Unemployment Persistence
Re-Evaluating the Natural Rate

A CEPR conference on 'Unemployment Persistence and the Long Run: Re-Evaluating the Natural Rate' was held in Vigo, Spain on 30 November/1 December 1997. The organizers were Dennis Snower (Birkbeck College, London and CEPR), Brian Henry (London Business School) and Jenni Greenslade (London Business School). The objective of the conference was to examine the observed movements in European unemployment and to assess the degree to which these movements could be explained by changes in labour-market equilibria, as opposed to dynamic unemployment processes.

The conference brought together a group of academics from Europe and North America with interests in the analysis of high levels of unemployment to examine radically different approaches to unemployment. Policy implications were also considered, as was the question of whether a new policy agenda was required to combat the high European unemployment levels. The deliberations continued the theme developed in a previous conference and workshops, but extended the debate to look in depth at developments across Europe and North America.

Stephen Nickell (University of Oxford) presented 'Unemployment: questions and some answers'. In almost every OECD country, unemployment was lower in the decades following the Second World War than in any other period of comparable length. The paper set out to consider why this was so, and to ask what factors had caused the enormous variations in cross-country unemployment rates. Nickell offered some tentative answers to the second question and some speculation about the first.

The paper set out a general dynamic model in which the long-run solution for unemployment depended upon industrial turbulence, the replacement ratio, the terms of trade, the skills mismatch, the mark-up, the tax wedge and the real interest rate. This relationship comfortably explained the four-fold rise in unemployment from the early/mid-1960s to the late 1980s/early 1990s. There remained, however, a feeling of dissatisfaction, since it was hard to tell a seven-variable story simply, and because the variables that were really needed – for example, those close to capturing exogenous shifts in search effectiveness – were not observed. On the reasons for the rise in unemployment, Nickell suggested that, by the 1980s, either employers found it far harder to get the workers they wanted from the unemployed pool, or the unemployed were much less enamoured of the work on offer. The problem was to provide a persuasive empirical analysis of these apparently large shifts in behaviour. Dennis Snower emphasised the importance of lagged dynamic effects, arguing that the outcome depended upon the extent of nominal inertia. Robert Gordon (Northwestern University, CEPR and NBER) suggested that a cross-sectional estimation be conducted to analyse the impact across countries.

Brian Henry (London Business School) presented 'Dynamic adjustments versus the natural rate: A story of UK unemployment', written with Dennis Snower (Birkbeck College, London and CEPR). This paper challenged the standard account of UK unemployment, namely that the major swings in unemployment over the last 25 years were due predominantly to movements in the underlying empirical 'natural rate of unemployment' (NRU). The authors' explanation of the medium-term movements of unemployment was that the trajectory of unemployment through time reflected the cumulation of labour-market 'impulses' and their associated streams of 'dynamic' responses, with each shock followed by a prolonged adjustment process. Their paper thus sought to refocus analysis of the labour market away from exclusive attention to equilibria and towards the dynamic movements between them. They contended that, because these movements took place over long periods of real time, they were important objects of analytical attention.

The analysis suggested that an important part of the movement of UK unemployment was due to the interaction between structural shifts in exogenous variables (such as the real oil price and the interest rate), and mutually reinforcing lagged labour-market responses operating on employment, wage determination, labour supply and capital accumulation. In contrast to the conventional interpretation, the authors argued that the UK's NRU had been reasonably stable through time, whereas the major unemployment swings had arisen from the cumulation of one-off shocks and their associated dynamic repercussions. Discussion centred on the nature of shocks that might yield a transitory, but possibly prolonged, effect. Edmund Phelps (Columbia University) gave examples of many such shocks, including EMU and real interest rates. David Rose (QED Solutions, Ontario) pointed out that separating the equilibrium level and the dynamic path was a difficult empirical task.

