A CEPR conference to examine the mechanisms underlying the dynamics of unemployment was held in Cambridge on 16 December 1996. The participants, all with interests in the analysis of high levels of unemployment, set out to examine radically different approaches to unemployment, especially unemployment persistence. Issues such as the effects of the labour market reforms of the 1980s and early 1990s on the response of unemployment to macroeconomic fluctuations, and the importance of asymmetric responses to favourable and unfavourable shocks, were central topics of discussion, as was the question of whether a new policy agenda is necessary to combat Europe’s high unemployment levels. Dennis Snower (Birkbeck College, London and CEPR), Brian Henry (London Business School) and Jenni Greenslade (London Business School) were the conference organizers. Regular workshops on these issues are planned for the future.

Marika Karanassou (Birkbeck College, London) presented ‘Is the Natural Rate a Reference Point?’, written with Dennis Snower (Birkbeck College and CEPR). This paper explored the two common concepts of the natural rate of unemployment: (1) the stable, long-run equilibrium rate, and (2) the equilibrium rate at which there is no tendency for this rate to change, given the exogenous variables. Using the second concept, the paper showed that the natural rate is not necessarily a reference point – i.e. a value towards which the equilibrium unemployment rate tends with the passage of time – in multi-equation labour-market models containing lagged endogenous variables and exogenous variables with non-zero long-run growth rates. Since lagged endogenous variables are a common feature of labour-market models, and non-stationary exogenous variables are common on account of economic growth, the authors argued that the issue was likely to be of considerable practical importance. This casts doubt on the usefulness of the natural rate hypothesis as a predictive tool. The authors believed that the way out of this box was to estimate multi-equation models of the labour market and to use these to predict the unemployment rate. Richard Jackman (London School of Economics) questioned the nature of the wage-setting function. He suggested that changes in the NAIRU are wage-dominated (with a vertical wage schedule) but believed that the wage schedule implied by the Karanassou and Snower model may be horizontal. There was also the issue that, in the profit function, with adjustment costs, expectations play a role.

‘Business Cycles and the Labour Market: Can the Theory Fit the Facts?’ was presented by Stephen Millard (Bank of England) and was written with Andrew Scott (London Business School and CEPR) and Marianne Sensier (University of Oxford). The paper forms part of an ongoing research project that attempts to evaluate how well small analytical models can explain certain ‘stylized facts’ of UK data. Specifically, the authors were interested in seeing how well four recent models of the labour market performed against the data. Of the models they examined, only the Cho and Cooley (1994) model was found able to generate sufficient volatility in total hours. It did this, however, by making employment highly volatile and average hours much less volatile. While this matches the stylized facts in the United States, employment and average hours in the United Kingdom appear equally volatile. Only Gali’s (1995) model was able to offer a convincing explanation of the unemployment rate. Different models may perform better in matching different features of the data.

Julia Darby (University of Glasgow) suggested using spectral analysis to detrend the data. She believed that using total hours instead of average hours might improve the findings. The authors’ paper was based upon representative agents, and she suggested that they could try using sectoral, instead of aggregate, data. Rod Cross (University of Strathclyde) suggested including non-wage income in a labour-supply function.

‘Insider-Outsider Conflict in Equilibrium Unemployment Theory: Does it Matter?’ was presented by Chris Pissarides (London School of Economics and CEPR) and written with Dale Mortensen (Northwestern University). They showed that in a rational model of job creation and job destruction with job creation costs, the wage path does not influence the final equilibrium outcome. Therefore, insider power, which leads to higher wages for existing employees, does not matter for equilibrium, if there is a period when workers are outsiders. The bargaining outcome with newcomers to the job is such that it offsets the effects of insider power. The paper demonstrated that, if the wage of newcomers cannot fall sufficiently to achieve this, job creation rates will fall and job destruction rates rise, leading to more unemployment in equilibrium. They also show that it is socially optimal for the newcomers to share the job creation costs with the firm and that, generally, the job destruction rate is too low in equilibrium. The job destruction and creation rates are optimized only when the wage rate can be chosen by a centralized body that represents the interests of the unemployed.

Alison Booth (University of Essex and CEPR) gave a detailed account both of how this paper built upon the existing literature and of its contribution, particularly in terms of the job creation costs and efficiency aspects. In response to a question on the rationale for renegotiation, Pissarides stated that timing did not have an impact on the equilibrium outcome. The influence of firing costs would be addressed in the future.

The paper by Andrew Hughes Hallett (University of Strathclyde and CEPR), entitled ‘Is There Convergence in a Two-Speed Europe? Evidence from the Labour Markets’, was written with Maria Demertzis (University of Strathclyde). They examined whether Europe’s emerging ‘two-speed’ framework reflected genuine differences in market structures, underpinned by relatively immobile labour, or whether it was really the result of poor economic management in the periphery. Using the models of German leadership and of locational competition, they showed that the origin of the two-speed divide lies in bargaining behaviour. The core enjoys German leadership, and the periphery locational competition. As a result, convergence in competitiveness between the two groups was negative in the 1980s and painfully slow in the 1990s. The greater the differences in cost structures, the more such a two-speed regime will emerge as a result of policy leadership, and the greater the conflict between real and nominal convergence. This conflict is ultimately incompatible with further integration. The authors argued that, to prevent its competitive edge being eroded, the core may well prefer to preserve its two-speed regime.

Ray Barrell (National Institute of Economic and Social Research, London) questioned whether there was a real convergence problem in 1992. He believed that the integration of goods and capital markets was important. Issues raised included the likelihood of German leadership (in wage settlements), the consequences of locational competition (in terms of production processes), possible spillovers in bargaining and the appropriate level of aggregation.

