table 5. offshoring-related displacement
rates for it and non-it workers.
non-it Workers Displaced
not Displaced 5,704 712
Displaced 227 61
A chi-squared test of the hypothesis of equality
between displacement averages is rejected at
the p<.01 level X2( 1)= 27. 5, p<.01.
can provide personal services directly
to overseas customers.
Table 7 reports the results of the
primary regressions from the survey
data relating the personal services
provided in one’s occupation to the
likelihood of offshoring-related displacement. The results in column
( 1) support the hypothesis that employment in a job providing personal
services significantly decreases the
likelihood of being displaced due to
offshoring (t= 3. 5). Somewhat surprisingly, the coefficient estimate on
salary level is negative and significant, suggesting workers with higher
salaries are less likely to be offshored
(t= 2.09). However, in the absence of
human-capital data, the salary term in
the regressions also reflects human-capital variables (such as education
and experience). We therefore interpret the negative coefficient as indicating that, conditional on job level,
workers with more human capital are
less likely to be offshored, an effect
that dominates any direct gains from
offshoring more expensive workers.
The results also suggest that older
workers (t= 4. 9), males (t= 2. 21), and
workers in simpler jobs (t= 2. 15) requiring less firm-specific capital are
likely to be offshored.
After including industry dummies
in column ( 2), the coefficient esti-
mate on gender is no longer signifi-
cant, indicating our earlier estimate
on gender may have reflected high off-
shoring intensity in such industries
as IT with a higher fraction of men
and low offshoring intensity in such
industries as health care with a high-
er fraction of women. However, the
estimates on the other coefficients
remain significant. Dummy variables
for state and race in column ( 3) do not
significantly alter the coefficient esti-
mates on any other variable.
Although about 15% of firms in the
U.S. offshored in 2007, firms in high-
table 6. Probit analysis, employer offshoring.
( 2) ( 3)
number of employees .006 .006
Local cost of doing business –.001 –.001
Geographic expansion 1. 55
Geographic expansion Impersonal –.015
controls Industry Industry Industry
Pseudo-r2 . 11.07 .09
n 26,568 4,041 4,041
Huber-White standard errors are clustered on firm and shown in parentheses. p<.01
employer offshores job type
a composite skill index taken from blinder. 4 Higher values indicate that less face-to-face contact or
physical presence are needed for job.
column ( 1) is a probit regression of employer and job characteristics against the likelihood the employer
offshores a particular type of job.
column ( 2) is a probit regression of employer and job characteristics against the likelihood the employer
offshores a particular type of job, only for employers are offshoring.
column ( 3) is similar to column ( 2) but includes variables related to employer expansion plans.