Robert Gordon (Northwestern University, CEPR and NBER) presented his paper 'Price, wage and unemployment dynamics in the recent experience of the United States'. The paper provided a unified framework in which inflation depended on inertia, demand and supply. The benign behaviour of inflation and unemployment in the United States in the mid-1990s was interpreted as the direct converse of the positively correlated 'twin peaks' of US inflation and unemployment in the mid-1970s and early 1980s. The decade between 1973 and 1982 was dominated by the macroeconomic effects of the adverse food and energy price shocks, which were largely reversed in 1982–86. Gordon's interpretation of the 'Cinderella' economy of the mid-1990s was that three beneficial supply shocks had combined to create 'twin valleys' of inflation and unemployment. These shocks were: (1) acceleration in the rate of computer price deflation; (2) reduction in the relative price inflation of medical care; and (3) measurement improvements which had reduced measured inflation relative to true inflation. The estimation of wage equations indicated that the same factors that had held down price inflation had also held down wage inflation. There was thus no need to appeal to factors specific to labour markets to explain why wage inflation had been so well behaved in recent years. The author explained that a period of outstanding economic performance was only the first step in guiding policy-makers; also needed was a prediction of the future behaviour of the special factors. Dennis Snower questioned whether the extent of the changes in computer and medical care prices and in the measurement of prices accounted fully for the movement in the Phillips Curve, and whether the exchange rate might also have an important role to play.

The paper by Edmund Phelps (Columbia University), entitled 'Lessons in natural-rate dynamics' built upon previous work on the dynamic model of the natural rate, such as the author's 1994 book. In this paper, attention was paid to these medium-run dynamics, with all frictions stripped away in order to focus on what Phelps believed was the crucial driving force behind the equilibrium motion of unemployment in an open economy, namely the accumulation of private assets net of liabilities – private wealth. Thus the Phelps-Winter customer was banished in this model. Only physical capital that was frictionless, costless and instantaneously transportable in and out of the country was permitted. The turnover-training model of wages and unemployment, as rebuilt by Hoon and Phelps, was used. For simplicity, however, the hiring/training cost of each new employee was taken to be constant, rather than an increasing function of the hire rate. In the equilibrium case, with these simplifications, the unemployment rate jumped on to its equilibrium path and, in the future, the equilibrium rate was driven by the evolution of the private wealth of the representative worker.

Robert Gordon questioned why US unemployment was currently low, given that stock market wealth was high. The flow of the benefits from wealth, rather than the current market value of workers' homes and durables, was more relevant. Another participant questioned why wealth dynamics should affect the movement in unemployment given that the unemployed hold only a small fraction of the average national wealth per head.

'A comparison of inequalities across French and US labour markets', presented by Daniel Cohen (CEPREMAP, Paris, and CEPR) provided a theoretical comparison of the bargaining power of French and American workers and an investigation of the channels through which they were affected. Using (the log of) wage inequality as a criterion, Cohen found that the United States was a much more inegalitarian society, with a coefficient of variation about 60% higher than the French coefficient. Taking the computed value functions as a criterion, the US was still more inegalitarian, but by a margin of only 15%. This difference was not uniform across various segments of the labour force: French labour markets were more inegalitarian along the age axis, whereas US labour markets were more inegalitarian along the diploma axis.

The author found that, for all segments of the labour market, job turnover was the dominant feature, always explaining more than 80% of the discrepancy between the welfare of an employed and an unemployed worker. A difference arose only in the segment of workers with no diplomas, in which the bargaining power of French workers was about 20% higher. For all the other segments, there was no difference in the wage-determination process in the two countries. Job turnover, rather than lack of wage flexibility, therefore appeared to be the critical discrepancy. Henri Sneessens, (IRES, Université Catholique de Louvain and Faculté Libre des Sciences Economiques, Lille) initiated discussion on the possibility that bargaining power might be underestimated. Attention also focused on the issue of job tenure, with more job-to-job turnover taking place in the United States, where it was quicker for workers to be rehired. Stephen Nickell questioned whether differences in France could be overstated because of higher employment protection in that country.

