European integration and external constraints on social policy: is a social charter necessary?

Ermisch, John

EUROPEAN INTEGRATION AND EXTERNAL CONSTRAINTS ON SOCIAL POLICY: IS A SOCIAL CHARTER NECESSARY?

The completion of the single European market after 1992 could have implications for the conduct of social policy in member countries of the European Community (EC). In particular, more integrated capital and labour markets could limit member countries' freedom of action in social policy.

Limitations could arise because businesses are mobile in the longer term, and an integrated European market will increase their mobility. The European Commission has expressed concern that industries in countries with the least social protection would have a competitive edge, thereby attracting more businesses to these countries and eventually reducing social protection in other member countries, as a consequence of economic competition. The European Commission has called this |social dumping', which creates a tendency for a |levelling down' of social protection in member countries. Operating in the opposite direction would be a tendency for workers to leave countries with poor social protection.

The European Commission's concern with this social dimension of economic integration, particulary the danger of social dumping, has been recognised by the European Commission in the |Community Charter of the Fundamential Social Rights of Workers', or |Social Charter' for short. In the preamble of the final draft (14 September 1989), the Charter states that |...the completion of the internal market must offer improvements in the social field for workers of the European Community, especially in terms of freedom of movement, living and working conditions, health and safety at work, social protection, education and training;....' The Social Charter can be viewed as an attempt to |level up' social protection, assuring that there is no |retrogression compared with the situation currently existing in each member state' (as the preamble of the Charter puts it). The Social Charter's freedom of movement principles also attempt to improve labour mobility within the EC.

All EC governments other than the UK endorsed the Social Charter in a |solemn declaration' (on 9 December 1989). The UK government is concerned that the Social Charter would increase unemployment, and this concern is examined in the paper.

In order to implement the principles contained in the Charter, the Commission adopted a |social action programme' in November 1989. The programme adopts the |principle of subsidiarity', according to which the Community acts where the objectives to be reached can be achieved more effectively at Community level rather than at national level. It is, however, possible that the Social Charter itself violates the principle of subsidiarity.

The purpose of this paper is to examine the conditions under which social policy would be constrained by European economic integration and to assess whether a Social Charter is needed. The first section discusses some principles to follow in deciding whether the EC or a member state should have the power to make decisions in particular areas of policy. It provides a framework for interpreting the |principle of subsidiarity'. The second section examines the potential for a direct effect of social benefits on the movement of people within the EC. In the third section, the impact of the taxes used to finance social policy on the location of businesses and people and the incidence of these taxes are investigated. The degree of labour mobility in response to differences in real wages between EC countries is demonstrated to be crucial in deciding whether a Social Charter is necessary. Thus, a substantial fourth section of the paper examines the evidence on the responsiveness of labour mobility, and it suggests little need for a Social Charter.

1. What level of government for which

functions?

Decisions about government functions, such as the provision of education or social security, could be made at a number of different levels of government (for example, at the local council level, at the national level or by the EC). What is the best level of decision-making for each function? In order to answer we need a framework for evaluating what level would be best. It is assumed that collective decisions at each level are made by majority rule.

The minority on a particular issue or policy suffer a welfare loss because the majority decision goes against their wishes. Such welfare losses are reduced by delegating decision-making (power) to communities that are relatively homogeneous regarding preferences in a particular policy area. (1) Homogeneity is most nearly approximated at the extreme local level; that is, |home rule'. Thus, maximum political consensus requires small political jurisdictions.

Consensus is not, however, the only consideration. If the population affected by the policy is larger than that jurisdiction's constituency, its decisions will affect the welfare of others. Similarly, its population's welfare will be affected by decisions made in other jurisdictions. Failure to take these effects into account will result in inefficient decisions (i.e. some people can be made better off while not making others worse off). Air pollution control is a clear example. It may improve the air of neighbouring communities as well as the jurisdiction deciding about it. Ignoring these benefits results in under-provision of air pollution control. In economists' jargon, their are |external effects' of community decisions.

In policy areas in which collective decisions result in benefits or costs for neighbouring communities, the jurisdiction for decisions should be large enough to include the total population affected by its policies. Thus, for some government functions, a jurisdiction much larger than that justified by |home rule' is required.

Fiscal equivalence

For most goods provided by government it is not feasible to exclude people who do not pay for the good from consuming it. Air pollution control is a good example. The trade-off between consensus and external effects for collective goods such as these can be balanced by defining political jurisdictions for each collective good so that there is a match between those who receive the benefits of that good and those who pay for it. This match has been called the principle of |fiscal equivalence' (Olson, 1969).

Thus, this principle entails a separate jurisdiction for every collective good with a unique |catchment area' for its benefits. (2) When there are decreasing costs per head of providing the particular collective good, decisions would be made in a jurisdiction consistent with the principle, but the provision would be contracted out to more efficient productive units, either public or private. Even in the absence of contracting-out, the fact that a larger population could provide the collective good at a lower cost per head is irrelevant if the benefits of the good do not extend to a larger population. (3)

If, however, there are increasing costs of production, then provision should be organised in smaller, minimum cost per head jurisdictions, and, in addition, a system of grants from a larger jurisdiction, encompassing all beneficiaries of the collective good, to the smaller government units should be set up. These inter-governmental grants would be just large enough to compensate the smaller unit for the external benefits produced by the collective good. Education may be example of such a good. A very large school system may be too bureaucratic and cumbersome, but because of migration the benefits of education extend beyond jurisdictions of a size that would minimise cost per student.

Redistribution

Social policy is often concerned with redistribution of resources. Redistribution raises an additional criterion for the best jurisdictional boundaries: these should contain the appropriate donors and beneficiaries. Whereas maximal consensus requires population homogeneity, redistribution requires population heterogeneity, particularly regarding income.

Mobility of donors and beneficiaries makes it difficult to maintain the desired population mix for redistribution purposes. This suggests that jurisdictions for redistributive policies should be defined to minimise mobility across the boundaries, implying a large jurisdiction in terms of population covered, perhaps as large as the EC if migration between member states is responsive to taxes and benefits.

But how are the redistributive goals decided upon? Preferences concerning these are likely to vary with national/regional |culture', and average income is also likely to affect these decisions. For instance, altruism (caring about the living standards of the less well-off) suggests that higher income regions would want to redistribute more (the |marginal utility' of a person's own income declines with income). On the other hand, voters may be more inclined to vote for more redistribution if they feel they may need it some day, due to uncertainty about their own income, so that poorer regions may vote for more redistribution. (4)

Whatever the reason that different regions/countries choose different amounts of redistribution, welfare losses would result if redistribution policies are decided (by majority rule) at a broader level of government, like the EC, rather than in jurisdictions more homogeneous in terms of cultural background and living standards. (5) Within Europe there are many distinctive cultures and a wide range of living standards. Consensus on redistributive aims would favour each of these setting up a regional or national government for redistribution purposes.

