EU employment and productivity growth patterns have diverged sharply over recent years. Compared with the first half of the 1990’s, the period since then has witnessed a significant increase in the contribution of labour to EU GDP growth but unfortunately this has been accompanied by a reduction in the contribution from labour productivity, with labour productivity growth having come down by about one percentage point. From a purely growth accounting perspective, the 1 percentage point decline in EU labour productivity emanates from 2 sources. Firstly, 50% can be attributed to a reduction in the contribution from capital deepening i.e. lower investment. Secondly, the remaining 50% appears to emanate from a deterioration in total factor productivity i.e. a decline in the overall efficiency of the production process. On top of this, cyclical conditions are estimated to have depressed annual labour productivity growth by around 0.5 percentage points in recent years.
By comparison, over the same timeframe, the US has been able to combine a strong employment performance with acceleration in labour productivity growth. Against this background, this section investigates to what extent the recent slowdown in labour productivity growth may merely reflect a response to a series of positive shocks to labour supply and jobs emanating from structural reforms and employment-friendly wage developments.
Graph 2, as a starting point for the analysis, decomposes labour productivity growth into its two components, with the US and the Rest of World included for comparison purposes. The productivity growth slowdown is evident, with the EU’s long established superiority in terms of labour productivity growth having disappeared over recent years.
Graph 2: Decomposition of Labour Productivity Trends
 |
The benign interpretation of the observed productivity growth trends sees the recent deterioration in performance mainly as the mirror image of structural labour market improvements. Under this view the EU may now simply be in a transition phase whereby wage moderation and positive labour supply shocks may have initially created a negative trade-off between employment and productivity growth, basically via a temporary decline in capital-labour substitution. However, the dynamic adjustment path towards a new equilibrium with higher employment and lower structural unemployment will also involve capital accumulation that should eliminate the trade-off over the medium-term. The more pessimistic view, on the other hand, is that the labour productivity growth slowdown reflects a genuine negative shock, either in the form of a decline in total factor productivity growth or additional pressures on capital productivity; clearly, in such a scenario, prospects for a recovery of labour productivity growth are much bleaker.
Obviously, both interpretations are likely to contain an element of truth, posing the analytical challenge to derive inference on the relative magnitude of the employment and productivity shocks and their respective consequences for overall productivity and employment developments. The picture is complicated by a third possible factor, namely aggregate demand. Both fiscal consolidation and precautionary household savings could have contributed to a decline in growth and, in particular, of productivity growth.
|
|
We have employed both a structural VAR analysis and a simulation using the Commission’s QUEST model to study these three shocks, shocks to employment, shocks to productivity and shocks to aggregate demand and to measure their relative importance for productivity and employment. What is of specific interest in the context of this paper is the dynamic response of productivity to structural employment shocks. In technical terms, we use a structural VAR (SVAR) methodology, based on Stock and Watson (1988) and Blanchard and Quah (1990), for the identification of structural shocks. The intuition for shock identification in Blanchard and Quah is based on the idea that demand shocks only have temporary effects while supply shocks have permanent effects. Stock and Watson extend this approach and allow for separate supply contributions from labour and productivity (TFP). In order to identify different supply contributions, namely those coming from employment and those coming from productivity, additional identification criteria must be introduced. Stock and Watson use long run restrictions implied by the neoclassical growth model for that task. The neoclassical growth model appears to be suitable, since there are at least three important features in the long run trends which are compatible with this model:
· A close trend correlation between the growth of labour productivity and capital intensity.
· Capital intensity and productivity grow at a similar rate in the long run.
· If one looks over long periods of time and across the EU and the US, the employment rate appears to be unrelated with productivity growth.
Using the neoclassical growth model this leads to the imposition of the following long run restrictions:
· The labour market shock can have short and long run effects on employment, productivity and inflation.
· The productivity shock can have long run effects on productivity and inflation but only short and medium run effects on employment. This constraint arises from the assumption that real wages are indexed to productivity in the long run.
· The demand shock can have a long run effect on inflation only but not on employment and productivity. No long run constraint is imposed on inflation.