David Rose (QED Solutions, Ontario) presented 'Inflation and unemployment in Europe and North America: Asymmetry versus hysteresis', written with Hamid Faruqee (International Monetary Fund) and Douglas Laxton (International Monetary Fund). A model and methodology for simultaneous estimation of the Phillips Curve and the Non-Accelerating Inflationary Rate of Unemployment (NAIRU) was specified in a small system of equations, and was applied to the data for seven countries in Europe and North America. The authors found that a model with modest asymmetry in the Phillips Curve fitted the data better than a linear alternative in most cases. They also found that this model could explain part of the rise in the NAIRU in many countries over the past few decades as a macro phenomenon, related to the increased volatility of the business cycle.

The model with asymmetry had some interesting policy implications. Interpreting the NAIRU as the 'equilibrium' unemployment rate consistent with stable inflation, the authors noted that this value must lie above DNAIRU (deterministic NAIRU), namely the value that would be obtained in a world without shocks. The difference between these values, which they dubbed the a-shift, depended, among other things, on the degree of asymmetry and the degree of volatility in the macro cycle. Their estimates suggested that the a-shift had been growing through the sample in most countries, but may now be falling in countries where stabilization policies have been more successful in establishing both credibility in the stable-inflation monetary policy, and a new direction with respect to fiscal policy. Dennis Snower questioned whether stochastic variations around a non-linear short-run Phillips Curve implied a non-vertical long-run Phillips Curve. David Mayes (Bank of Finland) asked whether the convexity of the curve had changed, to which David Rose replied that there was some empirical evidence of increased convexity.

Marika Karanassou (Queen Mary and Westfield College) presented 'Surprising properties of the natural-rate hypothesis', written with Dennis Snower. The paper examined whether it was possible to find a unique empirical estimate of the natural rate of unemployment, representing the rate towards which unemployment is tending. In the standard, single-equation models of natural-rate theory, such a unique rate was simply the unemployment rate at which unemployment adjustment dynamics have worked themselves out. The authors argued, however, that single-equation models were at best an encapsulated summary of the unemployment behaviour that emerged from a multi-equation labour-market system containing labour demand and labour supply equations.

For an empirical model of the UK labour market, they computed various estimates of the long-run unemployment rate and found that these differed substantially from one another. Since each of the estimates was equally compelling as a potential reference point towards which the actual unemployment rate could move, the search for a unique, empirically assessable natural rate of unemployment consequently must be abandoned. Rather than viewing unemployment dynamics in terms of movements towards an identifiable reference point, it might be more useful to focus instead on labour-market shocks and the lagged behavioural adjustments to them. Given that each shock affected the labour market by feeding through a network of lagged adjustment processes, unemployment movements may then be viewed as a cumulation of past shocks and their associated dynamic sequence of labour-market responses. Stephen Nickell suggested that, since capital accumulation and working populations evolved at different rates across countries and time periods, unemployment may be independent of these factors. Dennis Snower considered that this depended upon the time frame, and that long-run restrictions should not be imposed in the medium run.

Gylfi Zoega (Birkbeck College, London) presented 'Education and the natural rate of unemployment'. Written with Michael Orszag (Birkbeck College, London) and Edmund Phelps (Columbia University), this paper used Phelps's 1994 turnover-training model to predict differences in the rate of unemployment across education groups. There were two types of labour which differed in their level of education. The authors found that the rate of unemployment in each group depended on the rates of autonomous inflow into and out of employment, the responsiveness of hires and quits to changes in the rate of unemployment and the unemployment-benefits replacement ratio. Unemployment among the less educated was higher for three reasons. First, the less educated faced a higher unemployment-benefit replacement ratio since their wages were lower, and hence they were more prone to quit and less likely to accept jobs. Second, the less educated could perform fewer tasks in the firm and hence were more often turned down for jobs. In turn, this reduced the rate of inflow into employment and raised the group's unemployment rate. Third, the less educated may be more prone to quit because of a variety of social problems affecting those in the lower levels of the education distribution disproportionately.

