Computer Operations Request #344a

Date:	August, 1983

Purpose:  Update variables SPL100, XSPL100, and create RSPL100. These variables
          contain "Size of Place of Residence in Hundreds" corresponding to
          the geography codes of GEOG75, XGEOG77, and  GEOG70 respectively.
          (SPL100 was originally created in COR #337 for Wisconsin residents
          only.)

          The codebooks, XTRACT dictionaries, index and alpha list of variables
          will be updated to reflect the following:

          SPL100    448 -  452 (based upon GEOG75 - updated to include all U.S.)

          RSPL100   460 -  464  (based upon GEOG70)

          XSPL100  2559 - 2563  (based upon XGEOG77)

DESCRIPTION OF THE STEPS.

   The first step was to extract population counts and geography codes from
the 1970 Census, First Count, File B. The layout of this file is explained in 
DUALabs Technical Document ST-1 (CDE library number 198). The computer program 
DECOMPRESS writes 3600 character (place) records for the selected areas 
(States), in this case for all states. From these records we need only the 
following fields

   1. 1970 State 2-3

   2. Place  19-22

   3. Count of all persons 121-128

The Geography Codes are more fully explained in "Geographic Identification 
Code Scheme - PHC (R)," 1970 Census of population and housing. They consist of 
a two-digit State Code, a three-digit County code, and a four-digit Place code.
For the present purpose, only the State and Place codes matter, since a place 
code is unique within a State. Since a Place may be spread over more than one 
county, the population counts should be aggregated over counties. This 
database is available in compressed form on tapes DE1159, DE1160, DE1161, 
DE1162, and DE1163. The DECOMPRESS utility is used to retrieve the information 
from the tapes and to extract the "Summary Type 3 records: Place total summary
records (includes towns in New England)." A total of 23,208 place records was 
written on DE3685. (Stored in CDE as a SWH Master.)

   The next step was to write the population counts in hundreds in the relevant
fields of the Wisconsin Data Tape. The relevant Census information (State, 
Place, Population count) was first aggregated over Counties and stored in a 
large table that allowed fast look-up of population counts for a given 
geography code (in the technical literature known as a hash-table or 
scatter-table; see N. Wirth, Algorithms and data structures = programs, 
Prentice-Hall). Then, for each case in the Wisconsin data, the geography codes 
in GEOG75, XGEOG77, and GEOG70 were used to look these population counts up in 
the table and write them in SPL100, XSPL100, and RSPL100 respectively.
The latter variables were set to all nines if a given geography code could not 
been found. This turned out to be the case for places with a missing place 
code and for missing geography codes. Preliminary tests also revealed that New 
England towns were, countrary to documentation, NOT included in the Census 
extract. Since only 64 New England towns were referred to in the Wisconsin Data,
these were entered by hand using "Characteristics of the Population, Vol. 1, 
Part A: Number of Inhabitants, Table 31: Land Area and Population of Places of 
2,500 or more for the U.S. and of Towns of 2,500 or more for the New England 
States, column 1, 1970 Population." The program entered these 64 cases just as 
if they were Census records in the large table.

   The updated Wisconsin Data were written on tape DE3686 and some checks were 
made. If the code was not found and if the PLACE was zero or 9999 or STATE 
greater than 56 no diagnostic was printed. Place codes were found for 
pseudo-State-codes 66, 73, 91, 93, 94, 95, 96 but these are non-US state codes 
and thus no SPL100 value was written.