Does active labour-market policy increase employment? Theoretical considerations and some empirical evidence from Sweden.

I. INTRODUCTION

In the search for remedies for the persistently high unemployment in Western Europe there has been a growing interest in so called active labour-market programmes (henceforth denoted ALMPs). These are usually defined to include a wide set of measures to improve the functioning of the labour market that are directed mainly at the unemployed: (i) job broking and placement services performed by employment offices with the purpose of making the matching process between vacancies and job seekers more efficient; (ii) labour-market training in order to upgrade and adapt the skills of the labour force; and (iii) direct job creation, that may take the form of either public-sector employment or subsidization of private-sector work (OECD, 1993).

It has been noted that most EU countries spend relatively little on ALMPs as compared to `passive' unemployment benefits and early-retirement pensions (see Table 1). This has been advanced as an important explanation of unemployment persistence in these countries (e.g. Layard et al,, 1991). As a consequence, several policy documents have endorsed a shift of expenditures in favour of active measures, often viewing Sweden with its traditional emphasis on active labour-market policy as an example to follow (OECD, 1990; European Parliament, 1993; Employment in Europe, 1993; OECD, 1994; Presidency Conclusions, 1994). These reports also recommend that active expenditures should be directed to labour-market training rather than to job-creation measures and that programmes should primarily target outsiders in the labour market, such as the long-term unemployed, young entrants, and other difficult-to-place people.

As can be seen from Table 1, the emerging consensus on a shift towards more active measures has so far produced only very modest results. For the 12 earlier member countries of the EU, the share of labour-market expenditures allocated to ALMPs (measured as an unweighted average) increased only from 29.8 to 32.8 per cent between 1985-9 and 1990-3. For `the model country', Sweden, the share of active expenditures actually fell substantially, from 70.6 to 55.6 per cent, between 1985-9 and 1990-3, implying that active measures did not keep up with the dramatic rise of unemployment. As concerns various types of ALMPs, the share of active expenditures devoted to training has increased slightly Table 2) for the earlier EU countries (unweighted average), although the construction of this measure hides large differences between countries (e.g. large rises in Belgium as well as Britain and a large fall in Denmark).

[TABULAR DATA OMITTED]

Notwithstanding the ongoing policy discussion, our knowledge of the employment effects of ALMPs is very limited. They have usually not been incorporated in a consistent way in theoretical models of the labour market. On the empirical side there exists a substantial body of micro research trying to evaluate the impact on individual participants in various programmes. One problem with this literature is the diversity of results (OECD, 1993). Another is that it does not incorporate the behavioural effects on non-participants, which is necessary in order to evaluate the general-equilibrium consequences. This can only be done through macro studies examining the relationships between various aggregate variables, such as unemployment or real wage levels, on the one hand and ALMPs on the other hand.

So far only a small number of such macro studies have been performed. A few exploiting mainly cross-section variations between the OECD countries, originating with Layard el al. (1991), have come up with favourable effects of ALMPS (see OECD, 1993; Calmfors, 1994; or Katz, 1994). In contrast, time-series studies of wage formation in Sweden have often found ALMPs to increase aggregate wage pressure, which suggests that regular employment (excluding participation in programmes) may be adversely affected (Calmfors, 1993a; Skedinger, 1994). Both these sets of studies, however, suffer from the drawback of a very limited number of observations. A vivid illustration of this has been provided by Forslund and Krueger (1994): when they reran the cross-country unemployment equation that Layard et al. (1991) estimated for 1983-8, for 1993 instead, the ALMP variable both lost its significance and changed sign. The explanation is that a few earlier low-unemployment countries with heavy emphasis on active labour-market policy have changed to high-unemployment ones (mainly Sweden, but also Finland and Norway).

The present paper focuses on the employment effects of training and job-creation programmes. There are two main aims. The first is to structure theoretically the various employment effects that such ALMPs may have. The second aim is to add to the empirical macroeconomic evidence by exploiting pooled time-series and cross-section data from Swedish regions. Since there have been considerable inter-regional variations in both unemployment rates and active labour-market policies in Sweden, a study based on this data set may be very relevant for judging the effects of large-scale ALMPs also in the European context of high unemployment. We shall devote particular attention to the questions of whether training and job-creation programmes have different effects and of how the macroeconomic outcomes are affected by the extent of targeting. These important issues have so far hardly been addressed in the empirical macro studies, although there does exist some evidence that training programmes may be more favourable from the point of view of regular employment than job creation schemes (Forslund, 1992; Edin et al., 1993; Heylen' 1993).

II. A THEORETICAL FRAMEWORK

FOR THE ANALYSIS OF ALMPS

ALMPs can fulfil two basic functions (Calmfors, 1995). The first is `to keep the unemployed going' in general during recessions and to help them maintain or even increase their skills. The basic idea is to exert a positive effect on the effective aggregate supply of labour. This is perhaps the aspect that has been stressed the most in recent years when the unemployment problem has been seen mainly as one of general excess supply of labour (see e.g. Layard et al., 1991, or Wyplosz, 1994). The second - and perhaps more traditional - way of regarding ALMPs is as a means of overcoming structural imbalances in the labour market by adjusting the structure of labour supply to demand. This is the way that active labour-market policies were originally seen in Sweden and the US in the 1950s and 1960s (Fackforeningsrorelsen och den fulla sysselsattningen, 1951; OECD, 1990). This aspect of ALMPs seems now to be receiving increasing attention again in connection with the discussion about an ongoing shift in labour demand from unskilled to skilled labour (e.g. Jackman, 1994; OECD, 1994).

(i) An Aggregate Analysis of ALMPs

The natural aggregate set-up for analysing ALMPs is a modified version of the Layard-Nickell-Jackman (1991) labour-market framework distinguishing between a wage-setting and a labour-demand relationship. The main modification is that we shall distinguish between regular employment and participation in programmes. Based on Calmfors and Lang (1993, 1995) and Calmfors (1994) we can write the relationships for wage setting and regular labour demand (excluding programme participation) as:

w =f (n, y, A) (1)

n = g(w, y, B),(2)

where w is the real (product) wage, n is the rate of regular employment as a proportion of the labour force, y =p/(p + u)=p/(l - n) is the proportion of the job seekers without a regular job that are participating in ALMPs (which we shall denote the accommodation ratio of labour-market policy), p is participation in ALMPs as a proportion of the labour force, u iS open unemployment as a proportion of the labour force, and A and B are vectors of other variables.(1)

The labour-demand relationship is assumed to follow from profit-maximizing behaviour of firms. The wage-setting relationship can be thought of as being derived either from a bargaining (monopoly-union) model or from an efficiency-wage framework. In both …