Within the context of the model, the recent deterioration in the relative unemployment/wage rate of the less educated could be attributed to the information revolution which had made firms more selective when hiring from the pool of less educated workers, rather than relying on biased technological change in the traditional sense of affecting the ratio of marginal products. There was some discussion about the impact of the minimum wage on the outcome. Furthermore, such a model potentially allowed for a large role for longer-run unemployment dynamics via education. The authors believed that education provided part of the explanation for the rise in unemployment.

Torben M Andersen (Aarhus University and CEPR) presented a paper entitled 'Persistency in the Danish labour market', written with Svend Hylleberg (Aarhus University). The aim of the paper was to identify different sources of persistency in employment adjustment. Based on a dynamic labour-market model, an explicit distinction was made between real and nominal (prices and wages) propagation mechanisms. The theoretical analysis provided the basis for an empirical analysis of wages, prices and employment for the manufacturing sector in Denmark from 1973 to 1993. The authors found that nominal rigidities prevailed in the short run and that nominal propagation mechanisms played a larger role than real propagation mechanisms. As the half-life of temporary shocks was about 11 quarters, it followed that endogenous propagation mechanisms were important, although they could not fully explain the observed persistency.

The finding that nominal adjustment failures had a larger effect than real adjustment failures had several important implications. First, it might be misleading to base measures of the structural unemployment rate on recently observed unemployment rates. Second, policies directed towards making the labour market more flexible, and so reducing inertia in the adjustment process, should focus more on the incentives underlying wage and price formation than on employment adjustment. Finally, given substantial inertia – particularly nominal inertia – there is a role not only for general demand management policies in smoothing employment, but also for monetary policies in speeding up the adjustment process. Simon Burgess (University of Bristol and CEPR) commented that it might be interesting to consider the effects of linear symmetric adjustment mechanisms. Edmund Phelps questioned whether some persistence in real shocks was needed. It was also suggested that it would be important to evaluate the relative importance of the real and nominal adjustments.

'Labour supply, the natural rate and the welfare state in the Netherlands: the wrong institutions at a wrong point in time' was presented by Coen N Teulings (University of Amsterdam and Ministry of Social Affairs, The Netherlands) and written with Lourens Broersma and Jan Koeman (both Ministry of Social Affairs, The Netherlands). This paper analysed the causes of the Dutch miracle. Low wage increases in the Netherlands, by comparison with the rest of Europe, were the main factor explaining the fast employment growth. Two factors might explain the lag in wage growth: the 1982 Wassenaar agreement between trade unions and employers; or the realignments in the welfare state.

A small macroeconomic model for the Dutch economy was estimated to analyse these issues. The residuals of the wage equation were not systematically negative for the post-1982 period. These residuals did have a large impact on the employment history, however. There was a clear effect resulting from the generosity of the welfare state, but the data did not allow the magnitude of this effect to be established precisely. The impacts on the Dutch economy of the oil price shock and the social experiment of more generous welfare benefits were felt during much the same period (1975–84) and so it was difficult to disentangle their effects. Stephen Nickell pointed out that the Netherlands had the highest productivity in the world. He commented that if wages grew in line with productivity, then unemployment would appear constant in the model.

The next paper, 'Duration distributions and unemployment dynamics', was presented by Michael Orszag (Birkbeck College, London). Economic policy frequently relied on constructs which related changes in unemployment, away from its natural rate, to other variables, such as growth rates (Okun's Law) or inflation (the Phillips Curve). Such constructs ignore the distribution of unemployment. In Orszag's view, this was misleading for economic policy because distributional issues – such as the number of long-term unemployed – had implications for both growth and inflation.