There is a conflict between |home rule' and achieving a given region's or nation's redistributional goals if there is mobility across jurisdictional boundaries in response to the taxes and benefits used to carry out the redistribution. Such mobility would need to be taken into account by jurisdictions in deciding the levels of social benefits and taxes (for example, see Rothenberg, 1970). If, therefore, we wish to determine whether the current member countries of the EC will be constrained in their redistribution policies, international mobility within the EC after 1992 is a crucial issue.

The next section examines the extent to which European people may move across national borders in response to differences in social benefits. This discussion ignores the impacts of the method of financing these benefits on mobility and social policy constraints, but that is taken up in the third section.

2. Impact of social benefits on migration

There are no studies of migration within Europe which measure the impact of social benefits on migration flows, and in any case the responsiveness of migration to economic variables may change after 1992. It is, however, possible to use studies of migration between states of the United States to infer the maximum responsiveness that can be expected in Europe after 1992.

The measured response of inter-state migration in the USA to social benefits is likely to provide a maximum for Europe because America is much more homogeneous than Europe, with a common language and culture throughout the states, and it has had a more mobile population than most European countries. Coupled with the factors favouring mobility is a large variation in social benefits among states. For instance, payments under the main welfare benefits programme, Aid to Families with Dependent Children (AFDC), which primarily helps one parent families, vary enormously among states. In 1979, the state maximum monthly AFDC benefit payment for a family of one needy adult and three dependent children varied from $120 in Mississipi to $524 in Vermont (Blank, 1988, Appendix I). Even in neighboring states like Wisconsin and Illinois, the difference in the maximum payment was $195 in 1979.

This wide variation in AFDC benefits certainly provides large incentives for the population of one parent families to move between states. Other than education, it is the largest social expenditure programme in states and the benefits to recipients are transparent compared to the benefits of many other social programmes. These considerations suggest that we are more likely to find migration in response to AFDC benefits than to any other social policy. On the other hand, one parent families are less mobile than many other groups, such as young men prone to unemployment. Nevertheless, it appears likely that the geographic mobility response to inter-state differences in AFDC benefits is likely to be the higher than the mobility response to any social policy difference found between countries in Europe.

One attempt to measure this response studied the movement of families receiving AFDC between three groups of states: high, medium and low AFDC benefit states (Gramlich and Laren, 1984, table 4). The rates of migration (|transition rates') between these groups of states over a five-year period are shown in table 1. It indicates that while inter-state migration is a rare event (less than 10 per cent moved between the groups of states in the five-year period), when AFDC families do move between states, they are much more likely to go to a state with higher AFDC benefits (note that the entries below the principal diagonal in table 1 are more than twice as large as those above the diagonal). Gramlich and Laren point out that even though migration is very sluggish, this blas in movement can alter the inter-state distribution of the AFDC population substantially over time.

Blank (1988) takes a more direct approach, using data on individual one parent families headed by the mother to study the joint decision of whether or not to receive AFDC and the region in which to live. (6) Her analysis makes it possible to estimate directly the responsiveness of the probability of leaving a region to differences in AFDC benefits and wages. It suggests that a 10 per cent increase in AFDC benefits in a region increases the probability of remaining in the region by 1 per cent (an |elasticity' of 0.1). This is not a very large impact, but it could still be important in the longer run.

To investigate this, the hypothetical |equilibrium' proportion of the AFDC population in each group of states implied by the migration rates in table 1 is computed. These are |equilibrium' in the sense that when these proportions are achieved they would be maintained, (7) and they are shown in the first line of table 2. Blank's estimate of the impact of benefits on the probability of moving to another region is used in conjunction with the pattern of migration rates in table 1 to estimate how the hypothetical equilibrium distribution changes when AFDC benefits change in one region. The impact of a 10 per cent increase in benefits in low benefit states on the equilibrium distribution of the AFDC population is examined. The second line of table 2 shows the resulting equilibrium distribution. It appears that even in the very long run the impact of a 10 per cent rise in AFDC benefits is small. (8) Furthermore, it takes a very long time to converge to the new distribution. It would take about 40 years for half of the adjustment in the proportion of the AFDC population in low benefit states to the 10 per cent rise in benefits to take place.

As suggested above, this small and sluggish change in people's location in response to benefits is likely to be much higher than would take place in Europe. Language and culture represent formidable barriers to movement between European countries in response to differences in social benefits. Moreover, responses to differences in social welfare payments like AFDC are likely to be larger than to differences in other social benefits, because the benefits of welfare payments to individual families are large and transparent relative to other social policies.

Although one parent families are less mobile than other groups, such as young unemployed men, this analysis suggests that such differences between member countries of the EC in social policies would not directly encourage much migration between countries. This favours autonomy for individual countries in deciding social policy.

There could, however, be indirect effects which work through the taxes that need to be levied to finance the social policies (assuming that other public expenditure is not reduced to compensate). Taxes can affect after-tax returns to capital, which is mobile between countries. They also may affect take-home pay, and even though there is little responsiveness to social benefits by the population for whom social benefits are an important part of their income, there may still be large responsiveness by the population to net real wage differences. This would be the case when the part of the population receiving large social benefits is small and/or different from the rest of the population. For instance, the mobility of poor people, for whom these benefits are particularly relevant, may be constrained by their poverty because of the costs of migrating to another country and borrowing constraints.

3. Impact of taxes on location of economic activity

The taxation used to pay for collective goods and redistribution could constrain social policy because businesses and people may move in response to net (after tax) returns and wages respectively, and governments take this responsiveness into account in social policy-making. The incidence of the costs of providing the collective goods and redistribution also depends on the mobility responses and the type of tax used.

In this section, we focus on the effects of taxation, ignoring any impact of the benefits of the social expenditure on location. The benefits are likely to be small for firms, and the previous section suggests that they are also probably not an important factor in people's migration between member countries. Taxes on labour (payroll taxes), capital (corporate or property taxes), or products (VAT on the |origin principle') are examined first.

Labour mobility and the effects of differences in indirect taxes

Payroll taxes, like employers' National Insurance contributions in Britain, are a common method of financing pensions and social security in European countries. The economic analysis of tax incidence and industrial location can be used to investigate the impact of the tax on industrial location. (9) Other laws and regulations that increase the cost of labour, such as employment protection legislation, act as an implicit tax on the use of labour. As such, they would have effects similar to a payroll tax.

The analysis assumes, as is very plausible, that capital is perfectly mobile, in the sense that the after-tax return on capital is equalised acros s countries. It is also likely that the development of a single market increases consumer responsiveness to relative prices, and this is taken into account in the analysis. (10)

When workers move in response to differences in pay between countries, then a higher payroll tax in a country discourages capital investment in the country. Although the tax on labour encourages capital to come in to substitute for higher cost labour, the tax on labour also raises the cost of production, which reduces demand for the product and production, thereby discouraging capital investment. The latter effect dominates when the price elasticity of demand for the country's products is larger than the elasticity of substitution between capital and labour in production, and it is plausible that this would be the case in the integrated European product markets.

