Skill-biased technological change and endogenous benefits: the dynamics of unemployment and wage inequality

Abstract

In this article, we study the effect of skill-biased technological change on unemployment and wage inequality in the presence of a link between social benefits and average income. In this case, an increase in the productivity of skilled workers, and hence their wage, leads to an increase in average income and hence in benefits. The increased fallback income, in turn, makes unskilled workers ask for higher wages. As higher wages are not justified by corresponding productivity increases, unemployment rises. Generally, we show that skill-biased technological change leads to increasing unemployment of the unskilled and to a moderately increasing wage inequality when benefits are endogenous. The model provides a theoretical explanation for diverging dynamics in wage inequality and unemployment under different social benefits regimes. Analysing the social legislation in 14 countries, we find that benefits are linked to the evolution of average income in Continental Europe but not in the US and the UK. Given this institutional difference, our model predicts that skill-biased technological change leads to rising unemployment in Continental Europe and rising wage inequality in the US and the UK.

Acknowledgements

We are grateful to Philippe Aghion, Nicole Gürtzgen, Martin Hellwig, Melanie Lührmann, Lia Pacelli and Thomas Zwick for valuable comments and suggestions and to Jenny Meyer and Iliyan Stankov for excellent research assistance. The usual disclaimer applies.

Notes

¹ See, e.g. Gottschalk and Smeeding (1997), Katz and Autor (1999), Acemoglu (2003) and Green et al. (2003). Other factors affecting the relative demand for skills that have been identified in the literature are organizational changes (e.g. Lindbeck and Snower, 1996; Falk, 2002) and globalization of goods and labour markets (Borjas et al., 1996; Fenstra and Hanson, 1996; Tombazos, 1999; Baldwin and Cain, 2000).

² See, e.g. Krugman (1995), Katz and Autor (1999), Blau and Kahn (1996) and Acemoglu (2002). Few authors argue that both the stylized differences between the US and Europe and the labour market rigidities used to explain these differences are overstated in the above literature (e.g. Gregg and Manning, 1997; Nickell, 1997; Nickell et al., 2005).

³ Krugman (1995) is such an exception. He uses the ad hoc assumption that due to labour market institutions, the wage for unskilled labour is proportional to the wage for skilled labour.

⁴ Admittedly, it has been noted in the literature that institutions have changed in reaction to shocks, such as skill-biased technical change. There has been a tendency towards deunionization in the US and UK while, at least in the beginning of the 1980s in Continental Europe employment protection was strengthened and benefits of the unemployment insurance have become more generous (Blanchard, 2006).

⁵ See, e.g. Foster (1998): ‘Absolute versus Relative Poverty’ and the other contributions to the session on ‘What is Poverty and Who are the Poor?’ in the American Economic Association (AEA) articles and Proceedings Issue of the American Economic Review of May 1998.

⁶ In fact, most benefit systems have unemployment insurance elements that depend on the level of past earnings rather than the general income level of the economy. Note, however, the following: first, unemployment insurance benefits are generally limited in duration. So, in the long run, it is social benefits that constitute the fallback income. Second, for unskilled workers, unemployment insurance benefits may easily fall short of the level of social benefits. In this case, the payment is increased to this level.

⁷ Gürtzgen (forthcoming) shows that the impact of unions on wages is largest at the lower end of the skill distribution.

⁸ See, e.g. Siebert (1997), Katz and Autor (1999) or Acemoglu (2002). For Germany, a detailed analysis of the employment and wage development from 1975 to 1990 is performed in Fitzenberger (1999).

⁹ We are aware that the view that increasing unemployment in Continental Europe and increasing wage inequality in the US and the UK are two sides of the same coin (namely, skill-biased technological change) is not beyond controversy (e.g. Nickell and Bell, 1996; Gregg and Manning, 1997). Yet, there seems to have emerged a large consensus among many economists that this view explains at least parts of the intercontinental differences (e.g. Cahuc and Zylberberg, 2004, Chap. 10).