Beginning with a general Markov chain model, Orszag derived a continuum model of employment and unemployment, the steady states of which had distributions corresponding to those used in the microeconometrics of transitions. These steady states differed considerably from traditional natural-rate theories in which only the aggregate unemployment rate settled down to its steady-state level. The author demonstrated the use of the model in calculating steady-state unemployment rates for comparison with natural-rate theories. He also analysed convergence to steady-state equilibrium and discussed possible extensions. Simon Burgess suggested that the flows into and out of unemployment, with their associated probabilities, be included in the model, and Stephen Nickell commented that it would be beneficial to include wages.

'Unemployment persistence' was presented by Alison Booth (University of Essex and CEPR) and written with Wiji Arulampalam (University of Warwick) and Mark Taylor (University of Essex). They estimated dynamic panel-data models of unemployment incidence, in order to disentangle the effects of unobserved individual heterogeneity and true state-dependence. They also controlled for the 'initial conditions' problem that arose when the start of the observation period did not coincide with the start of the stochastic process generating individuals' unemployment experiences. These problems could be properly addressed only with panel data, such as the new British Household Panel Survey. If there was no state-dependence in the incidence of unemployment, then short-run policies to reduce unemployment would have no effect on the natural rate. If there was true state-dependence, however, then policies that reduced the incidence of short-run unemployment would have longer-run effects by reducing the natural rate.

The authors estimated their models for a sample of men for the period 1991–5. They found strong evidence of state-dependence, especially for more mature men (defined as those aged 25 and over in 1991). This finding was consistent with the 'scarring' theory of unemployment, which suggested that individuals' previous unemployment experiences had implications for their future labour-market behaviour. Some discussion, initiated by Daniel Cohen, concerned a possible sample-selection problem. Alison Booth confirmed that if a person became employed, he/she would remain in the sample, but that the data did not take into account a person who may have changed jobs several times during the survey period.

Juan Dolado (Universidad Carlos III, Madrid, and CEPR) presented 'Hysteresis and the sources of shocks: A cross-country analysis', written with Manuel Balmeseda (CEMFI, Madrid) and David López-Salido (Banco de España). Their paper used long-run identifying restrictions on a three-variable system, containing output growth, real-wage growth and the unemployment rate, to isolate three 'structural' shocks which drove business-cycle fluctuations in a sample of 16 OECD countries during 1950–94. These shocks were interpreted variously as aggregate-demand, productivity and labour-supply disturbances. Two competing hypotheses were assumed for the stochastic behaviour of the unemployment rate: trend stationary, or integrated of order one.

The authors found that the observed dynamic adjustment of the variables in response to structural shocks was much more consistent with the implications of a model with full hysteresis in unemployment than with models of partial hysteresis. In particular, aggregate demand and productivity had almost mirror-imaged positive/negative cumulated effects on output and unemployment, while labour-supply shocks tended to increase unemployment, and productivity shocks were most important in explaining fluctuations in real wages. The authors also concluded that productivity and aggregate-demand shocks explained most of the fluctuation in real output, the former being particularly important in the Anglo-Saxon and Scandinavian economies. Most remarkable, however, was the contribution of aggregate-demand shocks to the forecast error variance decomposition for unemployment for a large subset of countries, comprising the above-mentioned economies and some central European economies.

Some discussion took place about the assumption of modelling unemployment as an I(1) process. Although this may be a good approximation for the economies and sample periods used, in the long run it would be an I(0) process; it was possible, however, that unemployment over a specific period might be interpreted as being I(0), but with a mean shift.

Henri R Sneessens (IRES, Université Catholique de Louvain and Faculté Libre des Sciences Economiques, Lille) presented 'Demand-supply interactions and unemployment dynamics: is there path dependency? The case of Belgium, 1955–1994', which was written with Fatemeh Shadman-Mehta (IRES, Université Catholique de Louvain). The fact that the proportion of capacity-constrained firms in European economies had remained fairly small suggested that capacity shortages could not be the direct and single cause of unemployment persistence. Inferring from this observation that low investment rates played no role in explaining the persistence of high unemployment rates, however, may fail to take into account the relationship between demand expectations, capital accumulation and real rigidities.

