Labour Economics
Unemployment and Wages

What accounts for the persistence and duration of unemployment? Why does the wage-setting process fail to cushion market economies from large swings in unemployment? How do interest rates, productivity growth, capital market imperfections and the exercise of market power in labour product markets affect unemployment? How do the process of job search, the creation of new firms, and segmentation of labour markets influence the time-path of unemployment? Over the past decade economists have provided increasingly divergent answers to these questions. Some of the main proponents of the rival unemployment theories discussed their predictions and policy prescriptions at a joint CEPR/NBER conference on `Unemployment and Wage Determination' in Cambridge, Massachusetts, on 17/18 October. The conference was organized by Dennis Snower, Co-Director of CEPR's Human Resources programme, and Joseph Stiglitz, Professor of Economics at Stanford University. Financial support was provided by a grant from the Margaret and Robert Merrell Foundation to the CEPR International Foundation and by the UK Department of Employment to CEPR. Funding to the Bureau was provided by the Starr Foundation.

Papers presented at the conference included theoretical and empirical analyses of the labour market, focusing on where the different approaches lead to observationally distinct predictions and differing policy proposals. Major contributors to the current literature on unemployment presented the cases for their different research agendas and at the same time sought common ground. There was broad agreement that an adequate theoretical explanation of unemployment must account for a variety of empirical observations:

During recessions, wages do not fall far enough to prevent large increases in unemployment, and laid-off workers are unwilling or unable to work for lower pay in order to keep their jobs.
Current and past unemployment are closely related, and this `unemployment persistence' is more pronounced in Europe than in the US.
Productivity and the size of the labour force may affect unemployment in the short run, but apparently not in the long run: the US and several European countries have witnessed massive long-run growth of productivity and the labour force, but their unemployment rates have always returned to within a range of approximately 2-6%.
Changes in unemployment and the interest rate are often inversely related.
Vacancy and unemployment rates have typically been inversely related over the post-war period in the US, but not in Europe.
Individual workers' chances of finding employment typically fall as the duration of their unemployment increases; and the longer is the average economy-wide duration of unemployment, the smaller is the influence of unemployment on real wages.
Reductions in employment usually take the form of lay-off rather than reductions to hours worked; this tends to be more pronounced in the US than in Europe.
In upswings, employment rises in high-wage relative to low-wage industries while the corresponding relative real wages fall.

Efficiency Wage Theories
The first session was devoted to `efficiency wage' theories, in which unemployment arises because it is not in the firms' interest to reduce wages to the market clearing level. Joseph Stiglitz (Stanford University and NBER) opened the session with a paper on `Unemployment and Efficiency Wages: The Adverse Selection Model', written jointly with Andrés Rodrigues. A firm may be unwilling to reduce its wages if it expects that its best workers will then quit and other good workers will be discouraged from applying. Workers have better information about their own productivities than their employers, so the more productive workers have the more attractive fall-back opportunities and are therefore willing to demand a higher wage and accept a greater risk of unemployment. Firms therefore find it worth while to pay higher wages than are necessary to clear the labour market. In this context, Stiglitz and Rodrigues find that the labour market equilibrium is characterized by a single efficiency wage if there are no search costs and applications are unlimited; but in other cases the equilibrium involves a wage distribution. They also analyse how unemployment can arise when workers have no information about their productivities, but firms can acquire it by testing. Then firms will pay wages above the market-clearing level to acquire a larger applicant pool, to discourage other firms from `skimming' their best workers and to avoid hiring the worst workers.
Andrew Weiss (Boston University) suggested extending the model of testing to the case where firms minimize their expenses by using random examinations. Further, a model that failed to break the pooling equilibrium might be more realistic and interesting.
In his paper on `Interest and Wealth in Incentive Wage Modelling of the Natural Rate of Unemployment', Edmund Phelps (Columbia University) sought to account for prolonged periods of high unemployment with a dynamic model of workers' shirking and saving decisions. In particular, he showed how high unemployment may be associated with a high rate of interest. A higher interest rate implies that the future benefits from not shirking are discounted more heavily, so a higher wage is required to discourage shirking. Over the long run, however, productivity increases arising from labour-augmenting technological progress lead to increases in wealth and wages, which have countervailing effects on shirking. This helps to explain why the unemployment rate is independent of productivity in the long run.
Christopher Pissarides (LSE) commented that Phelps's model translates the traditional demand-for-leisure effects into unemployment changes. If productivity and human capital grow at the same rate while physical capital grows only at the savings rate, however, productivity growth may reduce shirking and increase the labour supply.

