Differences Between the March 1995 and March 1996 Annual Demographic Files

A. Geography: The substate geography identified now reflects the June 30, 1993 metropolitan area definitions. The migration data were also revised to reflect these definitions.

B. Matching: Due to the difference listed above, the March 1996 file is not matchable to the 1995 file. It will be matchable to the 1997 file.

C. Revised Earnings Topcoding: The topcoding of the last year's earnings variables was revised this year. Instead of topcoding them at 99,999 as was the case, the four variables used the following topcodes:

Variables Topcode Mean Earnings

a. ERN-VAL $150,000 $576,372

b. WS-VAL 25,000 183,748

c. SE-VAL 40,000 154,528

d. FRM-VAL 25,000 53,067

Records that were topcoded will have a value greater than the topcode value representing the mean earnings for topcoded individuals with similar characteristics. Aggregate income totals from the public use file should now match the data published by the Census Bureau.

D. One Year Ago Migration Data: These data returned to the file this year.

E. Health Insurance Items: Person's records on this year's file contain the edited data from the new items introduced last year. The health insurance information with the same names as 1995 has been recoded from the new items to replicate 1995 and earlier data. The household record once again contain household health insurance status recodes. While these were created from the new health insurance items, they are as consistent as possible with the 1994 and earlier recodes.

F. Household Identification Number
(H-IDNUM): Due to the phase-in of the 1990 Census based design and the introduction of the June 30, 1993 metropolitan definitions the size of the household identification number has changed. It is now 15 characters long.

G. Sample Reduction: Beginning January 1996, the CPS national sample was reduced from 56,000 to 50,000 eligible housing units. This reduction was necessary because of budget cuts and affected seven states (Illinois, Massachusetts, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania) and two substate area (Los Angeles and New York City). This reduction also decreased the size of the November supplemental Hispanic sample that is added to the March ADF. The number of households on the March 1996 file is 63,339 compared to 72,152 or 1995. This reduction had a minimal effect on the accuracy of National estimates. It had a larger effect on the accuracy of state estimates in those states with reductions.

H. Revised Race Edit and Allocation: In January 1996, the Bureau of the Census revised its procedures for editing and allocating the race variable. This change involved the allocation of all "Other" responses for race into one of the 4 main race categories (White, Black, American Indian, and Asian-Pacific Islander). This revision was made to offset a major increase in "Other" responses which was causing severe underestimates of the American Indian and Asian Pacific Islander populations. The result, as shown below was an increase in the estimates for these two racial groups that brought them more into line with independent estimates. Also, there will be some minor disruptions in other statistics. Due to this, one should use caution when making comparisons between the 1995 and 1996 data, especially for those identifying all four race groups.

Annual Demographic Survey (March 1996 CPS) Public Use File Page

CPS Home Page