At a CEPR/IEEG workshop on ‘Rising Inequalities’, held in La Coruña on 14/15 February 1997, participants reviewed recent theoretical and empirical work explaining shifts in inequalities in income and unemployment. The workshop was organized by Daniel Cohen (ENS, CEPREMAP and CEPR) and Thomas Piketty (CEPREMAP and CEPR). Financial assistance was provided by the Pedro Barrié de la Maza Foundation and the Instituto de Estudios Economicos de Galicia.

In ‘Tax Incentives for Youth Employment’, Michael Kremer (MIT) studied the consequences for work incentives and unemployment of a cut in marginal tax rates for youth. Data from the CPS (Current Population Survey) showed that the number of workers for whom work incentives were distorted per dollar of revenue gained was six times greater for 17 to 21 year-olds than for older workers. As a result, a reduction in marginal tax rates on the young, combined with revenue-neutral increases in marginal tax rates on prime-age workers, would increase work incentives for youths by six times as much as they would decrease work incentives for older workers. Moreover, the benefits of improving work incentives are greater for youths because of their higher labour-supply elasticity. But reducing marginal tax rates on the young is problematical in that it may distort the choice between education and work. If education subsidies are close to optimal, a small reduction in marginal tax rates for the young will cause second-order welfare losses from distortions of education decisions and first-order welfare gains from the improvement in youth work incentives. Conversely, if the problem of distortion of educational decision is important, then tax reductions could be restricted to age groups in which few people attend school.

Andrea Ichino (European University Institute, Firenze, and CEPR) noted that there exists an important literature on tax credits to employers. Such credits are more advisable if the problem lies on the demand side, while cuts in marginal tax rates for employees are preferable if the problem lies on the supply side. In an equilibrium framework, however, tax credits to employers and employees have the same effects. Thomas Piketty noted that when a government cuts marginal tax rates on the young, this may be because decision-takers have in mind a very crude model of linear tax rates, whereas Kremer’s paper pointed to the importance of hazard rates. Finally, Daron Acemoglu (MIT and CEPR) suggested that since child poverty is a more important problem than youth unemployment in the United States, it would be better to subsidize poor families with children.

‘Sustaining Fiscal Policy Through Immigration’, presented by Kjetil Storesletten (IIES, Stockholm, and CEPR), examined the effect of immigration on fiscal policy: Storesletten argued that changing immigration policy and admitting more young immigrants is an alternative to an increase in tax rates and/or a cut in spending when increasing social security obligations create a need for fiscal reform. The condition for immigration to have a fiscal impact is that immigrants differ from natives. In fact, they do differ from natives in age, fertility and employment. In a general equilibrium model, the following fiscal impacts of new working-age immigrants were identified: government spending per capita decreased, tax revenue per capita rose, and debt per capita decreased. The paper also computed the smallest increase in annual immigration needed to balance the budget, given that the government is free to choose the age structure of new immigrants and that current tax and spending policies remain unchanged. The increase was modest relative to the fiscal reform necessary if immigration policy was kept unchanged.

Daron Acemoglu noted the importance of other general-equilibrium effects, such as employment effects. Gilles Saint-Paul (Universitat Pompeu Fabra and CEPR) argued that the results would be the same if the growth rate of population were to increase: only the first generation would benefit from the increase because the capital-to-labour ratio would eventually return to the same value.

In ‘What Level of Redistribution Maximizes Long-Run Output?’, Roland Bénabou (New York University and CEPR) discussed the relative efficiency and long-run effects of two redistributive policies: progressive income taxation and education subsidies. Both policies have costs (due to distortions in agents’ effort) and benefits (due to the lack of credit and insurance markets). Bénabou analysed this trade-off, both theoretically and quantitatively, in a stochastic model of human capital accumulation with endogenous labour supply and missing credit and insurance markets. A first result showed that each income-tax or education-finance policy can be associated with a combination of consumption taxes and investment subsidies that restores investment to its optimal level. A second result was that progressive education finance was more efficient for maximizing long-run output and led to less inequality than progressive income taxation. Finally, when simulating the model with parameter estimates from the empirical literature, long-run output was maximized by an average marginal tax rate of 25% or by an equalization rate for education expenditures of 60%. In both cases, the top one-third of households subsidized the education of the bottom two-thirds.

Gilles Saint-Paul suggested using a richer taxation structure to run similar simulations. François Bourguignon (DELTA) noted that the conclusion that more public expenditure leads to more mobility seems to contradict evidence: the level of public expenditure for education is lower in the United States than in Europe, while intergenerational mobility is higher.

