elasticity of 1.5. This estimate depends upon the calibrated value of the between-firm elasticity. To assess the sensitivity of the estimate to the between-firm elasticity, I vary the calibration. The ratio of the two elasticities is roughly constant across calibrations, indicating that the way in which firms compete for workers is different from the way in which firms compete for workers from self-employed.
Finally, I estimate the migration elasticity using migration flows from the 2012 census. The model relates the share of migrants between any origin and destination pair to their relative wages and location amenities. To disentangle their roles, I take advantage of the sectoral component of the minimum wage law. Spatial variation in industrial composition meant that the average effective minimum wage varied across space. I use this variation as an instrument for wages and estimate a migration elasticity of 1.4. This value is comparable to estimates for China (1.5) and Indonesia (3.2).
Taking Stock
The model makes three important predictions about wage markdowns. First, wages are more competitive in small firms. This is because smaller firms have less oligopsony power when setting wages. Second, wages are more competitive in markets with more self-employment. If the rate of self-employment is higher, then either the local productivity in self-employment, and hence income, is higher or there are large job search frictions. Both force firms to post higher wages to attract workers. Third, migration has an inverse-u relationship with markdowns. In areas with low emigration rates, workers face high costs to migrate away. This allows firms to retain workers at lower wages. High immigration areas are characterized by either high amenities or high average productivity. These features give firms more ability to mark down wages.
The Spatial Distribution of Labor Market Power
Most workers are paid between 66-71% of marginal product. Wages are less competitive in rural labor markets. Figure 1 plots the average wage markdown against population density. Lower values for the markdown indicate less competitive wages. Wages are more competitive in districts with a higher population density. The markdown asymptotes around 71.5%, which is the most competitive wage that a firm would pay based on the estimated labor supply elasticities.
Figure 1 Wages are more competitive in districts with a higher population density
The reason that rural labor markets are less competitive is more nuanced than higher firm concentration. As noted above, self-employment plays an important role in diminishing labor market power. Figure 2 plots the equilibrium markdown against firm employment share for the average rural and urban labor market. Comparing two firms of equal employment share in a rural and urban market, the rural firm would pay the more competitive wage. This is largely due to the higher rate of self-employment. The reason that rural markets are less competitive is that the average worker is employed in a firm with larger market share.
Figure 2 Rural firms pay more competitive wages
Policy Implications
Using the model, I can quantify how output and welfare would change from paying competitive wages, reducing job search costs, or reducing migration costs. While I don’t consider any specific policies, these comparisons shed light on which types of policies are likely to see the biggest effects. The largest output gains come from reducing job search costs. This makes sense; a defining difference between sub-Saharan Africa and other developing countries is the higher rate of self-employment.
The most surprising result, however, is that reducing migration costs causes output to fall. How can this be? For starters, the urban wage premium isn’t that large. The rural-urban wage gap is primarily driven by differences in self-employment earnings. More importantly, when workers choose where to live and work, they consider the job and location amenities as well as the wage. This creates a wedge between the output and welfare maximizing labor allocations. A unilateral reduction in migration costs makes a range of possible work-life outcomes possible for workers that may not be more productive.
At first glance, this result seems to conflict with growing evidence that reducing migration costs leads to a more productive allocation of labor. This study adds to that literature by explicitly allowing workers to value things other than wages when making migration decisions, breaking the necessary link between lower migration costs and higher output. I can recover the standard result in two ways. The first is by additionally changing wage competition so that firms pay competitive wages. This increases the relative value of urban, high productivity areas enough to redirect workers there. The second way is to only reduce migration costs in the direction of the city, thereby only increasing the relative value of migrating to higher productivity areas.
Samuel Marshall is a PhD candidate at the University of Warwick.
Join the Conversation