Do labor market institutions matter for business cycles?∗ - European ...

3 CROSS SECTIONAL EVIDENCE 8
3.2 Collapsing information
Since we are not exploiting the time series dimension of the data, we have 19 data points, 13
LMI indicators and 8 macroeconomic indicators making the number of degrees of freedom
prohibitively small. For this reason, we collapse the information contained in the institutional
and macroeconomic variables into interpretable factors. Using factors, rather than
the original variables saves degrees of freedom and allows us to take into account possible
interactions among institutions, reducing the potential bias due to omitted variables. For the
institutional data we select as many factors as needed to capture always at least the eighty
percent of the variability of the original data. With this criterion we find that four principal
components describe fairly well the sample of indicators 7 . Also to control for endogeneity of
LMI we construct principal components using the initial condition of institutions measured
as the average value of institutions in the sample between 1960 and 1970.
Table 2 displays the correlation of the four factors with the original variables. The first
factor, explaining 38% of the variance, is highly correlated with all labor market indicators,
thus it can be interpreted as an index of overall rigidity. The second factor, explaining
22% of the variance, isolates the role of unions in an environment where employment protection
is weak. The factor is negatively correlated with the employment protection indices and
highly and positively correlated with union density, concentration and centralization and
the minimum wage index. We label this factor as the union factor. Note that the union
factor does not capture collective wage bargaining directly - the correlation between the
index and union coverage or the level at which bargaining takes place is negative. Rather,
it captures other activities that may indirectly affect wages and employment flows such as
the power to organize strikes or the representation of workers in the organization of the
production process. The third factor, explaining 11% of the variance, is labelled the wage
setting factor, as it positively correlates with variables extension, EXT , coverage, UC,
concentration, CONC, government intervention, GOV INT and minimum wage legislations
while it negatively correlates with coverage and replacement rates. The wage setting factor
captures the role of unions in wage bargaining activities. Finally, the last factor, explaining
almost 10% of the variance, strongly correlates with EPL in short term contacts, and the
other EPL measures and replacement rates. Hence, this component may be interpreted as
residually identifying a flow restrictions factor.
3.3 Cross country evidence
3.3.1 Statistical factors
Table 3 presents Spearman partial rank correlations between the business cycle statistics and
alternative LMI factors, when we condition on macroeconomic factors. These correlations,
7 The principal components are extracted from non-orthogonal and standarized data.