Human Capital Investments and Individual Earnings

"Ninety percent of an economist's job is doing the work, not getting the answers."  Orley Ashenfelter

In this assignment you will explore earnings differences across individuals and evaluate the influence of various proposed determinants. You will begin with several initial steps to gain familiarity with the data for this exercise as well as software tools for the course. In subsequent steps you will implement the statistical earnings function, develop estimates of the returns to human capital investments and consider the effects of union status.

There will be sufficient in-class time devoted to getting everyone started on the empirical work and you are encouraged to consult with classmates; however, completion of the exercise and written report is your individual responsibility. General suggestions for the written report include: organize data into meaningful displays (do build nice tables), be somewhat selective when reporting regression statistics (do not paste everything that is available to cut), analyze and apply critical insight (do think, do not blindly trust everything regression output appears to suggest), communicate your findings with a clear and concise report.  Keep the due date in mind as you make steady toward completion.

ü      download data and create workfile

The data for this exercise is available for downloading as CPS97 in MS Excel workbook format.  There are two worksheets, one with variable definitions and the other with data.  The cross-sectional data are taken from the March 1997 Current Population Survey (CPS), a monthly household survey that generates detailed information on thousands of individuals.  See the BLS-Census site for more information on the data source.  There are several ways to convert the data to EViews workfile format.  Use the Demonstration instructions in Help as a guide.

ü      initial data exploration

Begin with the entire sample and note the total number of observations. Determine the mean, median, standard deviation, minimum value and maximum value for WKLYEARN, HRLYWAGE and LNWAGE. Use the EViews GENR command to generate exper = age – edu - 6.  Obtain variable means for the set of human capital variables (education, experience, experience squared) and other core wage determinants (union, female, immigrant, msa, regions, and firm size). 

What do the mean values for UNION, FEMALE, IMMIGRANT, MSA and FIRM_LARGE tell us about the composition of the full sample?

Determine the simple correlation between LNWAGE and each of these variables.  How would the interpretation of regression coefficients differ from these simple measures of correlation? Explain this general point clearly.

ü      regression analysis using the full sample

Use LNWAGE as the dependent variable in your evaluation of earnings differences. Obtain regression results using the set of human capital variables and other core wage determinants as your explanatory variables.  Present regression results including number of observations, adjusted r-squared, residual sum of squares, explanatory variable names, regression coefficients and t-statistics (or standard errors).

What do the estimated coefficients on UNION, FEMALE, IMMIGRANT, MSA and FIRM_LARGE tell us about current wage patterns?

What are the returns to education and returns to experience in the context of your regression results?

ü      exploring union wage effects

Sort the full sample by UNION to form two separate subsamples.  You may want to save each subsample to a separate object to facilitate your work.  Compute the mean, median and standard deviation for LNWAGE in each subsample and subsample means for the core human capital variables and other wage determinants.  Interpret the mean LNWAGE difference between union-nonuion groups in the context of observed subsample means for other variables.

Obtain regression results for your union-nonunion subsamples. Use the same set of explanatory variables as before with the exclusion of the UNION variable. Note individual coefficient differences of interest and discuss the implications of these differences in coefficients across the union-nonunion subsamples.

Traditional econometric practice would advise you to perform a Chow test at this point to evaluate whether it is appropriate to combine the subsamples. What constraint (imposed by the full sample regression) is relaxed by the subsample regressions?  What do your Chow test results indicate regarding the validity of this constraint?

Based on your findings, and the results of the Blinder-Oaxaca decomposition, what are the primary sources of earnings differences between union and nonunion workers?

ü      further reflection and reporting

Repeat the steps above using FEMALE or IMMIGRANT or MSA or FIRM_LARGE as the basis for subsamples. 

In the context of your selected application, identify inadequate measurement problems, omitted variable concerns and opportunities for improvement.  You are encouraged to use additional variables provided with the CPS97 data to explore these issues further.

Your report should be typewritten with neatly presented diagrams, tables and text.

ü      related references

Ernst Berndt. "Analyzing the Determinants of Wages and Measuring Wage Discrimination: Dummy Variables in Regression Models," in The Practice of Econometrics: Classic and Contemporary, 1991.  Farrel Bloch & Mark Kuskin, "Wage Determination in the Union and Nonunion Sectors," Industrial and Labor Relations Review, 1978. George Borjas, "The Economics of Immigration," Journal of Economic Literature, 1994.  Mary Corcoran & Greg Duncan, "Work History, Labor Force Attachment and Earnings: Differences between the Races and Sexes," Journal of Human Resources, 1979.

 

update 8/29/06