The more responsive are workers to pay differentials, the more that the tax discourages investment. If workers are very responsive, they are better able to avoid the tax burden, placing a larger burden on capital, thereby discouraging capital investment in the country levying the higher payroll tax. Taxes on capital or production also have a stronger discouraging effect on capital investment, the more that worker mobility responds to pay differentials. (11)

Apparently, this is the effect that worried the European Commission when they expressed concern about |social dumping'. Social policy formulation would be constrained by this potential movement of businesses in response to higher taxes, and it would be likely to result in lower levels of social benefits and protection (for example, lower social security benefits and pensions). The Social Charter was the Commission's response. Its |levelling up' of social protection would keep businesses in the countries with higher benefits (the richer countries) while protecting those benefits.

The degree of worker mobility is crucial in determining whether there are such constraints on social policy and a need for a Social Charter to guard against social dumping. Taking the extreme case, in which workers are not responsive to pay differences between countries, a higher payroll tax in a country would not affect business location there. Workers in that country would see their wages fall by the amount of the additional payroll tax; in other words, workers would bear the full burden of the tax. (12) This would come about because the demand for labour would fall as a result of the higher payroll tax. At each quantity of labour, businesses would offer a wage that is lower by the amount of the increment to the tax. If workers' mobility responded to pay differentials, some would leave the country, thereby reducing the supply of labour and offsetting some of the fall in the wage. If, however, workers were not mobile, their wage would be remain lower by the amount of the additional tax. As the wage cum tax paid by firms would be the same as before, there would be no incentive for capital to leave the country.

This account assumes that wages in the higher tax country can fall relative to other countries. If, however, such falls are resisted, then unemployment would increase in the higher tax country. But note that declines in wages relative to other countries need not entail a decline in a country's real wage, only a smaller increase than in other countries.

Although it may be extreme to assume that workers are immobile, low responsiveness to pay differentials between countries is plausible, even after 1992, because of the language and cultural differences between member countries of the EC. Rational workers would move to equalise their personal (or family) welfare across countries. As not being able to speak one's native language nor live in a familiar culture represent large reductions in personal welfare, it would take much larger pay in another country to compensate for this loss in welfare. In the aggregate, therefore, it is likely to take large pay differences to elicit small amounts of movement between countries, and the evidence presented in the next section of the paper supports this hypothesis.

In these circumstances of low worker mobility between EC countries, the richer countries, with more social protection, are not putting themselves at a competitive disadvantage relative to poorer countries. More social protection, paid for by higher payroll taxes, is offset by wages being lower than they would otherwise be. In other words, a country's workers have paid for their social protection. As noted earlier, regulations that increase labour costs can be considered as an implicit tax on labour, like a payroll tax. The costs of such regulations are, therefore, also passed on to the workers when labour mobility between countries is not responsive to pay differentials, and these regulations do not affect capital investment patterns.

Payroll taxes are preferable among the indirect taxes in that they distort capital allocation among countries least, particularly when labour mobility is low. For a given tax yield, taxes on production or capital usually discourage capital investment in the taxing country more than a payroll tax. When labour mobility is not very responsive to pay differentials, this is clearly the case, with taxes on capital discouraging investment most (for a given tax yield). (13) This is why the discussion has focussed on the effects of the payroll tax, and only mentioned the effects of taxes on production and capital in passing. The alternative of direct taxation is explored next.

Labour mobility and the effects of differences in direct taxes

The sensitivity of capital movements to differences in after-tax rates of return suggests that income from capital should be taxed on the residence principle; that is, it should be taxed in the country in which the owner lives, not in the country in which the capital is invested. This makes investors indifferent between domestic and foreign assets when their pre-tax rates of return are the same, so that their pattern of investment among countries is not affected by different direct taxes on capital income. (14) Such differences could affect in which country capital owners live, but this will not affect the allocation or cost of capital among countries.

Differences in direct taxes on earning (e.g. income taxes or employees' payroll taxes) can, however, affect the allocation of both workers and capital among countries. As with indirect taxes, the effects crucially depend on the mobility of workers. The effect on capital investment in a country of a higher tax rate on earnings is exactly the same as the effect of a higher payroll tax (see Ermisch 1991). It discourages capital investment in the higher tax country when workers' mobility responds to pay differentials. It does so by encouraging workers to leave the country, which raises pre-tax wages and the cost of production in the higher tax country relative to that in other countries. The size of this effect depends on the degree to which people respond to differences in after-tax pay between countries. If they are not very responsive, then the discouraging effect on capital investment is small. In the extreme, when they do not respond at all, neither capital investment nor pre-tax wages are affected.

Thus, as with indirect taxes, differences in direct taxes are not likely to have a large effect on capital investment, because of the likely low responsiveness of workers' mobility to differences between countries in after-tax pay. In other words, |social dumping' is unlikely to be important, social policy will not be constrained by the potential mobility of businesses, and the need for a Social Charter is questionable.

Impact of a Social Charter

Suppose that the Social Charter is implemented nevertheless. What would be its effects? Not surprisingly, these depend on the mobility of labour.

The Social Charter would put pressure on poorer countries to raise their social benefits, and assuming that other public expenditure is not reduced to compensate, this would lead to an increase in taxes to pay for them. The analysis of the incidence of taxes has implications for who will bear the burden of introducing the Social Charter. Whether the industries in a country are, on average, capital-intensive or labour-intensive is important in judging the incidence of a particular tax. The poorer countries of the EC also tend to have more labour-intensive industries.

Even when labour moves in response to pay differentials, higher payroll taxes generally reduce wages. This is because the additional tax encourages substitution of capital for labour, thereby reducing the demand for labour. There is also an output effect, which operates through the responsiveness of product demand to the price of the product. This output effect could work in the opposite direction to the substitution effect if production in the country raising the tax is capital-intensive, but when production in the taxing country is labour-intensive, payroll taxes definitely reduce wages. With labour-intensive production, product taxes also reduce wages, and taxes on capital are more likely to reduce wages, the more labour-intensive is production.

Higher direct taxes on earnings can also reduce pre-tax wages in the tax-raising country when production is labour-intensive and labour is mobile in response to differences in take home pay. But, in these circumstances, pre-tax wages would fall by more in the countries receiving the labour whose migration was encouraged by the higher direct tax rate. Thus, workers in the labour-importing countries would be adversely affected by the higher direct tax rate in a labour-intensive country. This scenario is likely when the demand for the product is very responsive to its relative price and labour is mobile.

Thus, higher social benefits in poorer (labour-intensive) countries are very likely to lead to a reduction in take home pay (or a rise in unemployment) in these countries. If higher social benefits are paid for with higher payroll or direct taxes, then low mobility in response to pay differentials means that the cost of these benefits would primarily be born by the country's workers in the form of lower after-tax pay.

Wage rigidity

The analysis has taken a |long term' perspective. It has assumed that adjustments in relative pay between countries occur eventually. Slow adjustment of a country's relative wage to an increase in a tax or regulatory cost would complicate the adjustment process, particularly when migration responds to unemployment differentials, and increase the cost of adjustment, in terms of higher unemployment, but it would not alter the main conclusions.

