Europe is worried that its regional disparities are growing, not disappearing. Since 1987 annual spending on regional policy has doubled to 14 billion ECU, and further increases were agreed at the 1992 Birmingham summit.

There may be something to this fear: some of the Commission's own studies suggest an increase in regional disparities in the 1980s. During this decade Europe undertook two major policy changes, each with potentially profound effects on regional disparities. Both the `Southern enlargement' and the implementation of the internal market programme represented significant liberalizations of trade and factor movements. Conventional analysis suggests that such liberalizations should favour convergence of output (although not necessarily of income) across regions. The convergence hypothesis has received some empirical support, but recent theoretical work by CEPR Research Fellows such as Paul Krugman and Tony Venables suggests that if increasing returns or `agglomeration effects' are important, trade liberalization may not lead to regional convergence. Instead, by lowering the costs of moving goods around Europe, liberalization may encourage firms to move production away from the peripheral regions in order to take advantage of economies of scale at more central locations.

Neven and Gouyette first study the behaviour of regional output per head relative to EC average output. The behaviour of this `relative output' variable over time indicates how quickly output in a region responds to a shock hitting that region. Regional output in Europe recovers extremely slowly: after 30 years, only 25% of a shock will have been absorbed. The regions studied by Neven and Gouyette are smaller than US states and so should absorb shocks more quickly, but an American study by Blanchard and Katz indicates that states respond much more quickly, taking only 10 years to absorb 43% of a shock.

Though useful, this exercise doesn't tell us directly whether European regions have converged or diverged over time. The simplest definition of convergence involves the change over time in cross-section dispersion, usually of output per capita. The Commission uses this measure in its reports on regional developments in the Community. Convergence is said to occur if the standard deviation of per capita output (across regions) falls over time. Whether with this or more sophisticated measures,Neven and Gouyette obtain results on regional convergence that are essentially the same.

They calculate the standard deviation of per capita output across EC regions for each year between 1980 and 1989. The standard deviation for all regions rises and falls erratically in the early 1980s, but shows a tendency to converge after 1984. Studying Europe as a whole is misleading, however: convergence looks very different if the North and the South of Europe are considered separately. The standard deviation of per capita GDP across regions in the North falls sharply in the second half of the 1980s. In the South, on the other hand, dispersion fell in the early 1980s, but increased after 1985.

The importance of distinguishing between the South and the North might seem a little surprising: most discussions of convergence or cohesion focus on the `centre' versus the `periphery'. Neven and Gouyette test whether the centre-periphery distinction is useful by recalculating their convergence measures after splitting their sample of regions according to the value of a `periphery index', which measures the relative accessibility of European regions. The results deteriorate: the distinction between North and South matters for convergence, that between centre and periphery does not.

What might explain the convergence among the Southern regions in the early 1980s and among the Northern regions after 1985? The accession of Spain, Greece and Portugal to the Community and the inception of the single market programme represented a major shock, affecting the entire Community: perhaps some regions adjusted more quickly to the shock than others.

The US experience suggests that factor movements play a major role in convergence. When a US state is hit by a negative shock, unemployment initially rises. This causes an exodus of workers: unemployment returns to a normal level because of this fall in labour supply, not because new jobs are created. This argument suggests that the lack of adjustment observed in the EC South might be due partly to a relatively immobile labour force. Neven and Gouyette test this hypothesis using 1985 data on migration flows between pairs of regions within countries (there is no comparable data on flows between regions in different countries). Relative wage rates seem to influence the number of migrants moving from one region to another. Holding relative wage rates constant, the tendency to migrate between regions seems stronger in the North than in the South, and is particularly strong in the United Kingdom and weak in Italy.

Neven is Professor of Economics at the Université de Lausanne, co-director of the Industrial Organization programme at CEPR and Assistant Editor of Economic Policy.