The paper by Richard Blundell and Ian Preston (both University College London), entitled ‘Consumption Inequality and Income Uncertainty’, was motivated by the debate over the use of consumption versus income in studies of inequality growth. In a formal inter-temporal setting, where borrowing and saving are allowed, it is usually believed that consumption expenditure reflects expected lifetime resources better than current income. This paper argued that, though neither consumption nor income is perfect as a measure of inequality, both measures can be combined to understand changing trends in inequality. More precisely, the paper showed that a comparison of the relative growth in income and consumption inequality can separately identify changes in the permanent and transitory components of inequality: a faster growth in income, than in consumption, inequality suggests a growth in short-run income uncertainty; an acceleration in the growth of consumption inequality suggests a rise in the permanent component of inequality.

Using data from the 1968–92 UK Family Expenditure Surveys, Blundell and Preston showed there had been an acceleration in expenditure inequality (indicating an increase in permanent inequality), and that income variance had increased faster than expenditure variance in the latter half of the 1980s (suggesting rising short-term uncertainty). Younger cohorts therefore faced greater permanent inequality and greater short-term income uncertainty. François Bourguignon asked why only income and consumption variances had been used, and not the covariance between income and expenditure, because this covariance could be informative about the variance of permanent income.

In ‘Low-Paid Workers: Some Figures for Italy, 1977–95’, Andrea Brandolini and Paolo Sestito (both Banca d’Italia) addressed two questions: is there a low-wage problem in Italy? and has the situation worsened over the last decade? Using the Bank of Italy’s survey of households’ income and wealth, the paper documented changes in the overall earnings distribution and looked at the relationship between low wages and poverty. The analysis showed that earnings inequality decreased in the 1970s and the 1980s, but increased between 1989 and 1993. The particularly large rise in inequality between 1991 and 1993 could be explained by the abandonment of the wage-indexation mechanism that had been responsible for the earlier decline in inequality. Earnings inequality for prime-age non-farm workers began to rise after 1983, however, well before the overall increase in inequality began. Moreover, the share of inequality explained by between-group inequality had declined since 1990. Another result was that holding a permanent job seemed no longer sufficient for avoiding poverty.

Dominique Goux and Eric Maurin (both INSEE, Paris) examined the comparative effects of the French system of post-school training on wages, workers’ mobility and wage inequalities in their paper entitled ‘Train or Pay: Does it Reduce Inequalities to Encourage Firms’ Investments in Workers’ Training?’. They pointed to some unique features of the French system: in particular, since 1971, firms have had to devote a fraction (1.5%) of their wage bill to training their employees, or pay a tax. The authors examined data from the French Survey on Education and Qualifications and were able to match the workers and their firms, thereby enabling them to control for firm-specific compensation policies.

On average, the paper found that firms spent more than the legal minimum on training, but the level of expenditure depended on firm size. The percentage of trained workers increased with position and education, but the probability of being trained did not depend on education. Concerning employee returns to training, the paper showed that a trained worker was paid 5% more than the equivalent untrained worker. When taking into account firms’ wage and training policies, however, the return to training fell to 3%; and when controlling for both firms and individuals, the return to training fell close to zero. Finally, company-sponsored training was found to increase job tenure. Daron Acemoglu noted that the corresponding German survey inquired whether the training would be useful in another firm, whereas in France no attempt is made to ascertain whether the training is specific or general.

Daron Acemoglu’s own paper, ‘Good Jobs Versus Bad Jobs: Theory and Evidence’, was motivated by the presence of persistent and large wage differentials among identical workers in different industries and occupations. This led him to ask: what determines the composition of jobs?; why does this composition change over time? and; why do some labour markets have more bad jobs than others? The paper presented a search model in which different types of jobs have different creation costs – those that cost more to create will pay more. Thus the economy is naturally composed of good and bad jobs. Acemoglu showed that the composition of jobs will always be suboptimal (too many bad jobs) because good jobs cost more to create and firms do not necessarily receive the full marginal product of their investment. A further finding was that an increase in unemployment benefits improves job composition and may also improve welfare. This is because some workers, who were previously accepting bad jobs, will now prefer to wait for a good job, and this change in search behaviour induces creation of more good jobs. An increase in the minimum wage will also lead to a better job composition. The empirical evidence suggests that, in the United States, the impact of higher unemployment benefits and minimum wages may be to increase unemployment somewhat but also to improve substantially the composition of jobs.

In ‘The Distribution of Earnings in Spain during the 1980s: The Effects of Skill, Unemployment and Union Power’, Manuel Arellano (CEMFI, Madrid), Samuel Bentolila (CEMFI, Madrid, and CEPR) and Olympia Bover (Banco de España) provided the first account, based on micro data, of the evolution of wages in Spain. The authors used the panel of monthly social security records of individual earnings from 1980 to 1987. During this period, high-skill real earnings increased by 1.42% while low-skill and median real earnings decreased (respectively by 0.37% and 0.22%); thus overall inequality expanded. This outcome was due to increasing dispersion in the top half, rather than the bottom half, of the wage distribution. In addition, the return to skill increased for high-skill and medium-skill workers and decreased for the low-skilled. Another finding was that wages rose with union coverage, more so for medium-skill workers than for the low-skilled.