If wages remained rigid however, capital investment would be discouraged in the country raising its payroll tax. If mobility was not responsive to differences in unemployment rates between countries, that country's population would bear the entire burden in terms of higher unemployment. If the unemployed were much more mobile than the employed, or people's mobility is much more responsive to unemployment rate differentials than relative wages, then a higher payroll tax (or regulatory cost) would affect wages and/or unemployment rates in other countries through emigration of the unemployed. In these circumstances, social dumping could occur, and this could affect decisions about social provision.

But even if labour mobility responds strongly to unemployment rate differentials, if higher social provision were paid for by higher direct taxes on earnings and worker mobility was not responsive to relative after-tax wages between countries, neither capital investment nor unemployment would be affected. As before, a country's workers would bear the entire burden of higher social provision in terms of lower after-tax pay.

Even a combination of wage rigidity, payroll tax finance and responsiveness of migration to unemployment rate differentials does not imply that a Social Charter is needed. If a country's relative wages are inflexible, say because of pay bargaining arrangements, it would be more efficient to reform these arrangements in the countries concerned than to impose a Social Charter on all of the countries of the EC.

4. Effects of relative pay on migration

It has been argued above that the responsiveness of migration to differences in after-tax pay between countries is crucial in deciding whether social policy will be constrained and whether a Social Charter is needed. What does the evidence suggest?

Size and direction of migration within the EC

The Treaty of Rome, which established the European Economic Community in 1957, secured freedom of movement among the original six members of the EC by 1 January 1970, at the latest. (15) Despite this legal liberalisation, mobility among the original six members has hardly increased since the formation of the Common Market. The total number of workers from one of the original six who were working in the country of another original member increased from about half a million before 1960 to about 830,000 in 1968, after which it remained constant until about 1980. During 1980-84, it decreased to about 650,000 (Straubhaar 1988, Figure 1), which represents only 0.75 per cent of the labour force of the original 6 EC countries in 1984.

The relatively small intra-EC migration cannot be attributed to an absence of real wage differentials within the EC. For instance, depending on the measure of real wages used, the standard deviation of real wages ranged from 12 to 20 per cent of mean real wages in the original EC countries during the 1970s; German real wages were 35-50 per cent higher than French real wages in this period, and Italian real wages were much lower than those in France (Tovias 1983). Thus, the migration response by workers in EC countries to pay differentials has been relatively small.

While small, the fact that two-thirds to three-quarters of these migrant workers from another EC country were Italians does indicate movement from low wage to high wage countries within the EC. Furthermore, during 1973-80, the number of Italians working in another of the original six EC countries working declined, which is consistent with the increase in Italy's real wages relative to other EC countries.

In contrast to the Italian movement throughout the original EC countries, other intra-EC migration was regional. Workers from Belgium, the Netherlands and Luxembourg primarily moved within the Benelux countries, and French (German) migrants mainly went to regions in Germany (France), Belgium and the Netherlands near the national border (Straubhaar 1988). Moreover, the Italian northward migration was occurring before the formation of the EC. These developments suggest that the formation of the EC did little to change the patterns of European labour migration.

Examination of the gross inflows to and outflows from the United Kingdom suggests that the UK's membership of the EC also had little immediate effect on migration between it and the other countries of the EC. The UK, as well as Denmark and Ireland, joined the EC in January 1973. Even though movement between the UK and the rest of the EC should have been freer after the UK joined, Chart 1 shows no change in migratory flows between the UK and the rest of the EC in the years following 1973. (16) During the 1970s there was a fairly steady outflow from the UK to the rest of the EC.

After 1981, that changed. There was a sharp increase in inflows to the UK from the rest of the EC in 1982, and during 1984-86 flows in both directions increased dramatically, with net flows favouring the UK. In these years, the EC's percentages of all out-migration from or in-migration to the UK also increased sharply (see Chart 2). While movement in both directions declined somewhat during 1986-89, it was much higher than during the period before 1982.

Greece joined the EC in January 1981 and Spain and Portugal in January 1986. There was a seven-year transitional period for Greece, after which Greeks have been allowed to work in any country of their choice in the EC. New emigrants from Spain and Portugal seeking wage-earning employment in other EC countries require prior authorisation up to the end of 1992.

These restrictions could be damping large potential movement from the Spain and Portugal, and they may have restricted Greek emigration up to 1988. Real hourly earnings during 1986-89 were only about of 30 per cent of UK levels in Portugal and 45 per cent in Greece (Ray 1990). These differentials provide large incentives for northern migration from the new southern members of the EC. The situation of Italy at the formation of the EC was somewhat similar, and so its experience after the formation of the EC may be relevant to these new members. While migration from Italy to the rest of the EC did increase during the 1960s, it was noted above that the scale of migration remained small.

More recent information available on EC countries' migration is primarily aggregate data, particularly net migration to or from a country. The inferences about the sensitivity of migration to relative economic opportunities that can be drawn from such data are very limited. As the market conditions for workers of different skills can vary, these conditions could play an important role in affecting the movement of people even though net migration is near zero. Nevertheless, if movement is sensitive to real wage differences between countries and large differences exist, then a bias favouring migration to high wage countries and from low wage countries of the EC is likely to be present in the aggregate data.

Table 3 groups countries of the EC by the level of their total hourly labour costs in manufacturing during 1986-89, as estimated from surveys by the Swedish Employers' Confederation (Ray, 1990). It shows the maximum and minimum annual net migration rates (as a per cent of the average population that year) during the 1980s and the cumulative migration rate during 1980-88 (as a per cent of beginning 1989 population). As expected, the high wage countries (other than Belgium) were net importers of labour during the 1980s, but the low wage countries were as well. Greece and Portugal, the countries with the lowest wages, have been net importers of labour since the mid-1970s, following a long period (at least back to 1960) in which they were exporting labour.

Thus, at least within the EC, net migration does not appear to have particularly favoured the high wage countries in recent years. Furthermore, with the exception of Irish emigration, the cumulative net flows over the 9-year period are not very large, adding at most 2 per cent to the population (Portugal). The 4 per cent reduction in the population of the Republic of Ireland because of emigrations is, however, fairly large, and Irish emigration is discussed further below.

Taking the entire period of UK membership of the EC (1973-89), cumulative migration between the UK and the rest of the EC was virtually in balance (+ 600 people). (17) It is clear from Chart 1 that this long term net balance has come about because, in the period since the low wage countries of Greece, Spain and Portugal became members, movement between the UK and the rest of the EC increased and the UK

PHOTO : Chart 1. Migration between UK and EC

PHOTO : Chart 2. Migration between UK and EC became be a net importer of labour from other EC countries (a net balance of 56,500 during 1981-89).

In that real wages in the UK are toward the middle of the range among the present membership of the EC, this pattern could be consistent with intra-EC movement toward higher wage countries. (18) Before these low wage countries became members, the UK was one of the low wage countries of the EC (along with Ireland), and that may explain the net movement from the UK to the rest of the EC during the 1970s. Nevertheless, the small scale of movement, even in recent years, suggests either low responsiveness of people to real wage differences among EC countries, or, at best, very sluggish responses.

