1988 old format versus 1988 new format or Bridge file ----------------------------------------------------- [Source: email from Eanswythe Grabowski, Unicon.com May 2001] Subject: Re: 88B Hello, In 1989, the way the data was processed, cleaned, and edited was changed. Some variable definitions changed. Some new variables (census recodes) appeared, others disappeared. In an effort to determine how much of the change between 1988 and 1989 was due to the new process rather than to a shift in the population, the Census took the raw 1988 data, processed it using the 1989 techniques and released it as a Bridge file. The sample sizes of the two files are different due to differing eligibility requirements. By comparing the 1988 with the 1988B file you can determine for yourself whether a change in some trend at that time is true, or just a trick of the computer. Now which should you use? If you plan to work with primarily recent files, stick to 88B. If you want to investigate trends in the 70's and 80's, use the 88 file. Finally, because some variables appear in one file and not the other, you may want to check which file gives you more of what you need. For a discussion about the bridge file, refer to Appendix W in the CPS Utilities data dictionary, section 5, in the 'manual' directory on your CD. I hope this clears up any confusion. If not, write me back. /**************************************************************************/ Excerpt from Section 5 March CPS Unicon documentation Contrasting the 1988 and 1988 Bridge files^M [Source: email from Gregory Weyland, Census Bureau July 1997]^M The 1988 "Bridge file" was the result of a complete rewrite of the ^M entire CPS processing system. In addition to doing a complete rewrite ^M of the March Supplement edit, we needed to incorporate the myriad of^M questionnaire changes that were made during the beginning of the 1980's, ^M we made a number of changes to our basic CPS processing system that was ^M over 20 years old at that point. We went back to our basic data ^M acceptance programs and made some revisions in what we would accept and ^M not accept as "good data" into the processing system. This is the cause ^M for the difference in the number of records on the files. While over 99 ^M percent of the records are the same, you will find additions and deletions^M between the files. ^M ^M As I said above, we completely rewrote the supplement edits. Extensive ^M revisions were made to the allocation procedures. For this reason, I ^M wouldn't think that you would want to try and take data from one file ^M and place it on the other file. The two files were to be treated as ^M complete datasets, not as a source of fixes for the original by the ^M bridge file. While as a general statement, it is true that the position ^M numbers are +-40 in most cases, the differences in records between the ^M files do not make such a computation viable. Based on what I said ^M earlier, I can't see why this should be a concern since you don't want ^M to use these files on an individual record basis. ^M