David Lopez-Salido (Banco de España) presented ‘Hysteresis and the Sources of Shocks: A Cross-Country Analysis’, written with Manuel Balmseda (CEMFI, Madrid) and Juan Dolado (Universidad Carlos III, Madrid, and CEPR). The paper used long-run identifying restrictions on a three-variable system containing output growth, real wage growth and the unemployment rate, so isolating three ‘structural’ shocks that drove business-cycle fluctuations in 16 OECD countries during 1950-94. The authors found that the observed dynamic adjustment of the variables in response to the structural shocks is much more consistent with the implications of a model with full hysteresis in unemployment than with a partial hysteresis model. In particular, aggregate demand and productivity shocks appear to have almost mirror-imaged cumulated effects on output and unemployment; labour supply shocks tend to increase unemployment; and productivity shocks are the most important ones in explaining fluctuations in real wages. They also concluded that productivity and aggregate demand shocks explain most of the fluctuation in real output, the former being particularly important in the Anglo-Saxon and Scandinavian economies. Competing hypotheses were assumed for the stochastic behaviour of the unemployment rate: trend stationary or integrated of order one. The latter appeared more plausible, thus supporting hysteresis theories of the labour market.

Simon Price (City University, London) agreed that the hysteresis model works, but suggested that the labour supply shock could be reinterpreted, replacing it by a measure of the natural rate of unemployment, so that it represents a shock to equilibrium unemployment. In this case, an increase in the natural rate would reduce wage pressure. He suggested using a nested model to test for over-identifying restrictions.

Marco Bianchi (Bank of England and CEPR) presented ‘A Nonparametric Analysis of Regional Unemployment Dynamics in Britain’, written with Gylfi Zoega (Birkbeck College, London). They estimated the probability distribution of relative unemployment at the county level in the United Kingdom for the years 1981–95. They found that the distribution was unimodal in all years, with a falling variance between 1989 and 1994. In any given year, there were counties with very high and very low unemployment rates, but they did not detect significant sub-groups in the data with either high or low rates. They used bootstrap methods to determine critical values for the two tails of the distribution and they analysed intra-distribution dynamics. By constructing confidence intervals for quintiles in the density which leave 30% observations in each tail of the distribution, they analysed the persistence of movements of counties between the three states of unemployment (low, average, high), corresponding to the middle and the two tails of the distribution. They found that positive shocks (those that reduce unemployment) tend to persist longer than negative ones. Transition probabilities also confirmed the high level of persistence of changes in relative unemployment, implying that the regional adjustment mechanisms of inter-county migration and capital movements are very weak.

Chris Allen (University of Limburg) commented that the small sample size may be a problem in determining confidence intervals. He suggested that the degree of asymmetry of the distribution could be considered, or a panel used to estimate the distribution if there were more than one mode. The persistence of positive and negative shocks was also dependent upon the business cycle. The authors responded that the data had been rescaled to deal with this.

The penultimate paper was entitled ‘What Determines the Natural Rate of Unemployment? And What Does Not?’. In it, Peter Westaway (Bank of England) examined the concept of the natural rate of unemployment both theoretically and empirically. He argued that, if the steady-state rate of unemployment actually can be affected by demand-side variables, then the conventional wisdom – which focuses completely on structural supply-side reforms of the labour market – may be misplaced. Surprisingly, many empirically estimated natural rates do include demand-side variables, such as the real exchange rate, world trade, the oil price and tax wedges. Indeed, it is argued that some of these demand-side effects would seem implausible, sometimes implying a natural rate that trends into the indefinite future. Westaway’s paper was based upon 1993 work by Manning, proposing that many of these effects emerge because of insufficiently rigorous theoretical underpinnings to the wage-determination process. The author showed the extent to which the natural rate can be affected by the demand side through relative factor prices, and suggested that this feature could be exploited to resolve the well-known identification problem afflicting structural wage equations. An empirical estimate of the UK natural rate has been derived using this approach.

Manuel Balmseda (CEMFI, Madrid) commented that the identification was satisfied in the CES case, but not for simpler technologies such as Cobb-Douglas. Discussion then followed on the role of the replacement ratio, the order of integration of unemployment, the validity of the bargaining model and the issue of whether demand-side variables could influence the natural rate.

Kevin Lee (University of Leicester) presented the final paper, ‘Identification and Estimation of Aggregate Wage and Employment Equations: An Application of the Structural VAR Modelling Approach’, written with Brian Henry (London Business School). Lee and Henry argued that, contrary to a widespread view, it is possible to identify the structural parameters of a model of wage and employment determination, and they suggested some devices for overcoming the typical difficulties that arise in models applied to aggregate data. Using a union-based model of wage and employment determination, derived in a static framework, the authors demonstrated that identification is still possible even in the presence of (unspecified) dynamics. Identification does not rely on the use of dynamics – indeed, the presence of dynamics is shown to be unimportant for the identification issue. Newly developed econometric methods were employed to estimate the relationships within a cointegrating VAR framework, using annual UK data for the period 1954-91. The empirical work showed that there was no evidence with which to reject the over-identifying restrictions suggested by the structural model, and it provided reasonable estimates of the uniquely identified structural parameters of the model.

Alan Manning (London School of Economics and CEPR) agreed that the approach proposed appeared to deal with the identification problem. Nonetheless, he argued that the estimated model still did not provide a satisfactory explanation of the trend rise in unemployment in the United Kingdom. A lively debate ensued, centred on the adequacy, or otherwise, of models which emphasise labour supply-side factors in the rise of equilibrium unemployment.