Reinforcing this conclusion is the fact that a relatively large proportions of these migrants are returning to their original country. For instance, during 1979-88, 45 per cent of the migrants coming into the UK were British citizens, and 35 per cent of people leaving the UK were not British citizens. Focussing on migration between the UK and other countries of the EC, 50 per cent of people coming into the UK from the rest of the EC in 1988 were British citizens, and 28 per cent of people leaving the UK for other countries of the EC were citizens of EC countries other than the UK.

The failure of the pattern or volume of migration among EC countries to change much after the formation of the EC may primarily reflect a tendency for trade in commodities to substitute for movements in factors of production, particularly labour migration. The formation of the EC increased trade between EC countries substantially, rising from 30 per cent to 52 per cent of all EC-trade between 1958 and 1973. According to the theoretical arguments of Mundell (1957), freer trade would reduce migration flows. Indeed, Straubhaar (1988) finds a significant negative correlation between intra-EC trade flows and intra-EC migration flows during 1958-80. (19) Markusen (1983) demonstrated a number of conditions under which freer trade would increase labour migration, but the similar production technologies and preferences among the original six EC countries (with the possible exception of Italy) were probably conducive to trade and labour migration being substitutes. Production technologies of the new southern members of the EC may be sufficiently different from the northern countries to create conditions under which freer trade could increase south-north migration within the EC (see Markusen, 1983).

Migration into the EC

The big change in European migration was the increase in migration to the EC from non-EC countries, particularly from countries that were to join the EC later, namely Greece, Spain and Portugal, but also from Turkey and North Africa. The Portuguese and Northern Africans mainly went to France, the Greeks and Turks primarily to Germany, and the Spanish to France, and to a lesser extent Germany. As consequence, only about one-fifth of migrants working in the original EC countries in 1974 were from another of the original six members.

Much of this migration into the EC was |organised migration' through recruitment bureau of the host countries in the countries of origin. Thus, the size and direction of this movement was primarily determined by policy in the destination countries rather than by spontaneous movement in response to economic opportunities. When European (and world) economic growth slowed after 1973, this policy went into reverse, encouraging large reductions in the number of foreign workers in the original EC countries.

But not all foreign nationalities returned home to the same extent. Between 1973 and 1980, there were declines in the number of Spaniards (200,000), Portuguese (100,000) and Greeks (150,000) working in EC countries, but not in the number of Turkish and North African workers (Molle and Van Mourik 1988). That is, workers from the poorest countries were more likely to stay on in the EC countries. In 1984, about 5.5 per cent of the total active labour force living in the original EC member countries were migrants from other countries. Only one-fifth came from another of the original six members, a further fifth came from one of the new members (Greece, Spain or Portugal), another fifth came from Turkey, and the remaining two-fifths came from other countries, primarily Yugoslavia, North Africa and non-EC European countries (Straubhaar, 1988, p.51).

The organised nature of this migration does not preclude the destination pattern of migration from countries outside the EC being responsive to relative pay. Indeed, econometric analysis indicates that the larger the difference between average wages in a northern European destination country and a southern European or north African origin country, the more foreign workers from that southern country present in that northern European country in 1980 (Molle and Van Mourik 1988). (20) This analysis suggests that a 1 per cent increase in an EC country's real wage differential would increase its labour supply by 0.3 per cent, purely because of the increase in migration from non-EC countries (an elasticity of 0.3). But this estimates of responsiveness clearly reflects the behaviour of EC governments in the 1960s and 1970s, and their difficult experience in assimilating foreign |guest workers' is very likely to reduce the future responsiveness of migration into the EC to real wages there. The reduction in foreign workers in EC countries after 1973 was probably not only a cyclical response, but also an attempt by EC governments to reduce reliance on foreign workers.

Variation in real wages among countries

It may appear that the sensitivity of migration to wage differentials could be gauged indirectly by examining the variation in real wages among countries of the EC and its change over time. Perfect labour mobility would tend to equalise real wages across countries, as people moved in response to real wage differences. But, as noted earlier, free trade can serve as a substitute for mobility of labour and capital. Under certain conditions, it also tends to equalise wages across countries in a customs union like the EC (Samuelson, 1948, 1949). Recent research (Mokhtari and Rassekh, 1989) suggests that freer trade among 16 OECD countries has made the most significant contribution to reducing the variation in real wages among these countries. Thus, less variation in real wages among EC countries over time need not imply that labour is mobile in response to real wage differentials.

Failure to exhibit a tendency toward convergence of real wages would, however, suggest low responsiveness of labour mobility to real wage differences between countries. While Tovias (1982) finds evidence of convergence in labour costs among five original EC members between the mid-1950s and mid-1960s, this process was reversed during the 1970s. (21) In the latter period, labour costs diverged, thereby suggesting poor responsiveness of migration to wage differentials, as well as factors (eg. increasing returns to scale) working against the tendency for freer trade to produce convergence of labour costs.

Direct estimates of the responsiveness of migration

So far we have tried to infer the degree of responsiveness of migration to real wage differentials by examining the size and pattern of European migration and whether there has been convergence in real wages. This section uses studies of migration between particular countries and of internal migration within countries to estimate the responsiveness, or |elasticity', of a country's labour supply to real wage differentials.

Migration between European countries Lianos (1972) studies Greek migration to West Germany, Australia, Canada and the United States during 1959-66. Focussing on Greek emigration to Germany, his analysis (his equation 9') suggest that a once and for all increase of 10 per cent in the German-Greek real wage differential would eventually decrease the Greek labour force by 0.67 per cent, or an elasticity of 0.067. (22) This response in the labour force is only that arising from a increase in gross migration from Greece to Germany. It ignores changes in return migration from Germany to Greece, which would probably decline. Also, if it was a fall in the wage in Greece that increased the German-Greek wage differential, that this response also ignores changes in migration between Greece and other countries as a consequence of changes in wage differentials. Thus, this elasticity measure may understate the total responsiveness of the Greek labour force to pay differentials between Greece and other countries. (23)

Analysis of the response of Irish migration to real wages in the Republic of Ireland relative to those in Great Britain may provide a better measure of the total labour force response. During the 1950s and 1960s, approximately four-fifths of Irish emigration was to Britain. Thus, the British-Irish wage differential is likely to be by far the most important pay differential affecting total Irish net migration. Walsh's (1974) econometric study of net emigration from Ireland during 1951 - 71 suggests that a one-off 10 per cent increase in Irish real wages relative to those in Britain would ultimately increase the Irish labour force by about 1.5 to 3 per cent. (24) In other words, the elasticity of the Irish labour force with respect to Irish real wages relative to Britain's appears to have been between 0.15 and 0.3 during this period. (25)

Walsh (1974) also finds that the unemployment rate in Ireland relative to that in the UK had a strong effect on Irish migration, and this is confirmed by extension of the analysis of Irish migration beyond 1971. The turnaround in Irish migration during the 1970s, when Ireland became a net importer of people, stimulated a re-examination of the analysis of the sensitivity of Irish migration to economic conditions. Keenan (1981) reports on a better series for Irish net migration, and have used the data that he provides to estimate models of Irish net migration which include the experience of the 1970s. (26)

These estimates indicate that the unemployment rate in Ireland relative to that in the UK is more important than relative pay in explaining the large changes in Irish net migration. (27) Indeed, the high unemployment rate in Ireland relative to the UK appears to have been primarily responsible for large emigration from Ireland during 1951-71, when net emigration totalled 543,000, amounting to half the Irish labour force in 1971 (Walsh 1974). Chart 3 shows the Irish net migration (as a proportion of the labour force) predicted by my estimated model if the unemployment rate in Ireland had remained constant at twice the UK rate, and actual net migration. The difference between the two series during 1955-71 shows the contribution of high Irish unemployment to net emigration. (28) It appears that 85 per cent of Irish net emigration during this period was attributable to the high Irish unemployment rate relative to the UK rate.

