«Der Open-Access-Publikationsserver der ZBW – Leibniz-Informationszentrum Wirtschaft The Open Access Publication Server of the ZBW – Leibniz ...»
One point deserves further discussion. Researchers often face a choice between various measures for the unionisation of a country. In particular a choice between union density and union coverage. The latter gives the share of employees covered by wage agreements and paid accordingly - irrespective of direct union membership. While it is generally diﬃcult to decide in favour of either union density or union coverage to be included in a model, we prefer union density in our setup. We believe union density has stronger implications for the bargaining power of employees. For a union’s clout in labour conﬂicts it is of crucial importance how many actual members it has since this directly determines the ﬁnances available through membership fees and the manpower in strikes and lockouts.5 A general and not surprising result emerging from table 3 is that the Anglo-Saxon countries are characterised by low unionisation relative to the sample median and that the Continental European, in particular the Scandinavian countries have relatively strong union membership shares in total employment. Thus, the following “sample split” exercise can, with some exceptions, also be seen as a crude comparison of diﬀerences in labour share inﬂuences between the English speaking world and Continental Europe.
Turning to the results if we split the sample according to union density (table 4) we Using union coverage to divide the two groups yield a largely similar grouping.
ﬁnd that poolability of slope coeﬃcients for the capital/output (ln(k)) and the ln(T F P ) variable seems not warranted for the countries that are classiﬁed as having high union density. Pooled Mean Group estimates of both slope coeﬃcients are signiﬁcant, but insigniﬁcant Mean Group estimates point to large cross-country variations in those estimates. In contrast, all three estimates for the trade openness coeﬃcient turn out to be signiﬁcantly negative. Furthermore, a Hausman test does not reject the null hypothesis of poolability for this latter variable. In comparison to the full sample estimates, the MPG and MG estimates of trade openness are approximately twice as high in absolute value.
Findings for the low union density countries show some interesting diﬀerences. First, FE, PMG and MG estimates for ln(k) and ln(T F P ) are negative and signiﬁcant. PMG and MG estimates are similar and a Hausman test does not reject poolabilty of the ln(k) and ln(T F P ) coeﬃcients. Secondly, trade openness does not seem to inﬂuence the labor share in the low union density countries since both the PMG and MG coeﬃcients are not signiﬁcantly diﬀerent from zero. This casts doubt on the negative and signiﬁcant static FE estimate.
It is important to discuss the results reported in table 4 in comparison to our ﬁndings for the full sample estimates. Remember that we do not reject the poolability hypothesis for ln(k) based on the full sample estimates shown in table 2, but we do reject the same hypothesis for the high union density countries. Similarly, we ﬁnd poolable coeﬃcients for trade openness in the full sample, but not in the low union density country group.
This discrepancy can be explained twofold: on a more technical point it must be noted that the MG estimator is particularly vulnerable to outliers as it simply computes the average of the country-speciﬁc slope estimates. Consequently, its precision increases with N, the cross-sectional dimension. Given that our sample split estimates are based on rather small subgroups, insigniﬁcant MG estimates may simply reﬂect small sample problems of the MG approach which in turn leads to a rejection of the poolability hypothesis. A second explanation brings us back to one of our main concerns, namely the treatment of cross-sectional heterogeneity. The question whether to pool or not to pool is naturally linked to the sample under consideration. If we allow for more homogeneity within a given sample as we do when countries are pre-grouped by a certain homogeneity criterion (union density), we would regard tests on the poolability of parameters as more powerful than in a sample with relatively heterogeneous countries. In this respect, the ﬁndings on poolability and non-poolability of parameters in the split sample may be preferred over those from the relatively heterogeneous full sample.
The upshots of the analysis that takes the inﬂuence of wage bargaining institutions into account are the following: First, for the countries with low unionisation (or low institutional bargaining power of employees) we ﬁnd a high degree of homogeneity with respect to inﬂuences on the labour share through the capital output ratio and technological progress. Furthermore, the trade channel seems to have no role in aﬀecting the relative compensation of employees in these countries. Secondly, for the countries that are classiﬁed as having powerful employees in the bargaining process, the picture is more mixed. However, there is more evidence that the trade channel is more important and may exert more downward pressure on the labour share than in countries with less inﬂuential employee institutions.
To bring theory and the empirical ﬁndings together, recall our above considerations with respect to how trade aﬀects the labour share. We argued that the trade channel not picked up elsewhere is the possible power shift in the wage bargaining process. Now assume that any given change in trade openness might exert a heterogeneous eﬀect across countries if there are diﬀerent institutional arrangements. Our results point to stronger negative trade eﬀects in countries with relatively many union members. A possible explanation for that ﬁnding may be the following. In countries where the bargaining power of employees is already low, or the wage setting process is simply not characterised by bargaining but by market forces instead, employees do not suﬀer a loss in relative compensation when the bargaining power of the employer increases, for instance due to an increase in a ﬁrm’s outside option. Put simply, if there is no union wage markup, the labour share is not vulnerable to shifts in relative bargaining power. In countries where the labour share was held at a higher level through strong unions, international cost competition and easier access of ﬁrms to the world’s labour supply brought down the labour share.