These three types of restrictions imply a triangular long run structure between the growth rate of employment (
), productivity (
) and inflation (
) on the one hand and the corresponding shocks to employment (v), productivity (e) and demand (d) on the other. If one defines the vector
|
|
and the vector 
, then the moving average representation of this model is given by:
with 
where the matrix A(1) shows the long run restrictions. Note, this particular structure is particularly suited to test for the short, medium and long run effects of an employment shock. Allowing for a non-zero long run productivity effect of an employment shock allows one to test for labour quality effects associated with a permanent change in the employment rate. A similar analysis of the employment effects of productivity shocks has been conducted by Gali (1999).
The empirical results are presented in two steps. In step one, the impulse responses from the estimated VAR are presented. These responses give the impact on employment and productivity of a unit shock to employment, productivity and demand. Recall that the identifying restrictions imply that temporary unit shocks to employment can have permanent effects on employment and productivity, while a unit shock to demand (inflation) can only have temporary effects.
In order to evaluate the quantitative magnitudes of these shocks, they are compared to similar shocks simulated with the Euro area QUEST model. This comparison is useful since it shows whether orders of magnitude from these shocks are similar when two very distinct empirical tools are used, with the VAR model imposing very little economic structure (apart from the long run constraints), while QUEST consists of explicitly estimated structural equations and estimated adjustment lags.
· Employment shock: A positive employment shock initially leads to an increase in productivity; however, this short run positive effect in the VAR model is partly spurious. In the medium and long run the effect on productivity is negative, i.e. an increase in employment is associated with a decrease in labour quality. Note, though, that this negative long run effect is estimated to be small: a shock which leads to a permanent increase in the level of employment of about 1% is associated with a long run productivity level effect of about -0.1%. Analysis based on QUEST model simulations yields fairly similar results to the VAR approach, but the negative impact upon the productivity level is slightly stronger (-0.3 instead of -0.1) over the medium term; moreover, the QUEST model analysis does not reveal any short run increase in productivity.
· Productivity shock: A positive productivity shock is associated, in the short-run, with a small negative employment effect. The order of magnitude of the employment effect is only about one tenth of the size of the productivity shock. By implication, this analysis suggests that a structural slowdown in labour productivity growth will, by itself, not be associated with an expansion of employment. Again, in the QUEST model analysis a qualitatively similar pattern to the VAR emerges, but the short-run negative employment response appears to be somewhat stronger.
|
|
· Demand shock: The demand shock is initially associated with a positive employment and productivity effect. This result appears quite plausible, since a demand shock is likely to lead to better capacity utilisation in the short run. As the demand effect fades away and employment is slow to adjust, the productivity effect turns negative and dies out within a year.
Graph 3: Impulse response analysis
A) Employment shock
B) Productivity shock
C) Demand shock
Graph 4: QUEST analysis
A) Employment shock
B) Productivity shock
In the second step of the empirical analysis, the shocks are cumulated over the period 1995q1 to 2003q4 [5] in order to derive an estimate for the structural component in employment growth and its likely impact on productivity and vice versa. The results of this exercise are depicted in Graph 5. The cumulated size of the employment shock over the period 1995-2003 is estimated at about 5%. Thus, roughly one half of the overall observed employment expansion over that period is attributed to structural trend improvements. According to the VAR approach the cumulated productivity cost of this structural employment expansion may have amounted to ¾ of a percent; the QUEST model simulations would put the productivity cost somewhat higher at 1 ½ per cent. Roughly translated into year-on-year figures using a mid-point between the VAR and QUEST estimates, this implies a reduction in annual productivity growth of around two tenths of a percentage point, equivalent to some 20 % of the observed total productivity growth slowdown, which could be attributed to positive structural shocks in the labour market.
Graph 5: Cumulated euro area employment shock 1995 Q1 to 2003 Q 4
A further interesting result of the VAR model relates to the question of the structural versus temporary nature of the productivity growth slowdown. Based on the underlying assumptions on the short, medium and long term impact of the various shocks, the VAR model attributes most of the decline in productivity to a structural trend decline in productivity growth. As can be seen from Graph 6, the autonomous shock to productivity explains a decline in the level of productivity of almost 5%, which would translate into an annual average productivity growth rate effect of the order of -0.6 percentage points. This is fully consistent with the growth accounting result given earlier of a decline in TFP of the order of a ½ a percentage point, with TFP considered to be a reflection of the structural component of the productivity trend.
