Buhong Zheng, "Poverty Comparisons with Dependent Samples", Journal of
Applied Econometrics, Vol. 19, Vol. 3, 2004, pp. 419-428.

The file poverty.zip contains 10 CPS data sets from 1990 to 1999.

1. These data are extracted from the official US Current Population Surveys
   (Annual Demographic and Income Supplement).  The CPS data can be accessed
   from the official web site at <www.bls.census.gov/cps> or from the CPS
   Utilities CD-ROM by the Unicon Research Corporation <www.unicon.com>

2. Each year's data contains 5 variables, which are spread over five lines:
   MIS -- month in sample; 
   FAMINC -- total family income; 
   HHID -- household id number;
   FMIPA -- The amount of public assistance and welfare payment;
   Fnumber -- Number of persons in family.

3. Each data set contains around 155,000 observations.

3. Data are all ASCII format and start from the first line.

4. The variables MIS and HHID are used to create the overlap portion of the
   data between two consecutive years.