To study the effects of unionization on wages, you construct a panel of N individuals and T time periods. We assume that each individuals job does not change during the period. After holding focus group discussions with workers, you find the following phenomena:
1. Ordinary workers are more likely to join unions than team leaders and managers.
2. Workers with bad productivity shocks tend to join the union.
3. The average wage has been increasing during the period regardless of whether a worker joins a union or not.
(a) If you run the following regression, which of the phenomena above do you think might lead to the most serious endogeneity problem in the regression?
most serious endogeneity problem in the regression? wage; = Bo + B1 union; + ui Where wage is the individual i's wage and union is a dummy variable on whether that individual belongs to a union. Note that in this regression, as the indexes to the variables suggest, instead of using a panel dataset with multiple time periods, you are simply taking the first period (t=1) and discarding the rest. (b) Realizing that there is an obvious endogeneity problem above, you update the regression above to become wageit = Bo + ai + B1 unionit + Uit where a; is a set of dummy variables for each individual in the panel dataset. How many of these dummies will you want to include in the model? What endogeneity problems has this regression solved? Also, what endogeneity problems has this regression not solved? (again, to answer this, please refer to the above listed phenomena 1 to 3) (c) Finally, you update the regression to be wageit = Bo + B1 unionit + d + 8t + Uit Where 8 is a set of variables for each period t in the panel. What endogeneity problems has this regression solved? Also, what endogeneity problems has this regression not solved? (again, to answer this, please refer to the above listed phenomena 1 to 3) most serious endogeneity problem in the regression? wage; = Bo + B1 union; + ui Where wage is the individual i's wage and union is a dummy variable on whether that individual belongs to a union. Note that in this regression, as the indexes to the variables suggest, instead of using a panel dataset with multiple time periods, you are simply taking the first period (t=1) and discarding the rest. (b) Realizing that there is an obvious endogeneity problem above, you update the regression above to become wageit = Bo + ai + B1 unionit + Uit where a; is a set of dummy variables for each individual in the panel dataset. How many of these dummies will you want to include in the model? What endogeneity problems has this regression solved? Also, what endogeneity problems has this regression not solved? (again, to answer this, please refer to the above listed phenomena 1 to 3) (c) Finally, you update the regression to be wageit = Bo + B1 unionit + d + 8t + Uit Where 8 is a set of variables for each period t in the panel. What endogeneity problems has this regression solved? Also, what endogeneity problems has this regression not solved? (again, to answer this, please refer to the above listed phenomena 1 to 3)
