Government Partisanship, Labor Organization, and Wage Inequality

...

2) the stronger labor organization, the lower wage inequality.

Since non-market institutional forces tend to dampen inequality, wage inequality ought to be lower in countries that features strong labor organization, while higher in countries that have weak labor organization.

4. Empirical Measures, Method, and Analysis.

To test the hypotheses just outlined above, this paper uses time-series cross-section data drawn from the twelve OECD countries[2] from 1970 to 1992 (276 observations). According to Beck and Katz (1995) who show that the generalized least squares approach of Parks produces standard errors that lead to extreme overconfidence, this paper employs OLS parameter estimates but replace the OLS standard errors with panel-corrected standard errors (PCSEs). This paper tests effects of government partisanship and labor organization on wage inequality, controlling for budget spending, economic globalization, some macroeconomic indicators, country-fixed effects, and time-specific effects. The models are as follows:

Theil it = β0+ β1 Theilit-1 + β2 LEFTit-1 + β3UNIONit-1 + β4CONTRYit + β5 YEAR it + u it (1)

Theil it = β0 + β1 Theil it-1 + β2 LEFT it-1 + β3 UNION it-1 + β4 DEF it-1+ β5 CONTRY it

+ β6 YEAR it + u it (2)

Theil it = β0 + β1 Theil it-1 + β2 LEFT it-1 + β3 UNION it-1 + β4 DEF it-1 + β5 TRADE it-1

+ β6 LOWWAGE it-1 + β7 FDI it-1 + β8 CAPLIB it-1+ β9 CONTRY it + β10 YEAR it

+ u it (3)

Theil it = β0 + β1 Theil it-1 + β2 LEFT it-1 + β3 UNION it-1 + β4 DEF it-1+ β5 INFLA it-1

+ β6 UNEMP it-1 + β7 GDPPC it-1 + β8 CONTRY it + β9 YEAR it + u it (4)

Theil it = β0 + β1 Theil it-1 + β2 LEFT it-1 + β3 UNION it-1 + β4 DEF it-1 + β5 TRADE it-1

+ β6 LOWWAGE it-1 + β7 FDI it-1 + β8 CAPLIB it-1 + β9 INFLA it-1 + β10 UNEMPit-1 + β11 GDPPC it-1 + β12 CONTRY it + β13 YEAR it + u it (5)

Theil it = β0 + β1 Theilit-1 + β2 INFLAit-1 + β3 UNEMPit-1 + β4 GDPPCit-1 + β5 CONTRYit

+ β6 YEAR it + u it (6)

Theil it = β0 + β1 Theilit-1 + β2 TRADEit-1 + β3 LOWWAGEit-1 + β4 FDIit-1 + β5 CAPLIBit-1

+ β6 INFLAit-1 + β7 UNEMPit-1 + β8 GDPPCit-1 + β9 CONTRYit + β10 YEARit

+ u it (7)

I use Theil index, rather than Gini coefficients, as a measure of wage inequality. The Gini coefficients are flawed due to a lack of consistent data across time and country, and indirect measure of the wage differentials[3]. If all incomes are equal, Theil’s T measure of equality is zero. As dispersion grows about the mean, T increases. It has a theoretical maximum of one given by the logarithm of the number of people in the population; that value is reached when one person has all the income. Annual Theil index is computed from the Industrial Statistics Data Base of the United Nations International Development Organization (UNIDO)[4]. Meanwhile, Theil at t-1 is for controlling for intertemporal correlations of wage inequality.

Government partisanship is operationalized as leftist share of seats in parliament (LEFT)[5]. To ensure the causal arrows flow as we expect, the lagged value for partisanship variable is used under the assumption that levels of wage inequality at time t must be caused by the degree of left partisanship at time t-1, i.e., the preceding year. The assumption applies to all other lagged independent variables. The degree of labor organization is measured by adjusted union density (UNION), i.e., adjusted union membership (total membership less self-employed, retried and unemployed) weighted by total dependent labor force (Ebbinghaus and Visser 2000a, 2000b)[6].

Government fiscal policies often affect macroeconomic situation. For example, leftist government may adopt deficit budget spending in favor of social welfare. So, the regression model includes level of budget balance (DEF) as a control variable, measured by tax revenue less government consumption (OECD, 1997).

Assuming that globalization effects on wage inequality are heterogeneous and uneven, I include in the model various indicators of integration into international markets (World Bank 1997). Trade (TRADE) is measured by import plus export as a share of GDP; LOWWAGE is imports from low-wage countries as a percentage of total import; FDI is measured by FDI inflow plus outflow as percentage of GDP; as for capital liberalization (CAPLIB) I used Quinn’s index of international financial openness raging from zero to fourteen with higher scores denoting more openness (Quinn 1997).

Introducing country-fixed effects is particularly important when analyzing the government partisan effects because the partisanship tend to vary more across countries than over time within them. The model also includes year dummy variables to take into account time-specific effects. If all OECD countries are subject to common shocks such as Oil Shock, this should be considered through a series of year dummy lest these shocks contaminate the regressors of direct interest.

5. Results.

The results of seven time-series cross-section analyses are reported in table 1. The batteries of country and year dummy variables are significant, and this demonstrates the importance of controlling for both unit and time fixed effects in analyses. An interesting thing is that country dummy is significant in models 1 to 5 that include government partisanship and labor organization, while year dummy is significant in models 5 and 6 which exclude those variables. This suggests that the partisanship and union density is a country-fixed historical tradition. In contrast, more year-specific than country-fixed variables are globalization variables (TRADE, LOWWAGE, FDI, and CAPLIB) and macroeconomic situation variables (INFLA, UNEMP, and GDPPC).

As expected from hypothesis 1 and 2, leftist partisanship (LEEF) and labor organization (UNION) have consistently negatively coefficients; they tend to dampen wage inequality across time and country.

Table 1 here

Figure 1 and 2 also illustrate that the leftist partisanship and