All Things Upwork

We recently completed a study the analyzed the impact of worker tenure, training, and country of origin on hourly wages. Our initial hypotheses were that there are positive returns to tenure (length of employment) and training (number of tests taken, scoring of tests taken), and that the worker’s country of origin affects wages.

Based on the results of several regressions to test our hypotheses, we concluded that there are positive returns to tenure for certain job types, particularly those that require technical expertise.


Overall, it appears that managers have been willing to pay a premium for tenure but they do so selectively. For example, a manager may be willing to pay higher for a developer with a longer tenure on oDesk, but may not be willing pay more for a data entry worker for a longer tenure.

Though oDesk does not provide formal training to workers, we defined “training” based on oDesk’s skill-specific tests.


We concluded that wages increase with the number of exams taken, higher exam scores yield higher wage returns, and some exams affect wages more than others.



Our analysis on country of origin led us to conclude that workers living outside North America earn wages higher than the mean in their home markets. The analysis also showed that North American workers earn higher wages than workers in other countries at a statistically significant level.

At a high level, it is clear that workers from the United States and Canada are earning wages that are higher than their counterparts from Eastern Europe and Asia. Upon closer inspection, however, the story is more subtle and complex than it first appears.


Much of the apparent country effect can be explained by observable differences in the labor forces. There may be something structural about the environment in each of these countries that prompts the differences in the makeup of the labor force, but it cannot be ignored that workers with similar characteristics around the world selling their services through oDesk are earning wages that are more similar than a simple analysis would imply. When the data is reweighted to take into account similarities in tenure and training, the mean wages in each country adjust to a more common level.


Coming back to our initial hypotheses, the data reinforces that there are positive returns to tenure (length of employment) and training (number of tests taken, scoring of tests taken). Additionally, while a worker’s country of origin affects their potential wages, similiarly-skilled providers on oDesk command wages that are statistically more alike than not.

Our results were based on a study of providers over a three-year period. Are you seeing these results played out in the marketplace? Has your experience been different?

Special thanks to Ashley Carroll, Ruth Bryson, Jeremiah Dillon and Brandon Paulson – MBA 2010 candidates at Stanford’s Graduate School of Business – for their research efforts and contributing this guest blog post!