International trade theory has typically ignored the costs of adjusting to a change in trade policy, focusing instead on static models and long-run conclusions. However, adjustment costs are central to much of the controversy over trade liberalization. This work measures the importance of the dynamics of adjustment in shaping the distributional consequences of trade policy, and characterizes the time it takes for the economy to complete the transition to the new long-run equilibrium. When these adjustment costs are taken into account, the benefits of trade liberalization can be substantially smaller.
The gains from international trade depend upon the ability of countries to specialize in the production of goods and services where they have a comparative advantage. Thus, the reallocation of workers and capital across sectors and firms is at the heart of the gains from trade. However, this reallocation is also at the center of tensions between economists and policy makers. On one hand, economists tend to emphasize long-run gains from trade within models where labor and capital are perfectly mobile across economic activities. On the other hand, policy makers tend to be concerned with the potentially slow and costly adjustments to trade policy as workers and capital transition to new sectors. When these costs are ignored, the gains from trade may be substantially overstated, especially from the point of view of current generations who will bear the largest chunk of adjustment costs, and who will have to wait longer to fully enjoy the benefits from trade liberalization.
Subsidizing workers in adversely affected sectors to switch to a new sector tends to outperform policies that retrain workers to enter new sectors.
In a recent paper, Dix-Carneiro (2014) estimates the barriers that workers face in adjusting to trade policy changes and assesses how these barriers affect the dynamic adjustment to the new long-run equilibrium. The paper also examines how the costs of adjustment are distributed across various types of workers, and identifies the types of workers likely to experience the highest adjustment costs.
The model features new generations of workers entering the labor force and existing generations of workers retiring every year. This dimension of the model lends flexibility to the economy, as young and more mobile workers gradually replace older and less mobile workers.
At each point in time, each worker decides which sector to work in based on the discounted present value of income that her choice will bring her. Sectoral and individual conditions change over time and, therefore, give incentives for some workers to switch sectors. The model developed in the paper considers three types of barriers to mobility between sectors: direct costs of switching sectors, individual-level comparative advantage across sectors, and imperfectly transferrable sector-specific experience.
we can assess the distributional consequences of trade at a level that is much finer than what is usually explored
The first barrier to mobility, the direct costs of switching sectors, may include factors such as firing costs, the cost of moving to a new geographical area, or search and matching frictions that induce spells of unemployment.
The second barrier, individual level comparative advantage, arises from the fact that the returns to worker characteristics (observable and unobservable) differ across sectors. Therefore, workers sort themselves across sectors based on their characteristics. For example, if the returns to education are higher in High-Tech Manufacturing, then more educated workers will tend to be relatively concentrated in that sector, and they will be reluctant to switch to a sector where their characteristics are less highly valued.
The third barrier to mobility, imperfectly transferrable sector-specific experience, is of particular interest to policy makers. The model accommodates the possibility that experience accumulated in a given sector is not perfectly transferrable to another sector. As a consequence, workers not only face switching costs when moving across sectors, they can also face direct wage losses from doing so. Imperfectly transferrable sector-specific experience has the potential to slow down adjustment in response to trade liberalization and to significantly hurt workers who are highly specialized in adversely-affected sectors.
A worker contemplating switching sectors will only do so if the discounted present value of income is sufficiently high to compensate her for the mobility costs she will incur. When computing the present value of working in each sector, the worker observes the wages she can collect, and forecasts how these wages will evolve over time. Sector-specific wages depend upon characteristics that are observable to the researcher (age, gender, education, sector-specific experiences) as well as on characteristics that are unobservable to the researcher (for example, innate aptitude to work in the High-Tech Manufacturing sector).
the median direct mobility costs that workers face in switching sectors range from 1.4 to 2.7 times annual average wages
The incorporation of extensive observable and unobservable worker heterogeneity is a distinctive feature of the model. From a technical perspective, this level of richness is necessary because the estimation of mobility costs requires the best possible estimates of counterfactual wages the worker could get by switching sectors. The incorporation of observed heterogeneity goes a long way in modelling these counterfactual wages, but when workers are also heterogeneous along dimensions that are not observed by the econometrician, the relevant counterfactual wage is known only to the worker – the econometrician can only observe the realized wage in the chosen sector. If workers self-select into sectors based on unobserved components of wages, this process must be carefully modelled if we hope to extract consistent counterfactual wages. One reward from the degree of detail this requires is that we can assess the distributional consequences of trade at a level that is much finer than what is usually explored.
