The firm you work for is non-unionized. A co-worker is considering forming a union and applying for certification. To make sure that this is worthwhile, she asks you about the impact of unions on wages of union workers.

A union organizer provides you with cross-sectional individual-level data and asks you to estimate the impact of unions on wages with the following equation:

ln(wi)= + Ui + Xi + i

where wi is worker i's wage, Ui indicates union status, and Xi is a vector of observable characteristics.

a) what (potentially observable) characteristics would you include in Xi? Why?

b) can you confidently say that your estimate of constitutes an unbiased estimate of the causal impact of unions on the wages of their members? Explain. If not, propose an alternative approach that does not suffer from bias.