Insiders, Outsiders and Unions
The second session focused on `insider-outsider' and union theories a natural counterpart to the efficiency wage theory, in which unemployment arises because workers' market power raises wages above the market-clearing level. In a paper on `Patterns of Unemployment: An Insider-Outsider Analysis', Assar Lindbeck (Institute for International Economic Studies, Stockholm) and Dennis Snower (Birkbeck College, London, and CEPR) extended the insider-outsider analysis to labour markets in which gross flows of workers into and out of firms far exceed net changes. In the presence of random productivity shocks, they find that labour turnover costs do not eliminate these large gross flows; rather they influence firms' hiring and retention rates. Unemployment persistence depends positively on the retention rate of incumbent insiders and the hiring rate of insiders from other firms but negatively on the hiring rate of entering outsiders. Their analysis also shows how differences in labour turnover costs, quit rates and labour force growth rates may help explain why unemployment persistence tends to be higher in Europe than in the US. It also indicates that in a slump firing costs may be expected to have a greater impact than the discouraged worker effect on unemployment persistence. Finally it explains how long-run adjustments in the number of firms, labour turnover costs and unemployment benefits can make the long-run unemployment rate independent of the levels of productivity and the labour force.

Christopher Pissarides suggested that a further explanation of unemployment persistence may be found in the persistence of hiring rates.
Stephen Nickell (Institute of Economics and Statistics, Oxford, and CEPR) presented his joint paper with Jari Vainiomaki and Sushil Wadhwani on `Wages, Unions, Insiders and Product Market Power', which used data from 800 UK manufacturing firms to examine the roles of insiders and market power in wage determination. He found that firm-specific factors such as productivity gains influenced wage increases, but he obtained no robust evidence of insider power effects. Conditions in the external labour market have an important influence on wages, particularly when firms have little market power. Product market share also has a positive impact on wages, which is more pronounced in large firms but is not influenced by union status.

Mark Bils (University of Chicago and NBER) noted that wages reflect only 15% of productivity, so there may be other explanations for high-productivity firms' tendency to offer high wages. He suggested improving the model's realism by also examining the effects of changes in capital and the number of insiders and differentiating between insider and entrant wages. Dennis Snower argued that it is inappropriate to measure insider power through the effect of lagged unemployment on the real wage. As the number of insiders increases, other things unchanged, insiders' job security may fall, while their market power increases. The first effect depresses wages whereas the latter effect raises them; the net effect is ambiguous.

In a paper on `Productivity Growth, Wage Setting and the Equilibrium Rate of Unemployment', Alan Manning (LSE and CEPR) argued that reduced productivity growth may be an important factor in explaining the long-run rise in OECD unemployment since 1973. In his model, incumbent employees have a higher chance of remaining employed than outsiders have of becoming employed. When falling productivity growth reduces incumbents' need to safeguard their employment this year to improve their chances of being employed next year, they exercise less wage restraint and unemployment increases. Manning then used this model as the basis for an expectations-augmented Phillips curve, which he then estimated for 19 OECD countries during 1956-85. His results indicated that the slow-down in productivity growth was an important factor in the rise of unemployment.

William Dickens (University of California at Berkeley and NBER) questioned the robustness of Manning's estimated growth coefficients. Assar Lindbeck noted that Manning's intertemporal substitution hypothesis is the opposite of the one proposed by Lucas: Lucas argues that the prospect of higher wages tomorrow leads workers to take more leisure today, while Manning argues that they will take less leisure.

Persistence and Costs of Adjustment
Bruce Greenwald (Columbia University) and Joseph Stiglitz (Stanford University) opened the third session, on persistence and costs of adjustment, with a paper on `Capital Market Imperfections and Labour Market Adjustments'. They attribute high and persistent unemployment to firms' inability either to diversify unsystematic risks or raise equity finance because of informational problems in financial markets. Such firms make partial adjustments to uncertain changes in business conditions, and Greenwald and Stiglitz view their labour market decisions as portfolio decisions in which their risk-and-return trade-off depends on their `wealth' positions. When a firm's output depends on its workers' effort, which is not directly observable by their managers, short-run adjustments to unfavourable shocks lead to reductions in employment rather than wages or hours, and such reductions will be concentrated on the smallest possible number of employee groups, as in lay-off policies based on seniority.

Dennis Snower suggested testing this theory's implicit empirical prediction: that wages and hours adjust more rapidly relative to employment in sectors where monitoring is more imperfect.

