Lutz Hendricks - UNC - Department of
Economics
# CPS Earnings Statistics

Purpose: Get students started with data handling.

Goal: Using Current Population Survey data (from IPUMS), construct measures of wage dispersion over time.

## Getting the data

I downloaded the data for you (from IPUMS).

A compressed file can be downloaded from here. It contains:

- The data in
`Stata`

format (extension `dta`

).
- Codebooks (extension
`cbk.txt`

or `xml`

).
- Program files for reading the data using various stats packages (extensions
`sps`

, `sas`

, `do`

).

## Your assignment

For each year:

Construct measures of wage dispersion: standard deviation of log wages; 90/10, 90/50 and 50/10 ratios.

- also break this down by schooling (high school dropouts, high school graduates, college dropouts, college graduates)
- graph these results

Estimate a Mincer equation using OLS. This involves regressing log wages on schooling dummies and a quartic in experience.

Universe: Men, aged 30–60, who work full time (at least 30 weeks per year) for wages.

Wage = earnings / weeks worked.

## Things to think about

The highest degree earned is not available in all years.

How to assign people to school groups?

The standard answer: years of schooling < 12: high school dropout, = 12: high school graduate, 13–15: college dropout, 16+ college graduate.

The coding of weeks worked changes over time.

Earnings are top coded (you don’t have to deal with this complication for this exercise).