COR683
May 5, 1999
Creating prestige scores (%5+) for 1990-basis occupation codes

Introduction
------------
In their 1997 paper, Hauser and Warren created an SEI index
for 1990-basis detailed occupation categories.  In doing so,
they adopted a measure of occupational prestige upon which
they regressed occupational earnings and education scores
to obtain the weights used in the construction of their SEI.
In this paper they report:

"Like most earlier investigators, we prefer a measure of the 
percentage of prestige ratings above a fixed threshold as a criterion 
variable, rather than a weighted average of arbitrarily scored
prestige ranks.  For example, Duncan (1961:119-20) chose to use
the percentage of raters who chose "excellent" or "good" on the
five-point scale offered in the 1947 North-Hatt prestige
study, rather than the prestige score...Nakao and Treas (1994:10)
came closer to Duncan's original procedure by analyzing the 
percentage of raters who chose ranks of 5 or higher on the 9-point
scale that was used in the 1989 GSS (as well as in NORC surveys
of the 1960s).  Our analyses are based upon the same prestige 
measure used by Nakao and Treas..."

Thus, rather than use the prestige scores developed by Nakao and
Treas in their 1994 work, Hauser and Warren recommend the
use of the raw percentages of persons rating the occupation a
5 or higher (on a 9-point scale).  Thus, it is *this* prestige
variable which we chose to include in the WLS codebook for
1990-basis occupation codes.  We call this prestige score the
"%5+" score.  (Please see Nakao and Treas 1994 for more detail
on how these data were collected, and how these "%5+" scores
were obtained.)

The "%5+" scores are not published, but rather were obtained 
directly from Nakao and Treas, and are included in a worksheet 
that is part of COR680 (see Column G of this table).  Included 
in COR683 is a SAS module which, given an occupational code as 
input, will return the "%5+" score for that occupation (COR683.SAS).  
Note that "%5+" scores are NOT split by industry and/or class-of-worker
codes.

Recommendations
---------------
In their 1997 paper, Hauser and Warren also made a recommendation
about working with these %5+ scores.  As they write:

"...we have also analyzed a logistic transformation in order to
reduce heteroscedasticity in the residuals from the regression of
prestige on occupational earnings and education..rather than using the 
simple logistic transformation of prestige, we used a started logit of 
the percentage of prestige ratings above the threshold (Mosteller and 
Tukey 1979:109-15):

     	ln[(yi+1)/(100-yi+1)]

This symmetric transformation eliminates the extreme or undefined 
values of the log transform that would otherwise occur when the 
observed percentage is at or near 0 or 100."

(In the above equation, yi=percentage choosing ranks of 5 or higher
for the ith occupation).

When using the "%5+" score as a measure of occupational prestige,
we recommend transforming the score into the started logit before
analysis.


References:

Duncan, Otis Dudley.  1961.  "A Socioeconomic Index for All
Occupations."  Pp. 109-38 in _Occupations and Social Status_,
edited by Albert J. Reiss Jr. New York: Free Press.

Hauser, Robert M. and John Robert Warren.  1997.  "Socioeconomic
Indexes for Occupations:  A Review, Update, and Critique."  Pp.
177-298 in _Sociological Methodology_, edited by Adrian Raftery.  
Cambridge: Blackwell.

Mosteller, Frederick and John W. Tukey.  1979.  _Data Analysis and
Regression:  A Second Course in Statistics."  Reading, MA: Addison-
Wesley Publishing Co.

Nakao, Keiko and Judith Treas.  1994.  "Updating Occupational Prestige
and Socioeconomic Scores:  How the New Measures Measure Up."  Pp. 1-72
in _Sociological Methodology, 1994_, edited by Peter Marsden.  
Washington, D.C.: American Sociological Association.