Overall Segregation indices with optional segregation curve for the case of two > groups (For multigroup cases see localseg)
dicseg unitvar groupvar [weights] [if exp] [in range] [, format(string) save sc nograph x(string) y(string) xtitle(string) ytitle(string) graph_options(string) ]
fweights, aweights and aweights are allowed; see weights.
It requires to have matsort, written by Paul Millar, previously installed; if not, install it writing in the command line:
. net install matsort, all from(http://fmwww.bc.edu/RePEc/bocode/m)
dicseg computes segregation indices across units in a two-groups context.
Micro data are required (data should not be collapsed by units).
The first variable unitvar indicates units (ex. occupations, census tracts, schools, ....).
The second variable groupvar identifies dichotomous groups such as gender (male vs. female), race (white versus non-white), ...
If option sc is specified it also draws the segregation curve, and creates new variables using x(string) and y(string) options (you can also use other graph options).
x represents the group with the lowest value in groupvar and y the group with the highest value.
Segregation indices (See Gradín, 2010):
. Duncan & Duncan, the Duncan and Duncan (1955) index of dissimilarity.
. KM, the Karmel and Maclachlan (1988) index.
. GE(c) with c=-1, 0, .10, .25, .50, .75, .90, 1, 2 the family of Generalized Entropy measures for different values of the segregation sensitivity parameter. GE(1) is Theil index.
Note: GE(c) is not defined for some c when one group has no observations at one or more units. In those cases only well-defined indices are reported.
. Hutchens Sq. root, the Hutchens square root index (Hutchens, 1991, 2004). Note that it is equivalent to GE(.5)/4
. Mutual Information, (Frankel and Volij, 2007, 2008) with natural logs instead of log(2).
. Gini, the Gini index.
Note that Duncan, Gini, and Hutchens are bounded between 0, when there is no segregation because both groups have the same distribution across units, and 1, when segregation is at its maximum because there is no overlap between both distributions.
Segregation Curve (if option sc is specified): as defined in Duncan and Duncan (1955) or Hutchens (1991).
format(string) to change numeric format
save to save results
sc to compute segregation curves
nograph to do not graph segregation curves (if sc was specified)
x(string) y(string) to create variables for segregation curves (if sc was specified)
xtitle(string),ytitle(string) title for segregation curves (if sc was specified)
graph_options(string) to change default graph options for segregation curve (if sc was specified)
r(seg) : segregation indices
. use segdata.dta, clear
. dicseg occupation white [aw=pwgtp], sc save
. dicseg occupation2 sex [aw=pwgtp], sc save
. dicseg occupation white [aw=pwgtp], save sc x(cumnonwhite) y(cumwhite) xt(Cumulative proportion of nonwhites) yt(Cumulative proportion of whites)
. dicseg occupation sex [aw=pwgtp], save sc x(cummale) y(cumfemale) xt(Cumulative proportion of males) yt(Cumulative proportion of females)
Carlos Gradín <firstname.lastname@example.org> Facultade de CC. Económicas Universidade de Vigo 36310 Vigo, Galicia, Spain.
Duncan, Otis D. and Duncan, Bervely (1955), A Methodological Analysis of Segregation Indexes, American Sociological Review, 20(2): 210-217.
Gradín, Carlos (2010), Conditional occupational segregation of minorities in the U.S., Ecineq Working Paper Series 2010-185.
Hutchens, Robert M. (1991), “Segregation Curves, Lorenz Curves, and Inequality in the Distribution of People across Occupations”, Mathematical Social Sciences, 21: 31- 51.
Hutchens, Robert M. (2004), “One Measure of Segregation”, International Economic Review 45(2): 555-578.
Karmel, T. and Maclachlan M. (1988), Occupational Sex Segregation - Increasing or Decreasing, Economic Record, 64, 187-195.
duncan if installed; hutchens if installed; localseg if installed; seg if installed