The estimated model implies elasticities of the Irish labour force with respect to Irish real wages relative to British of about 0.1, which is only slightly lower than those implied by Walsh's (1974) analysis of the earlier period, using a different measure of net migration. Thus, while there is evidence that migration between countries that are linguistically and culturally similar can be very responsive to relative unemployment conditions, the responsiveness to relative pay is fairly small. Between countries less similar in these respects, the responsiveness of the labour force to relative pay is likely to be even smaller.

Internal migration in European countries It is very likely that estimates of the sensitivity of migration within a country to differential economic opportunities would overstate the sensitivity of migration between EC countries, because linguistic and cultural homogeneity within a country favours mobility. American studies have often suggested that interstate migration is responsive to earnings differences between states, (29) but, as noted earlier, American people tend to be much more mobile than Europeans. A more relevant upper bound on the responsiveness of migration and labour supply to differences in pay can be obtained from studies of inter-regional migration within European countries.

The study of net migration from southern to northern Italy during 1958-74 by Salvatore (1977) provides information from which it is possible to estimate the responsiveness of the southern Italian labour force to north-south pay and unemployment differentials. During this period, average per capita income in the South was about 60 per cent of that of the North of Italy, and 1 million workers (about 17 per cent of the Southern labour force) emigrated from the poorer South of Italy to the North. (30) Again, unemployment differential appear to be more important in accounting for changes in migration than pay differences. Salvatore's (1977) analysis suggests that a 10 per cent increase in the North's real wage relative to the South's would reduce the South's labour force by 0.7 per cent, or an elasticity of 0.07. (31) This is only slightly smaller than the elasticity suggested by the experience of Irish migration, and it is very near the estimate from a recent study of internal migration in Great Britain (Pissarides and McMaster 1990).

This study of migration among Scotland, Wales and the 7 planning regions of England (the South East and East Anglia are grouped together) during 1961-82 allows estimation of the responsiveness of a region's labour supply to relative pay and unemployment rates. Using their estimated migration equation, in the long run, a 10 per cent higher regional wage (relative to other regions) increases regional labour supply to 0.6 per cent (an |elasticity' of 0.06). Their estimate of the labour force response to regional unemployment differentials also indicates a very sluggish response. For instance, it takes 10 years for migration to eliminate half of a regional unemployment rate differential.

Other studies of migration in Great Britain have produced mixed results concerning the effect of relative wages. Hughes and McCormick (1990) find, using data on individuals, that migrants move from regions of relatively low levels of real wages to regions with higher real wages. They find that regional wage differentials particularly affect the destination of people who have decided to migrate, but regional real wage levels do not appear to affect significantly the decision to leave a region. Pissarides and Wadsworth (1987) find that a region's relative wage has a significant effect on the migration decision, and a number of studies of net migration between Scotland and the rest of the UK also suggest that regional earnings are an important influence on migration. (32) Muellbauer and Murphy (1988) also provide evidence that higher relative earnings encourage migration to the South East. On the other hand, Mohlo's (1984) analysis of inter-regional migration flows suggests that relative earnings do not affect migration patterns significantly.

The aggregate migration equations in some of these studies have the level of a region's relative wage as an explanatory variable, rather than the change in the relative wage used by Pissarides and McMaster (1990). Presumably they are meant to suggest how labour supply adjust to wage differentials resulting from disequilibrium, but such an estimated net migration equation is not measuring the response of labour supply to relative pay, but rather the equilibriating compensating wage differentials, under the maintained assumption of perfect mobility. (33) Clearly, we do not need any econometric estimates of elasticities if we adopt this maintained assumption.

If regional wage differentials primarily compensate for other differences between regions (for example, lower pay in regions with nicer climates), then wage differentials would not affect migration. For instance, Pissarides and McMaster's (1990) analysis suggests that, in the long run, higher than average regional wages compensate for higher than average unemployment. A combination of disequilibrium and compensating regional wage differentials may explain the mixed results in estimating the effect of the level of a region's relative wage on migration. These other British migration studies provide insufficient information to calculate the elasticity of regional labour supply with respect to a region's relative wage.

As the elasticity estimates from the internal migration studies are likely to be an upper bound on the responsiveness of a country's labour supply to its relative wage, it appears that workers' responsiveness to inter-country pay differences is likely to be small. Estimates from Italy and Great Britain suggest elasticities of less than 0.1, and the studies of Irish and Greek migration discussed above also suggest elasticities in this range. The Irish case is also likely to overstate the responsiveness of labour supply to relative pay between other European countries, because of the linguistic and cultural similarity of Ireland and Great Britain, to which a large proportion of migrants went, and the long tradition of emigration from Ireland.

5. Conclusions

While clearly not conclusive, the limited evidence available suggest that the mobility of workers between member countries of the EC is not likely to be very sensitive to differences between countries in social benefits or after-tax pay. In these circumstances, if relative wages between countries are allowed to adjust, higher payroll or direct taxes in a country only have a small discouraging effect on capital investment, and the discouraging effects of taxes on capital and on production are also much smaller. In some countries, pay bargaining arrangements may impede relative wage adjustment. But, if this is the case, it would be more efficient to reform these arrangements than to impose a social charter on EC members.

When, as is preferable, social benefits are paid for through payroll taxes and/or direct taxes, the small responsiveness of labour mobility to pay differentials also implies that workers in each country bear virtually the entire burden of the taxes. In that they pay for and receive the benefits exclusively, decisions concerning the level of social benefits are best left to the population residing in each country. The option of voting and paying for higher social benefits is already open to those countries with poorer social protection, and the decision is not constrained by the forces of economic competition. Imposition of higher social benefits on such countries, through the Social Charter, would, therefore, reduce the welfare of the majority of the population in these countries. In other words, member countries of the EC should be given autonomy in setting their social policies, including those involving redistribution. A Social Charter does not appear to be needed.

The argument by the UK government that the Social Charter would raise unemployment is consistent with the analysis here if a country's wages relative to other countries are not allowed to fall when it raises taxes to pay for better social benefits. When relative wages can adjust, higher social benefits are paid for by low after-tax pay in the country rather than higher unemployment.