For the sample characterised by low union density, we also ﬁnd the impact of the capital output ratio and the Total Factor Productivity to be poolable and to exert negative and visible inﬂuences. This may well hint at a stronger inﬂuence of technological change in those countries. Adjustments in relatives shares of factors used in production could be swifter and the eﬀect of capital-augmenting technological change is visible in the labour share’s development. In the other countries, an organised workforce might better be able to shield itself from the forces of technological change. Yet, as argued above, globalisation makes this an ever more diﬃcult task.
The above results emphasize the virtues of caution in the interpretation of the driving forces of the labour share. Even if one cannot technically reject the pooling assumption for empirical models of the labour share, this does not rule out important diﬀerences in the eﬀects across countries or groups of countries.
The initial motivation for this paper was to shed some light on the key driving forces underlying the downward movement in labour shares across a variety of countries. More precisely, it was about assessing whether the explanatory variables exert the same inﬂuences in all countries; we wanted to test the pooling assumption on slope-homogeneity implied by almost all existing studies on the topic. For this purpose, we estimated the determinants of labour share movements with standard ﬁxed eﬀects models as well as in a dynamic heterogeneous panel framework. The latter allowed us to employ estimators which diﬀer in their assumptions on slope-homogeneity and to subsequently compare the results. Based on those PMG and MG estimators we ﬁnd the pooling assumption to be valid for two variables - the capital output ratio and trade openness. This ﬁrst ﬁnding lends important support to the theory on labour share movements along the lines of Bentolila and Saint-Paul (2003). However, as far as other explanatory variables often found in the literature go, the picture is more mixed. Total Factor Productivity, in particular, is found to have heterogeneous slope coeﬃcients across countries and no clear support for the pooling assumption is found.
In order to add more detail to our analysis and to address the role of institutional arrangements, we test for possible clusters among countries characterised by diﬀerently strong unions from which we assume implications for the wage bargaining process. We ﬁnd important diﬀerences in the coeﬀcient values and levels of signiﬁcance. For more market-oriented countries with lower union density, we see the labour share being driven by variables capturing technological change and shifts in the relative usage of factors of production. For countries with strong unions, however, we ﬁnd trade openness to be the most relevant explanatory factor for downward movements of the labour share. We conclude this is due to trade openness reducing the possibilities of unionised employees to secure a wage markup in the distribution of factor incomes.
Given the above results, we conclude that further research as well as scientiﬁc policy advise should take possible slope-heterogeneity and institutional arrangements into account when estimating models describing the labour share.
References Arpaia, A., Pérez, E., and Pichelmann, K. (2009). Understanding labour income share dynamics in Europe. European Commission Economic Papers, (397).
Atkinson, A. (2009). Factor shares: the principal problem of political economy? Oxford Review of Economic Policy, 25(1):3–16.
Bentolila, S. and Saint-Paul, G. (2003). Explaining movements in the labor share.
Contributions to Macroeconomics, 3(1).
Binder, M. and Bröck, S. (2006). On the relation between investment and economic growth: New cross-country empirical evidence. Mimeo, Goethe University Frankfurt.
Blanchard, O. and Giavazzi, F. (2003). Macroeconomic Eﬀects of Regulation and Deregulation in Goods and Labor Markets. Quarterly Journal of Economics, 118(3):879–907.
Buch, C. M., Monti, P., and Toubal, F. (2008). Trade’s impact on the labor share:
Evidence from german and italian regions. IAW Discussion Papers 46, Institut für Angewandte Wirtschaftsforschung (IAW).
Checchi, D. and Garcia-Peñalosa, C. (2008). Labour market institutions and income inequality. Economic Policy, 23(56):601–649.
Gollin, D. (2002). Getting income shares right. Journal of Political Economy, 110(2):458– 474.
Guscina, A. (2006). Eﬀects of globalization on labor’s share in national income. IMF Working Paper, 06/294.
Harrison, A. (2002). Has Globalization Eroded Labor’s Share? Evidence from CrossCountry Data. manuscript UC Berkeley.
Hutchinson, J. and Persyn, D. (2009). Globalisation, concentration and footloose ﬁrms:
in search of the main cause of the declining labour share. LICOS Discussion Paper Series, (229/2009).
Jaumotte, F. and Tytell, I. (2007). How Has the Globalization of Labor Aﬀected the Labor Income Share in Advanced Countries? IMF Working Paper, 07/298.
Pesaran, M., Shin, Y., and Smith, R. (1999). Pooled Mean Group Estimation of Dynamic Heterogeneous Panels. Journal of the American Statistical Association, 94(446):621– 622.
Pesaran, M. H. (2006). Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica, 74(4):967–1012.
Pesaran, M. H. and Smith, R. (1995). Estimating long-run relationships from dynamic heterogeneous panels. Journal of Econometrics, 68(1):79–113.
Visser, J. (2009). The ICTWSS database: Database on institutional characteristics of trade unions, wage setting, state intervention and social pacts in 34 countries between 1960 and 2007.