Estimation, results, and policy implications
The paper estimates the model using administrative data from Brazil (Relação Anual de Informações Sociais), which allowed the author to follow millions of workers over 10 years. With these data, we can track workers over time and across sectors, observe their key characteristics (age, gender, education, wage) and measure the number of years they have worked in each sector.
if Brazil were to … reduce tariffs on the sectors with higher technological content, the reallocation of workers across sectors could take up to 9 years.
Brazil is an interesting case study because it not only possesses data that allows the estimation of this demanding model, but it is also a relatively closed economy. Although the country went through a major trade liberalization episode in the early 1990’s, trade barriers are still high. Therefore, understanding the labor market effects of a hypothetical new round of liberalization in the country is of practical importance.
Estimates of the model point to significant barriers to the mobility of workers across sectors. Indeed, the median direct mobility costs that workers face in switching sectors range from 1.4 to 2.7 times annual average wages, depending on the sector of entry. In addition, experience accumulated in one sector is not perfectly transferable to another. For example, one year of experience accumulated in the construction sector is worth 2 to 4 times more in the construction sector than in the manufacturing sector. Importantly, direct inter-sectoral mobility costs are quite heterogeneous across the population, with women, unskilled workers and older workers facing significantly higher costs. These results have important consequences for the distributional consequences of trade liberalization: the identity of the winners and losers is determined not only by the initial sector of employment, but also by worker characteristics. For example, if tariffs are reduced in manufacturing, workers initially employed in manufacturing lose relative to workers in other sectors, but older workers in manufacturing lose even more.
the present value of the gains from trade are estimated to be 11- 26 percent lower compared to a situation where reallocation occurs immediately.
The estimated model can be used to calculate the duration of the transition following trade liberalization and the magnitude of the present value of the gains from trade. Results in Dix-Carneiro (2014) indicate that if Brazil were to implement a new round of trade liberalization and reduce tariffs on the sectors with higher technological content, the reallocation of workers across sectors could take up to 9 years. This delay in the reallocation of workers across sectors is consequential for the gains from trade: the present value of these is estimated to be 11 to 26 percent lower compared to a situation where reallocation occurs immediately.
one year of experience accumulated in the construction sector is worth 2 to 4 times more in the construction sector than in the manufacturing sector
The framework developed by Dix-Carneiro (2014) allows for the simulation of government policies designed to compensate workers adversely affected by trade liberalization. Subsidizing workers in adversely affected sectors to switch to a new sector tends to outperform policies that retrain workers to enter new sectors.
Current and future work
Finally, the findings in Dix-Carneiro (2014) are a big step forward, but we still have much to learn about the mobility of factors across sectors. The model and data used in this study only allowed the estimation of the costs workers faced in moving across sectors. However, one of the findings of the analysis is that the transitional dynamics and the extent to which the gains from trade are mitigated by the slow transition of the labor market depends crucially on the degree of mobility of factors complementary to labor such as capital.
wages keep falling in adversely affected sectors for years after the end of liberalization, leading to a long and costly transition
The most interesting simulation is when both workers and capital are imperfectly mobile across sectors. Capital may be imperfectly mobile because installed capital is difficult to move, but it depreciates over time. On the other hand, new investment can be allocated efficiently. As such, labor demand in an adversely affected sector will instantaneously shift down in response to trade policy, and it will keep falling gradually as new investment is directed toward non-affected sectors and installed capital depreciates over time. This can lead to a situation where wages keep falling in adversely affected sectors for years after the end of liberalization, leading to a long and costly transition.
Interestingly, a recent paper by Dix-Carneiro and Kovak (2015) shows that this pattern of adjustment is likely to occur in practice. They document that Brazilian regions that faced larger tariff cuts between 1990 and 1995 experienced a long period of gradual declines in wages relative to regions that faced mild tariff cuts. They rule out several possible mechanisms that could explain this pattern of adjustment, but show an array of evidence supporting the argument that labor demand kept falling in adversely affected regions for years after the end of the liberalization (relative to the national average). One plausible mechanism behind these dynamics in labor demand is slow adjustment in capital via redirection of investment and depreciation of installed capital in adversely affected sectors.
The natural next step in this literature is to develop models and collect the appropriate data that would allow us to estimate mobility frictions for both workers and capital.