Giuseppe Bertola (Princeton University, CEPR and NBER) and Ricardo Caballero (Columbia University and NBER) then presented their paper on `Efficiency and the Natural Rate of Unemployment: Labour Hoarding in Matching Models', which focused on the steady-state labour market equilibrium where firms undergo idiosyncratic shocks and hire workers only through a costly search process. Mobility costs impede labour from relocating to high-productivity sectors, while low-productivity firms tend to hoard labour, waiting for better times. This reduces equilibrium unemployment but also decreases economic efficiency. Job-security provisions and workers' bargaining power interact with the matching technology and sectoral shocks to determine the natural rate of unemployment, output, wage schedules, allocation of labour and labour hoarding. In particular, equilibrium output is positively related to mobility costs.

Bruce Greenwald objected to the unrealistic requirement that workers pass through unemployment to change jobs, whereas in practice most search occurs while they retain their current jobs. He also disputed the assumption that workers are homogeneous: search occurs precisely because firms seek good employees and workers seek good jobs. Stephen Nickell observed that in practice firms can alter the level of unemployment by adjusting hours without resorting to firing.

Search Theories
The fourth session considered `search' theories, in which unemployment arises because imperfect information prevents workers and jobs from being properly matched. Olivier Blanchard (MIT and NBER) opened the session with his joint paper with Peter Diamond, `Ranking, Unemployment Duration, and Wages'. `Ranking' refers to the practice that firms receiving multiple acceptable job applications hire the worker who has been unemployed for the shortest period. This clearly implies that workers' chances of securing employment fall as the duration of their unemployment rises, which accords with the empirical evidence. Blanchard showed that this effect becomes stronger as unemployment rises, while the presence of long-term unemployed per se has little effect on wages. Relative to random hiring, ranking raises wages and makes them less dependent on the level and more on the change of unemployment.

Reuben Gronau (Hebrew University of Jerusalem) argued that firms would realize that ranking raises wages and therefore have an incentive to avoid it, while the tendency of workers' acceptance wages to fall the longer they remain unemployed should raise their chances of gaining employment.

Christopher Pissarides (LSE) then presented his joint paper with Dale Mortenson on `Job Creation and Job Destruction in the Theory of Unemployment', in which they viewed each job as an irreversible investment by the firm, which must destroy some jobs and create others to alter its production mix. Their model is consistent with empirical findings that substantial job creation and destruction proceed throughout the business cycle and that much new job creation is undertaken by existing firms. They suggest that the vacancy unemployment flow data are potentially more useful in distinguishing between aggregative and reallocational shocks than data from Beveridge or vacancy supply curves, since the latter depend on a similar set of parameters.

Bruce Greenwald argued that this model like other search models pays inadequate attention to the microeconomic foundations of the matching technology and therefore assesses neither its responsiveness to cyclical shocks nor its interaction with firms' incentives to offer wage distributions.

In a note on `The Segmented French Labour Market', Edmond Malinvaud (Collège de France) considered the segmentation of the labour market into permanent and temporary jobs. In Europe this is not related to productive sectors, since the same tasks are often performed by employees on permanent and temporary contracts working side by side. In France, many workers holding temporary jobs move on to permanent ones, so temporary contracts do not appear to stigmatize the job holders. Temporary jobs may arise in response to a slack labour market: in a recession, workers may reduce their wage claims and be more ready to accept temporary jobs, while employers take the opportunity to save on labour turnover costs and increase their employment flexibility in response to uncertainties. Malinvaud also highlighted the conflict between labour market institutions that seek to achieve worker protection and those that aim to promote employment flexibility.
Olivier Blanchard suggested that technological change may have created a need for increased flexibility of employment, in which case the new regime will change little when tighter markets return. Also life-cycle considerations may induce workers to seek flexibility early in their careers but prefer the security of permanent jobs later.

Presenting their paper, `Labor Market Segmentation, Wage Dispersion and Unemployment', Kevin Lang (Boston University and NBER) and William Dickens (University of California at Berkeley and NBER) developed a model of a segmented labour market in which unemployment is more prevalent among low-skill workers who also take a disproportionate share of low-wage jobs. Their model predicts that wage differentials will often exceed differences in productive potential and that aggregate unemployment will be positively correlated with the average wage. Larry Katz (Harvard University and NBER) argued that there was no systematic relation between unemployment and wages in urban areas. He referred to work by Blanchflower and Oswald that documents a negative empirical association of wages and unemployment.

