Why Have Labour Market Outcomes of Youth in Advanced Economies Deteriorated?

This is a proxy for the relative present value of future wage streams from schooling – the relevant factor here – due to lack of data.

For the derivation see Layard (1982, endnote 41).

Weak but still present, e.g. because differentials in labour outcomes between skill-groups may reduce the strength of collective bargaining or lower union influence on wage-setting. Estimating the model with and without labour market institutions will shed some light on this issue.

Acemoglu (2003) explores the former possibility by linking the incentives to adopt new technologies to the degree of compression in the wage structure and, consequently, to labour market institutions. He argues that as institutional wage compression in Europe forces firms to pay higher wages to unskilled workers than in less regulated labour markets, there is a higher incentive for these firms to increase the productivity of the unskilled and, therefore, to adopt technologies complementary to them. However, under this hypothesis, technology should be less skilled-biased in Europe than in countries with more liberal pay setting structures, a prediction that does not find support in empirical evidence (e.g. see Goos, Manning, and Salomons 2009).

Excluding females can be partially justified on the assumption of strict segregation between male and female labour markets (see Bergman’s 1971, pioneering work and subsequent studies). The reasons for segregation stretch from sex discrimination on the demand-side to female self-selection in certain occupations on the supply-side. The problem here is that segregation declines in time and is more significant for prime-age adults than youths. Indeed, younger male and female workers are likely fairly close substitutes. Because in the countries and periods studied women have increased their participation in the work force and in education more than men, excluding women from the analysis can significantly alter the results. To address this issue without increasing the complexity of the model, lagged female labour force participation is included as an extra instrument in all reduced-form equations. For similar reasons, and to reinforce the overall explanatory power of the instrument set, the reduced-forms also control for migration rates and a time-trend.

Several of these variables introduce uncertainty into the model estimates. First, because of country and time differences in the share of young people at school and out of the labour force, the unemployment rate is likely a poor measure of labour market slack for youth. Second, while import penetration may capture the effects of outsourcing and cheaper imported goods and services on the share of low-skill employment, this indicator may also reflect country size and geographic remoteness from potential trading partners (imports tend to play a smaller role in large than smaller economies). Third, the variable of ICT capital services is inferior as an indicator of technical change relative to information on computer use, for which no data were available over the period of study. Lastly, none of the institutional variables account for age-specific provisions. For a more detailed description of the database, see Section A.1 in the Appendix.

Table 3 in the Appendix presents estimates of six alternative specifications. In comparison to the preferred specification, specification (1) replaces the time trend with time-dummies; (2) omits statistically insignificant institutions; (3) omits all extra instruments that enter only the reduced-form; (4) replaces the lagged youth unemp. rate with a time trend; (5) is a combination of (3) and (4); and (6) is a combination of (2), (3), and (4). While the results are qualitatively robust, standard errors inflate in all cases, making it more difficult to establish significance on key coefficients, which signals that the instrument set is overall weaker. For the first-stage estimates of the preferred specification, see Table 4 in the Appendix. See also Table 4.4 in Christopoulou and Ryan (2009) for estimates from a simpler specification and a smaller sample.

The sample mean of relative demand (i.e. youth/adult employment) is 26% overall years and 32% in 1985 (the earliest year common across all countries). To get a sense of scale, consider that starting from an initial value of 32%, a negative shift of 0.022% points every year for 10 years would decrease relative demand from 32% to 24% ().

This figure is the mean annual growth rate in the estimated relative supply shifts, i.e. it has been calculated by subtracting all wage effects from the estimated relative supply values, including the wage effect in the educational participation equation.

This has been calculated taking the relative wage coefficient in relative supply as equal to zero. The effect is sizeable despite the fact that the employment and education data do not exclude part-timers (i.e. there exists some overlap between the two variables, expected to be highest in English-speaking countries).

The average proportion of males over 24 years old enrolled in tertiary education for the countries examined was 5.5% in 1994, and just 1% point higher in 2001.

Note that the absence of a variable to capture expected future returns to education might make the lagged unemployment variable potentially look more important than it is. Note also that this variable may not be a valid instrument for the identification of educational participation; i.e. lagged youth employment may also be correlated with relative wages via the demand equation. However, this variable is not necessary for identification, and the results are largely robust to its exclusion (see specification (4) in Table 3).

This finding complements reduced-form evidence that minimum wages impact employment outcomes, from a new generation of studies (e.g. Portugal and Cardoso 2006; Meer and West 2012; Bachmann, König, and Schaffner 2012; Sabia, Burkhauser, and Hansen 2012) which explain variation in the data in ways that reconcile previous mixed results (Neumark and Wascher 2008). For example, Meer and West (2012) show that, minimum wages do not affect employment levels, they do however reduce job creation and, thus, affect employment dynamics. Therefore, they impact mostly new labour market entrants, notably youth.

Estimations with the relative unemployment coefficient varying between individual countries are not feasible due to identification failure.

Country-differences in levels and changes in indicators of labour market institutions of most relevance to the youth-specific analysis are illustrated in Figures 3 and 4 in the Appendix.

Time-varying information on school-to-work transition institutions is not available. However, some time-invariant figures of formal apprenticeship use by country from various sources are provided in Christopoulou (2008). Table 2.5 in that chapter shows that apprenticeship registrations for Germany and the Netherlands in the late 1990s amounted to 36.7% and 13.3% of the youth population, respectively, while the relevant figures for all remaining countries in the same period were around 10% or lower. Similarly, enrolment rates in upper secondary education programs with combined school and work-based training reported for year 2005 by the OECD (2008b, Table C1.1) also show Germany and the Netherlands to be at the top of the country ranking, scoring 44.2% and 18.3%, respectively. For qualitative differences in the institutional attributes of apprenticeship across the UK, Germany, and smaller continental European countries, see Ryan (2000).

Specifically, relative unemployment rates are 39.5%, 43.9%, 35.4% points (or 14.9%, 16.6%, 13.4%) higher than the equilibrium rate for the English-speaking group, Cont. Europe I and Cont. Europe II, respectively.

For the projected unemployment adjustment of two country groups (“liberal” and “coordinated”) based on comparable estimates, see Figure 4.3 in Christopoulou and Ryan (2009).