The conclusion that member countries should be given autonomy in social policy follows from the lack of responsiveness of workers' mobility to differences in social benefits or take home pay. The Social Charter also attempts to encourage labour mobility between member countries. While all of the attempts at harmonisation in the Charter work in this direction, it particularly provides that |The right of freedom of movement shall enable any worker to engage in any occupation or profession in the Community in accordance with the principles of equal treatment as regard access to employment, working conditions and social protection in the host country.' (para.2). It goes on to call for |Elimination of obstacles arising from non-recognition of diplomas or equivalent occupational qualifications;' (para.3).

It is unclear how effective these provision would be in increasing worker responsiveness to differences in pay and benefits. If they are effective, then we may need to reconsider the need for the other aspects of the Social Charter. But the experience of migration after the formation of the EC and the addition of new members in 1973 suggests that large changes in the responsiveness of migration to differential economic opportunities in EC countries is unlikely.

REFERENCES

Adams, D.W. (1980a), |Migration from Scotland to the rest of the United Kingdom', Discussion Paper

10, Economics and Statistics Unit, Scottish Office, Edinburgh. Adams, D.W. (1980b), |Migration from Scotland to the rest of the U.K. Revisited', Discussion Paper

Economics and Statistics Unit, Scottish Office, Edinburgh. Blank, R.M. (1988), |The effect of welfare and wage levels on the location decisions of

female-headed households', Journal of Urban Economics, 24:186-211. Ermisch, J.F. (1991), |Labour migration and the incidence of taxes', paper presented to the Fourth

Annual Conference of the European Society for Population Economics, Pisa, June. Flatters, F., Henderson, V. and Mieszkowski, P. (1974), |Public goods, efficiency, and regional

fiscal equalization', Journal of Public Economics, 3:99-112. Gramlich, E.M. and Laren, D.S. (1984), |Migration and income redistribution responsibilities',

Journal of Human Resources, 19:487-511. Graves, P.E. and Linneman, P.D. (1979), |Household migration: theoretical and empirical results',

Journal of Urban Economics, 6:383-404. Greenwood, M.J. (1985), |Human migration: theory, models, and empirical studies', Journal of

Regional Science, 25:521-544. Harberger, A.C. (1962), |The incidence of the corporate income tax', Journal of Political Economy,

70:215-240. Hughes, G.A. and McCormick, B. (1990), |Housing and labour mobility in Housing and the National

Economy (J.F. Ermisch, ed), Avebury-Gower, Aldershot (UK). Keenan, J.G. (1981), |Irish migration, all or nothing resolved?', The Economic and Social Review,

12:169-186. Lianos, T.P. (1972), |The migration process and time lags', Journal of Regional Science, 12:425-433. Markusen, J.R. (1983), |Factor movements and commodity trade as complements', Journal of

International Economics, 14:341-356. McLure, C.E. (1970), |Taxation, substitution, and industrial location', Journal of Political

Economy, 78:112-132. Mieszkowski, P.M. (1967), |On the theory of tax incidence', Journal of Political Economy, 75:250-262 Mohlo, I. (1984), |A dynamic model of interregional migration flows in Great Britain', Journal of

Regional Science, 24:317-337. Mokhtari, M. and Rassekh, F. (1989), |The tendency towards factor price equalization among OECD

countries', Review of Economics and Statistics, 71:636-642. Molle W. and van Mourik, A. (1988), |International movements of labour under conditions of economic

integration: the case of Western Europe', Journal of Common Market Studies, 26:317-342. Muellbauer, J. and Murphy, A. (1988), |Housing and regional migration to and from the South East',

Nuffield College, Oxford, mimeo. Mundell, R. (1957) |International trade and factor mobility', American Economic Review, 47:321-335. Myers, G.M. (1990), |Optimality, free mobility, and the regional authority in a federation', Journal

of Public Economics, 43:107-121. Olson, M. (1969), |The principle of fiscal equivalence: the division of responsibilities among

different levels of government', American Economic Review Proceedings, 59:479-487. Pissarides, C.A. and McMaster, I. (1990), |Regional migration, wages and unemployment: empirical

evidence and implications for policy', Oxford Economic Papers, 42:812-831. Pissarides, C.A. and Wadsworth, J. (1987), |Unemployment and the inter-regional mobility of labour',

Center for Labour Economics Discussion Paper 296, London School of Economics. Ray, G.F. (1990), |International labour costs in manufacturing, 1960-88', National Institute

Economic Review, Number 132 (May). Rothenberg, J. (1970), |Local decentralization and the theory of optimal government', in The

Analysis of Public Output (J. Margolis, ed), National Bureau of Economic Research and Columbia

University Press, New York. Salvatore, D. (1977), |An econometric analysis of internal migration in Italy', Journal of Regional

Science, 17:395-408. Samuelson, P.A. (1948), |International trade and the equalisation of factor prices', The Economic

Journal, 58:163-184. Samuelson, P.A. (1949), |International factor-price equalisation once again', The Economic Journal,

59:181-197. Schachter, J. and Althaus, P.G. (1989), Journal of Regional Science, 29:143-159. Straubhaar, T. (1988), |International labour migration within a Common Market: some aspects of EC

experience', Journal of Common Market Studies, 27:45-62. Sinn, H-W. (1990), |Tax harmonization and tax competition in Europe', National Bureau of Economic

Research Working Paper No. 3248. Tovias, A. (1982), |Testing factor price equalization in the EEC', Journal of Common Market Studies,

20:375-388. Walsh, B.M. (1974), |Expectations, information, and human migration: specifying an econometric model

of Irish migration to Britain', Journal of Regional Science, 14:107-120.

NOTES

( 1) See McGuire (1974) for a rigorous demonstration of this principle. ( 2) It does not, however, imply that separate bureaucracies or elected officials are needed for

every jurisdiction. A

collective good could be voted on and paid for by the appropriate jurisdiction, by the

administration of the provisions

could be handled by a larger jurisdiction. ( 3) The principle of fiscal equivalence implies that the jurisdictional boundaries should be drawn

to include all beneficiaries

of the collective good. When people are mobile, the number of people within those boundaries

would be affected by the

cost per head of the collective good, and when there are decreasing costs per head, movement of

people would affect

the cost per head. This fiscal effect of the movement of people across jurisdictional

boundaries could lead to inefficient

decisions about the level of provision of collective goods by the local jurisdictions, but this

need not be the case (see

Flatters, Henderson and Mieszkowski, 1974, and Myers, 1990). ( 4) Gramlich and Laren's (1984) analysis of variation in the average levels of welfare (AFDC)

benefits paid by different

states in the US suggest that, ceteris paribus, there is a weak positive relationship between

average income and

redistribution, supporting weak dominance of the altruism motive. ( 5) Furthermore, redistribution involving distant populations are not likely to be a source of

welfare: |charity begins at

home'. ( 6) She groups the states into twelve regional groupings, each of which has similar women's wages,

unemployment rates

and AFDC benefit levels as well as being close together geographically; see her Appendix III.