In a paper on `Inter-Industry Mobility and the Cyclical Upgrading of Labor', Mark Bils (University of Chicago and NBER) and Kenneth McLaughlin (University of Rochester) first noted that aggregate expansions are associated with upgrading in the labour market, while employment is more cyclical and real wages less cyclical in high-wage than in low-wage industries. Using PSID data, they estimated a market-clearing Roy model for 25 industries. They concluded that even if the market-clearing model can explain the above observations, it fails to explain the relative wage movements of movers versus stayers. Relative wage changes are also extremely unresponsive to cyclical or other fluctuations in employment across industries, which cannot be reconciled with their estimates of the market-clearing model. The observed wage gains of about 5-10% for workers entering high-wage, cyclical industries are consistent with some queuing at these industries.
Larry Katz recommended further testing with queuing models or a Roy-type model with compensating differentials or learning. He also cautioned that the data do not distinguish between selective lay-offs and those due to plant closure, which convey little information about workers' quality.

Empirical Studies
The final part of the conference was devoted to empirical assessments of the causes of unemployment in the US and Europe. In their joint paper with Larry Summers, `Britain Divided: Hysteresis and the Regional Dimension of Britain's Unemployment Problem', Edward Balls (Financial Times) and Larry Katz (Harvard University and NBER) observed that unemployment in the `North' has remained high despite a full recovery of the vacancy rate. Understanding the relative rise of unemployment in the North therefore requires an explanation of the unemployed's unwillingness or inability to take the jobs that are available, not of why few jobs are available. Explanations based primarily on labour immobility or on unions' setting excessively high wages are therefore inappropriate, since they imply counterfactually that the vacancy rate in the North should be very low. The authors instead identified a dramatic reduction in the demand for skilled physical labour as the proximate shock that increased Northern unemployment, while `social hysteresis' (due to changes in workers' attitudes to work) acted as the propagation mechanism.

Assar Lindbeck suggested explaining these results on persistence in terms of a segmented labour market in which the primary sector offers higher wages and experiences an unfavourable labour demand shock. High primary sector wages due to workers' market power may then account for the observed persistence; but the authors had no model of vacancies and their explanation of the relation between vacancies and unemployment was therefore incomplete.

Stephen Nickell (Institute of Economics and Statistics, Oxford, and CEPR) presented his joint paper with Richard Layard on `Unemployment in the OECD Countries', in which they found that both the level of unemployment and its responsiveness to shocks in the post-war period have depended heavily on unemployment benefit systems and wage-setting mechanisms. Unemployment persistence has depended on these factors and also on employment flexibility. Finally, both the nominal flexibility of wage bargaining and the average variance of nominal shocks have had negative effects on the impact of nominal shocks.

Assar Lindbeck endorsed the authors' conclusions concerning the magnitude and duration of unemployment benefits and the need for employer coordination in wage bargaining, but he criticized both the complexity and lack of predictive power of their underlying theoretical model and maintained that the connections between its theoretical and empirical parts were somewhat tenuous.

Jean-Paul Fitoussi (Observatoire Française des Conjonctures Economiques, Paris) presented his joint paper with Edmund Phelps, `An Empirical Study of Alternative Approaches to Unemployment: The Role of the Interest Rate'. Using data on employment, unemployment, production, productivity, real wages, unit costs of labour and the real interest rate in the US and France for 1900-87, they showed that only the real interest rate is cointegrated with unemployment, while Granger causality runs only from the real interest rate to unemployment in both the short and the long run. Further, their VAR analysis of the French data showed that a positive innovation in the interest rate had negative short-run and positive long-run effects on output, which was inconsistent with their theory; but it had the reverse effects on the real wage, precisely as their theory predicted. A VAR analysis may be inappropriate, however, since it describes the effects of an transitory real interest rate shock while the underlying theory considers an increase in the expected real interest rate.

Michael Woodford (University of Chicago and NBER) observed that the oligopolistic pricing model he had developed with Julio Rotemberg predicted interest rate changes to have the opposite effects on the real wage and output, by increasing incentives to deviate from collusive behaviour and hence increasing labour demand at any given real wage. Adding changes in the real exchange rate as an explanatory variable may yield more conclusive results. Other participants suggested that the inverse relation between the interest rate and unemployment may be due to macroeconomic shocks with sluggish nominal interest rates rather than to customer-market effects.

In sum, the conference highlighted how the efficiency wage, insider-outsider, and union theories provide complementary explanations for why wage movements fail to prevent large cyclical swings in unemployment. It explored channels whereby the interest rate, productivity growth and conditions on other markets particularly product market power and capital market imperfections affect the level of unemployment. Furthermore, it identified determinants of the persistence and duration of unemployment and, finally, it helped to clarify how labour market segmentation and the processes of job creation and job destruction influence both unemployment and vacancies.

A selection of papers presented at this conference will appear in a forthcoming Special Issue of the Review of Economic Studies.