The authors' objective was to show how a continuum of excess-supply equilibria could be obtained in models with real rigidities, similar to those traditionally used to analyse the determinants of equilibrium unemployment (the NAIRU), provided the effect of a 'capital gap' on wage formation and the relationship between capital accumulation and demand expectations were explicitly taken into account. Such a persistence mechanism might explain why current unemployment rates seemed to depend so much on past history, and why there was not a simple explanation that applied equally well to all European countries. The authors estimated an econometric model using post-war Belgian data, with special emphasis on the effects of the 1982–7 wage controls. They estimated the model using Johansen's maximum likelihood approach, thus imposing as few a priori structural restrictions as possible. Their results suggested path-dependency and were thus compatible with the implications of their theoretical model. Dennis Snower questioned whether there was a sales constraint, implying wage and price inertia, to which Sneessens replied that the model implicitly contained menu costs, and hence nominal inertia.

Giorgio Brunello (Università degli Studi di Udine) presented 'Regional Disparities and the Italian NAIRU', written with Claudio Lupi (Instituto di Stidi per la Programmazione Economica (ISPE), Roma). In this paper, the authors estimated the Italian NAIRU using annual data for the period 1951–96. They found evidence that aggregate wage-setting in Italy depended only on the rate of unemployment prevailing in the northern and central areas of the country. Southern unemployment did not affect wage pressure. There was evidence of a co-integrating relationship between unemployment in the northern and central areas, the tax wedge, the real interest rate and a measure of union power.

The NAIRU estimates were not very precise, and the width of the 95% confidence interval was close to two percentage points. The NAIRU had remained more or less constant in the past ten years, in association with a reduction in union power and in the real interest rate, which was almost completely compensated for by the increase in the tax wedge. However, aggregate equilibrium unemployment had increased by 8 percentage points and actual unemployment had increased by 15.2%. These numbers suggested that both regional-mismatch and demand factors had been important in the increase of Italian unemployment since the early 1980s. Dennis Snower suggested that if the mismatch was roughly measured by the difference between average and northern unemployment (and therefore between southern and northern unemployment), then changes in mismatch could be regarded as exogenous. If this were the case, the model could be extended to endogenise southern unemployment.

Hector Sala-Lorda (Universitat Autonoma de Barcelona) presented his paper on 'The natural rate and lagged adjustment processes in Spain'. The paper represented a first attempt at using Chain Reaction theory to analyse the Spanish labour market. The objective was to understand how much the dynamic effects of different kinds of shocks could explain unemployment in Spain. The analysis was done via two different types of exercises. First, Sala-Lorda used unemployment persistence and imperfect responsiveness measures to evaluate the degree of persistence of the effects of temporary and permanent shocks on unemployment. Second, he estimated which part of the unemployment trajectory could be explained by lagged adjustment processes and which by the impulses of growing exogenous variables. When analysing the contribution of lagged adjustment processes to the change in unemployment, he found that the dynamics of the system accounted for one-half of this change. The interplay between the succession of impulses and the dynamic responses may have been explained by the medium-run unemployment changes.

Sala-Lorda suggested that these results showed that Chain Reaction theory provided a useful framework for analysing the causes of unemployment and thus for deriving an appropriate design of unemployment policy. Juan Dolado suggested that, rather than looking at the entire business cycle, it might be beneficial to consider the peak and the trough of the cycle separately. Dennis Snower thought it might be desirable to ensure that there was a reasonably close fit between the Spanish unemployment 'story', factors such as the 1978 tax reform, the introduction of fixed-term labour-market contracts in 1984 and the empirical analysis.