Graph 6 also indicates that the autonomous productivity shock is unable to explain the increase in employment. Therefore, it is necessary to look separately at both shocks in order to give a complete picture of both the employment and productivity developments. However, concerning productivity, the overall conclusion from the analysis suggests that the decline in productivity growth is to a large extent structural in nature.
Graph 6: Cumulated euro area productivity shock 1995 Q1 to 2003 Q 4
The empirical results presented above are quantitatively broadly in line with other available evidence on structural labour market improvements as indicated by a trend increase in participation and a reduction in structural unemployment. Moreover, relating the productivity effect to real wage moderation also suggests that the estimated impact on short-run productivity developments is of a reasonable order of magnitude. A stylised number for real wage moderation in the past 10 years or so would put the average annual reduction in real efficiency wages at slightly less than ½ a per cent. Thus, back-of the envelope calculations would suggest that real wage moderation could, on average, have reduced annual actual labour productivity growth relative to its balanced steady-state rate by about two tenths of a percentage point, which is well within the range derived from the VAR and QUEST model approaches. Further corroborating evidence stems from growth regressions suggesting that about 25% of the productivity decline is due to the increase in employment[6].
In summary, and recalling that the overall slowdown in average annual productivity growth has amounted to about one percentage point, it emerges as a fairly robust result that only some 20 % of this reduction can be attributed to the dynamic response of productivity to positive structural shocks in the labour market.
Conclusions
In a nutshell, the analysis in this paper dismisses the notion of a genuine trade-off between employment and productivity growth. Obviously, misguided policies to exploit such a trade-off have to be avoided. However, there are no reasons to think that structural labour market reforms boosting employment will typically entail negative implications for longer-term productivity growth. In particular, this paper reaches the following conclusions:
•The negative relationship between productivity and employment in comparative-static considerations should not be interpreted as a genuine trade-off.
•However, all else equal, a move towards full employment is likely to see a widening of the labour productivity gap between Europe and the US.
•The dynamic response of productivity to positive labour supply and wage shocks may entail a temporary reduction in productivity growth rates, which, in principle could be considered as benign; in any case, the size of a negative effect of this type is estimated to be fairly small.
•The increase in employment since the mid 90s has indeed been to a significant extent the result of such positive labour market shocks, with about one half of the additional jobs attributed to structural improvements.
•Positive employment shocks can only account for a very small fraction of the observed productivity slowdown; consequently, the decline of labour productivity growth must be considered as predominantly caused by other factors and probably not just a temporary phenomenon. Indeed, a cyclical pick-up in labour productivity growth after the recent period of weak output growth should not divert attention from the “deeper” structural problem of a slowdown in trend productivity growth.
The implications of the above findings for the Lisbon strategy are straightforward: Indeed, “the more jobs the better” may serve as a simple catch-phrase characterising the principal goal of labour market reform efforts since there is no genuine trade-off – in the sense of a difficult decision to be made – between policies to raise the employment rate and policies to foster productivity growth. Of course, misguided policies attempting to exploit such a trade-off have to be avoided – if, for example, policy-makers promoted sectors with low productivity growth prospects, if they introduced unnecessary regulations leading to “over manning”, if they discouraged young people from pursuing further education, or if they used funds for public training programmes in an unproductive manner, then employment might be raised at the expense of longer-term productivity potential. However, none of these policies is advocated in the EU’s economic and employment policy framework and, in consequence, the employment strategy should not be blamed for the dismal productivity performance in recent years.
[1] See Pichelmann and Roeger (2004) for an analysis of potential growth in the EU.
[2] Labour augmenting technical progress is equal to Harrod-neutral technical progress when the capital stock grows at the same rate as output, thus leaving the capital-output ratio constant. For a Cobb-Douglas production function this “balanced” labour productivity growth rate is defined as TFP growth divided by the labour share.
[3] Gordon (1995) provides a neat theoretical and empirical investigation as to how a productivity-unemployment trade-off might emerge and how it will subsequently be eliminated through a dynamic path of capital adjustment.
[4] Obviously, misguided policies attempting to exploit a perceived trade-off have to be avoided, for example unnecessary regulations leading to “over manning”.
[5] It should be noted that this provides an estimate for the overall magnitude of the shocks, but not of the impact these shocks have had on the macroeconomic aggregates.
[6] See EU economy review 2003, Table A3.