function of institutional variables. As a consequence, there is no recognized general mea-
sure of flexibility with sufficient time and country coverage available. In a very recent
paper, Holden and Wulfsberg (2009) provide a very simple measure of downward real
wage rigidity for 19 OECD countries from 1973-1999. Unfortunately the country cover-
age differs from our sample and the data is not balanced. However, in their analysis, they
conclude that “real wage cuts are less prevalent in countries with strict employment pro-
tection legislation and high union density” (p. 605). In the light of this finding, we use an
index of union density (adjusted for the degree of corporatism) as a proxy for the rigidity
of real wages.41 Our theoretical model suggests that foreign labor market institutions
should affect domestic unemployment more when domestic real wages are rigid. Hence,
we interact the rigidity proxy rigidi with the foreign wage wedge b* and expect a positive
sign. Column (1) in Table 3 includes this interaction term into an unemployment regres-
sion of the type (31). The rigidity index itself has a positive but statistically insignificant
coefficient. Thus, the effect of wage rigidity on unemployment is not significant. However,
the interaction with the foreign wage wedge comes with positive sign and high statisti-
cal significance, indicating that the spill-over effect of labor market institutions is higher
when wages are more rigid. Adding additional labor market institutions to the regression
(column (2)) does not change the picture.
Second, we discuss the interaction between country size and the wage distortion. We
measure country size by population, just as in our theoretical analysis. This variable has
the advantage that it is exogenous. The logic is that the larger the domestic economy is,
the more strongly should it be negatively affected by bad domestic labor market institu-
tions and the less by foreign ones. Conversely, the larger the foreign economy is (weighted
by bilateral trade potentials), the more strongly should foreign distortions increase the
domestic unemployment rate while domestic distortions should be less important. Hence,
we expect that the coefficients on ln (pop) × b, ln (pop) × b*, ln (pop*) × b, and ln (pop*) × b*
should be positive, negative, negative and positive, respectively. Column (3) in Table 3 is
nicely in line with this sign pattern. However, statistical precision is not very high, most
likely due to the large degree of correlation between those interaction terms. Including the
degree of product market regulation into the regression (column (4)) does not alter the
sign of significant coefficients or their magnitudes and only partially improves statistical
accuracy. Column (5) focuses on statistically significant effects only. In line with our
theory, distortions are more harmful when they have their origins in large countries. In-
terestingly, the direct effect of the own and the foreign wage distortions is now negative.
There is also fairly strong evidence that - everything else equal - large countries have
smaller unemployment rates. This is also in line with the theoretical model, where larger
home markets are associated with fiercer competition, more varieties, and hence higher
productivity of the average firm and, consequently, with lower unemployment.
Third, we discuss the interaction between entry regulation42 and the wage wedge.
41The proxy is rigidi = union.densityi × low-corporatismi.
42 The measure of domestic entry regulation (pmr) provided by the OECD strongly correlates with
other openness measures (e.g., the share of trade over GDP), but has the advantage that it is unrelated
to geography and size.
40