¹⁰ There have been several attempts in the EU to harmonize social legislation–without much success, though. Two of the more successful attempts have led to the European Social Charter of 1989 and to the social protocol annexed to the Maastricht Treaty of 1992–both not signed by the UK. If compared to other policy areas of the European Community, the treaties on social standards remain vague.

¹¹ There were no additional discretionary increases between 1979 and 2001 (Cantillon et al., 2004).

¹² For a concise overview, see Uccello and Gallagher (1997) from where most of the following information is taken.

¹³ This assumption is also for simplicity. In Weiss and Garloff (2005), we consider the case where the union takes into full account the effects of the wage level on the level of benefits and the tax rate. The results are virtually unaffected.14

¹⁴ See, e.g. Cahuc and Zylberberg (2004, Chap. 7).

¹⁵ Equations 3 and 4 are two different ways of writing down the same result. In Equation 3, the focus is on the dependency of the unskilled workers’ wage on the level of (endogenous) benefits, while in Equation 4 the unskilled workers’ wage is shown as a function of the exogenous parameters of the model.

¹⁶ Throughout this article, the term ‘equilibrium’ is used to refer to the allocation which results from union wage setting, given the other institutional features of the model.

¹⁷ Existence: for w_u sufficiently small (respectively, sufficiently large), the right-hand side of the equilibrium condition 4 is larger (respectively smaller) than the left-hand side. As both sides of the equation are continuous in w_u , there must exist at least one value of w_u , , for which both sides are equal. Uniqueness: the left-hand side of condition 4 is strictly increasing in w_u whereas the right-hand side is strictly decreasing in w_u . Therefore, if a solution to condition 4, , exists, it must be unique.

¹⁸ It is easily verified that (∂(w _s /w _u )*/∂(a _s /a _u ))·((a _s /a _u )/(w _s /w _u )*) = ρ·((w _s /w _u )/(a _s /a _u ))^{ρ/( 1 − ρ )}/(1 − ρ + ((w _s /w _u )/(a _s /a _u ))^{ρ/( 1 − ρ )}) is unambiguously smaller than ρ, the respective elasticity that prevailed if wages were perfectly flexible.

¹⁹ See, e.g. Fenstra and Hanson (1996) and Baldwin and Cain (2000).

²⁰ In several further extensions of the model, we assess the robustness of our results and show that the central results are kept. These robustness checks are performed with respect to the bargaining regime, the risk neutrality of workers, and the production technology (compare the working paper version of this article with Weiss and Garloff (2005)).

²¹ We use social expenditures on unemployment, since a category for benefits alone does not exist so that this category comes closest to our needs. Using in addition social expenditures on housing and incapacity-related benefits does not change the principal conclusions. Results are available upon request.

²² It is likely that both real GDP per capita and real social expenditures per unemployed are trended. We use first differences in order not to run into the problem of a spurious regression.

²³ Austria, Italy, the Netherlands and Portugal are the exceptions to the rule.

²⁴ See Weiss and Garloff (2005).

²⁵ Unemployment rates by qualification are from Reinberg and Hummel (2005). The skill premium has been calculated from administrative data where the wage for unskilled workers is proxied by the lowest performance group of blue-collar workers (‘Leistungsgruppe 3, Arbeiter’) and the wage for skilled workers is proxied by the highest performance group of white-collar workers (‘Leistungsgruppe 2, Angestellte’), Source: German Federal Statistical Office.

²⁶ The finding that higher wage dispersion and unemployment are positively correlated (across age-by-education cells) in Germany is also found and discussed by Fitzenberger and Garloff (2008).

²⁷ Unemployment rates by educational attainment stem from the US Census Bureau (1975–2004) and Francesconi et al. (2000). Wages stem from the CPS (US Census Bureau, internet release, www.census.gov/hhes/www/income/histinc/incpertoc.html). Unemployment of the unskilled and wage dispersion are unlikely to contain a trend. While this is obvious for the unemployment rate at least in an asymptotic sense, it is plausible for wage dispersion, too. Nevertheless, we performed the same regression in first differences. Albeit less strongly, the results of a regression of the unemployment rate of the low-skilled on the wage premium in levels is significant for Germany (and insignificant for the US).