Returns to experience across tasks: evidence from Brazil

ABSTRACT

Using a rich Brazilian panel dataset and an occupation-task mapping, we investigate whether returns to experience depend on the types of jobs performed by workers. We find that returns to experience in non-routine tasks, especially returns to analytical tasks, are much larger than returns to routine tasks. This gap increases with schooling, suggesting that schooling and non-routine tasks are complementary in the human capital production function. These are important findings for developing countries similar to Brazil, where approximately 70% of workers’ tasks are routine.

KEYWORDS:

• Task-approach
• human capital
• returns to experience
• life-cycle wage profile

JEL CLASSIFICATION:

• J20
• J24
• J31

Acknowledgments

We thank Bruno Funchal for sharing the initial task-occupation mapping. We thank Juliano Assunção, Leonardo Rezende, Rodrigo Soares, Gabriel Ulyssea, Eduardo Zilberman, and participants in seminars at IPEA and the 4^th LACEA Labor Network Conference for valuable comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ We should note that our measure of task experience is different from the one used in Gathman and Schonberg (2010).

² Our findings should also help explaining cross-country differences in experience-wage profiles, as most workers in less-developed countries perform routine and manual tasks (World Bank 2012).

³ We drop workers employed in the agriculture, military and public sectors. We do not include workers with more than one simultaneous job. To avoid measurement error, we exclude workers with any variation in schooling across years. For workers with college education, we only include those who were 22 years old or less in 2003.

⁴ Our measure of actual experience does not include experience in the informal sector.

⁵ Formally, we use the task classification to decompose each observation of a worker’s total experience in RAIS data into the five task-experience measures (and the two more aggregated task-experience measures, routine and non-routine), according to the equation below:

$Ex{p}_{i,t}^k=\sum _o\left(OccTe{n}_{o,i}\ast Tas{k}_o^k\right)$

where, for each worker i in period t, experience in task k is defined as the cumulative sum (across four-digit occupations) of current and previous tenure in each occupation o multiplied by the task k content of the occupation. Note that the sum of $Ex{p}_{i,t}^k$ across the k tasks equals total experience,$Ex{p}_{i,t}$.

⁶ Note that in our sample, an observation unit is constituted of a worker’s log wage ($ln{W}_{i,t})$ and her task k experience ($Ex{p}_{i,t}^k)$ at any time of the year. Workers may have more than one observation per year if they changed jobs during that year. Note that observed variables that do not change over time, such as schooling, gender and race, are picked up by individual fixed effects.

⁷ Note that occupation fixed effects also account for task composition within an occupation.

⁸ In our main specification, we interpret coefficients associated with task experience as the speed at which wage grows for workers employed in each task. In the specification with industry and occupation fixed effects, we interpret coefficients associated to task experience as the average wage growth conditional on being in a certain industry and occupation weighted by a function of industry and occupational composition.

⁹ The CBO structure is coded into 614 four-digit occupations. Task data comes from Funchal and Soares (2013).