'Structural unemployment and the NAIRU in Austria: theoretical considerations and empirical results' was presented by Andreas-Ulrich Schuh (Ministry of Finance, Vienna) and written with Karl Pichelmann (Institute for Advanced Studies, Vienna). Their paper gave a brief survey of the main theoretical and empirical issues surrounding the NAIRU concept. According to modern labour-market literature, NAIRU was defined as the rate of unemployment at which inflation stabilizes in the absence of any wage-price surprises. Conventional thinking about the equilibrium unemployment rate assumed that, in the long run, the NAIRU is determined solely by supply-side factors in the labour market. The authors showed that quite complex adjustment dynamics may arise even in simple log-linear wage-price models. Furthermore, they surveyed a number of hysteresis mechanisms which could lead to permanent shifts of equilibrium unemployment over time, implying that a unique long-run NAIRU may not even exist.

In addition to these theoretical issues, the authors referred to two serious problems which might arise with empirical applications of the NAIRU concept. First, various empirical studies suggested that the results depended highly on model specifications. Second, a considerable amount of statistical imprecision was inherent in the results obtained from empirical estimates. For these reasons, they argued, policy conclusions drawn from the NAIRU concept must be judged with the utmost care, particularly in the many countries in which a number of labour-market measures, as well as monetary policies, were based on the concept. Dennis Snower questioned how important the static natural rate of unemployment, as well as the lagged adjustment processes, were in accounting for the observed movements in unemployment.

David Mayes (Bank of Finland) presented 'Unemployment in a small open economy: Finland and New Zealand', written with Jouko Vilmunen (Bank of Finland). This paper sought to make a contribution to the understanding of the causes of unemployment and its successful reduction by exploring the examples of two rather different small open economies, Finland and New Zealand. They hoped to shed light on three key issues: (1) the degree to which unemployment had been the result of slow adjustment to large external shocks; (2) the degree to which differences in labour-market structures could lead to different responses to shocks; and (3) the importance of the exchange rate and the external sector in resolving the problem.

For Finland, the bulk of the short-run adjustment to a shock to the long-run unemployment relationship fell on unemployment. The speed of adjustment of unemployment was also quite fast, whereas the response of real wages was perhaps surprisingly small: unemployment would adjust at the speed of 0.34 percentage points per six months to a unit shock, while the adjustment speed of real wages was as low as 0.01 percentage points per six months. In the case of New Zealand, the evidence was more difficult to interpret because of some uncertainty about the parameter estimates. Terms of trade, however, appeared to be an important adjustment channel to shocks to the long-run equilibrium in the New Zealand economy. On the relevance of productivity growth in New Zealand, which was raised in discussion, Mayes pointed out that, although it may be the case that the actual outcome for the service sector was more favourable than the impression given by the official statistics, there were many new jobs in lower productivity areas. Unlike many other countries, however, these jobs were not generally of a part-time nature. 

The final paper, 'Unemployment persistence in a small open labour market: The Irish case', was presented by Brendan Walsh (University College Dublin). The question underlying the paper was whether the behaviour of the Irish unemployment rate since the 1960s was best understood in terms of the lagged adjustment of a stable equilibrium unemployment rate (NAIRU) to shocks, or in terms of variation over time in the equilibrium rate of unemployment. Walsh presented evidence from a variety of sources to demonstrate that the relevance of the NAIRU concept was not easily established under Irish conditions. The price-inflationary process was generally agreed to reflect the influence of world inflation, transmitted through the exchange rate. Although domestic inflationary pressures, emanating from wage costs, had been shown to play a minor role in Irish price inflation, it had proved difficult to establish a link between these pressures and the rate of unemployment.

Alternative approaches to defining an 'equilibrium' rate of unemployment had drawn on the relationship between Irish and UK unemployment. A stable relationship appeared to have existed between these two variables until the mid-1970s. Although it subsequently broke down, this link now appeared to have been re-established, but with a significant negative trend in the gap between the two rates. This was attributed to favourable developments in the Irish economy, notably the steady improvement in competitiveness, changes in the wage-bargaining process and reforms in the social welfare system. It remained to be seen how rapidly and how far the Irish unemployment rate would fall in response to further rapid economic growth.