Joseph G. Hirschberg, Esfandiar Maasoumi, and Daniel J. Slottje,
"Clusters of Attributes and Well-Being in the US", Journal of Applied
Econometrics, Vol. 61, No. 3, 2001, pp. 445-460.

There are two data files and one program file. All files are zipped in the
file hmsfiles.zip. These are ASCII file in DOS format.

The two data files are labeled data_l.txt and data_f.txt.  

data_l.txt contains the standardized values of the quality of life
indicators from 1915 to 1995.  These series are defined to have means of
zero and standard deviations equal to one.  They are listed in columns which
are separated by single blanks and are given in the order listed below. Thus
each file is a matrix with 81 rows and 16 columns.  The appendix to the
paper provides the details as to the origin of these series.

Year = 1915 to 1995.
L1= Per Capita GNP in 1982$ 
L2= 1/Infant Mortality Rate Per 1000 Births 
L3= Male Life Expectancy At Birth 
L4= Female Life Expectancy At Birth 
L5= 100 - % Unemployment 
L6= Per Capita Disposable Income in 1982$ 
L7= Physicians Per 100,000 
L8= Surfaced Highway Miles per capita 
L9= % of Households With Telephones 
L10= % of Households With Radios 
L11= 100,000 per Homicide 
L12= % of Ages 5-17 in School 
L13= Newspapers per capita 
L14= Rate of Real GNP Growth (1982$) 
L15= % of GNP not for Defense 

data_f.txt contains the residuals for each of these series from 1915 to 1995
from the appropriate ARIMA model fit to each series; thus it also has 81
rows and 16 columns.  Note that the first observation for this series is set
to a '.' for all the series for which a model in the first differences is
used.  The order of the columns correspond to the series listed above.

The file clust.sas contains a program written in the SAS computer language.
This program will perform the clustering described in the paper using the
two data files.  This routine requires the availability of the IML, SAS
Graph, and SAS Stat modules.  The IML routine is used to compute the distance
matrix and perform the clustering. The Proc Tree routine is then used to
produce the dendrograms published in the paper based on a series of data
sets created in the Proc IML routine.  Because this program was written as a
research tool it is not a general purpose clustering program.  The authors
are not responsible for possible modification of this routine.  All results
in the paper were produced using SAS version 8 for MS Windows.