In her study, migration

rates are defined over a four-year period (1975-79), which is comparable to the five-year

period (1975-80) used by

Gramlich and Laren (1984). ( 7) If M is the matrix of inter-state migration rates, then the equilibrium proportions are the

elements of the vector [alpha], where [alpha]

is defined so that [alpha]M = [alpha]. ( 8) The assumptions used to make this calculation are likely to be favourable to a large impact. An

alternative set of

assumptions, which changes only the outflow rates, produces an even smaller changes in the

equilibrium distribution of

the AFDC population. ( 9) The conclusions that follow concerning tax effects are based on further development of McLure's

(1970) two-country,

two factor model, which builds on the analysis of Mieszkowski (1967) and Harberger (1962). A

companion paper,

Ermisch (1991), presents the model and demonstrates the main propositions. (10) In the formal analysis (Ermisch 1991), more consumer responsiveness in a single, integrated

market is approximated by

letting the elasticity of product demand with respect to the relative price of the two

countries products be larger than the

elasticity of substitution between capital and labour in production. A special case of this

would be to allow the demand

elasticity to approach infinity. (11) As McLure (1970) shows, this is true under the plausible condition, assumed earlier, that the

price elasticity of demand

for the country's products is larger than the elasticity of substitution between capital and

labour in production. (12) The analysis has abstracted from other sources of labour supply elasticity: changes in hours

and employment

participation by the home population in response to real wages. Men's hour'' elasticity tends

to be small and negative,

while women's hours' elasticity is small positive. Women's participation elasticity has been

larger and positive, but it

has fallen as women's labour force participation has approached men's. It is not clear whether

these responses would

be positive or negative on balance, but the net response is likely to be small relative to

potential changes from migration. (13) Furthermore, it is the EC's intention to levy VAT on the |destination principle' rather than

the |origin principle' (a

production tax); that is, goods that leave the country are exempt from the exporting country's

VAT, and are subject to the

importing country's VAT. Sinn (1990) argues that this may, however, be difficult after border

controls are removed, thus

making VAT more like a production tax. (14) The residence principle is, in theory, applied by most OECD countries, including members of the

EC. (15) For employed persons, the Treaty stipulated that |the freedom of movement shall entail the

abolition of any discrimination

based on nationality between workers of the member states as regards employment, remuneration

and other

conditions of work and employment.' (article 48ff.) (16) In Charts 1 and 2, the EC is constituted as it actually was in each year; that is, the

countries included in the EC change

over time. Migration between the UK and the Republic of Ireland is excluded from the UK

statistics on international

migration, used in Charts 1 and 2. (17) Over the period 1964-89, net migration between the UK and the rest of the rest of the EC

was +9, 100 people. (18) The published information does not give details on the individual countries of the EC who are

the origins and

destination of the migration. (19) This negative correlation is maintained when intra-EC trade is expressed as a proportion of

total EC trade, and intra-EC

migration is expressed as a proportion of total EC migration flows; the correlation coefficient

is -0.55 (Straubhaar,

1988, Table 1). (20) The analysis controls for distance between the two countries and for trade between them. The

destination countries

considered in the study are France, Germany, Netherlands, Belgium, Sweden, Austria and

Switzerland, and the origin

countries are Italy, Greece, Spain, Portugal, Yugoslavia, Turkey, Morocco, Algeria and Tunisia. (21) Luxembourg is excluded. Similar conclusions arise from examining the wage differential between

France and the

Federal Republic of Germany. (22) Estimation of the labour supply elasticity from net migration equation estimates is based on

the assumption that the

change in a country's labour force is equal to net labour migration into/from the country.

Thus, net migration should

depend on the change in relative wages rather than the level of the relative wage. In

equilibrium, both net labour

migration and relative wage changes should be zero. The net migration equation is, therefore, a

disequilibrium, labour

supply adjustment equation. Letting [M.sub.x] = net labour migration and dlog [W.sub.x] = the

proportionate change in country X's

relative wages and expressing net migration to the country's labour force ([L.sub.x]), a

natural expression for net

migration is [M.sub.x]/[L.sub.x] = d. Log [W.sub.x,] where x is the elasticity of labour supply

with respect to relative wages. Because of

response lags, a distributed lag model is likely to be more appropriate; thus, as in Lianos'

analysis, net migration in

period t is given by

[Mathematical Expressions Omitted]

where [xi] is a random term, and the constant term [C.sub.x] captures autonomous migration

(independent of economic

opportunities) to or from the region or country. (23) It should be noted, however, that Greek migration to Germany was by far the largest migration

flow during the period

under study. Furthermore, Lianos does not allow for the unemployment rate in Greece relative to

Germany to affect

migration. Analysis of Irish migration, described below, suggests that omission of the relative

unemployment rate from

the migration equation biases upwards the estimate of the elasticity of the labour supply with

respect to relative pay. (24) This range of estimates is based on the coefficients in Walsh's |extrapolative and |adaptive

expectations' models

reported in Table 3 of Walsh (1974), which he appears to prefer. The estimates assume that 75

per cent of migrants and

50 per cent of the total population are in the labour force. This is consistent with the

evidence that, in the mid-1960s, 86

per cent of male and 71 per cent of female migrants from Ireland to the UK were labour force

participants. Walsh's

models generally express migration as a function of the level of the relative wage. As

explained below (footnote 33), this

type of model implicity assumes perfect labour mobility and compensating wage differentials. In

order to use his

empirical estimates, I have assumed that the long run response of net migration to a temporary

(one year) change in the

relative wage implied by his equations is the total labour supply response to a permanent

relative wage change. (25) Estimates of Walsh's (1974) |information flow' model produce much larger elasticities, but my

own analysis of the Irish

migration data (described below) strongly suggests that these arise because he constrains the

constant term to be zero,

thereby inflating the coefficient on the lagged dependent variable. (26) I estimate models analagous to that of Pissarides and McMaster (1990). A country/region's

relative unemployment rate,

[U.sub.xt]/[U.sub.t], is taken as an indicator of existing disequilibrium in the labour market,

to which migration is also responding:

[Mathematical Expressions Omitted]

The estimate of the long run labour supply elasticity is a/(1 - [gamma]. In other words, the

long run labour supply response to the

relative wage is estimated as the sum of annual migration responses to a one-off permanent

increase in the region's

relative wage, holding relative unemployment in the region constant. This is the appropriate

calculation for determining

the structural, long run effect of relative wages on labour supply. (27) My analysis of the Irish migration data provided by Keenan (1981; Table A. 1) finds an

acceptable econometric model

virtually identical to that of Pissarides and McMaster's (1990). Using the new migration

series, which appears to be

preferable to the official series (and real non-agricultural earnings per hour for the Irish

wage series), the estimated

equation for 1955-77 is:

[Mathematical Expressions Omitted]

The estimate of [gamma] (of the previous note) was not near being statistically different from

zero; thus the lagged dependent

variable was dropped the equation. (28) The model implies that an Irish unemployment rate of about twice the UK rate would result in

zero net migration from

Ireland; this can be interpreted as an equilibrium relative unemployment rate. The predicted

series in Chart 3 allows for

actual changes in Irish real wages relative to British. (29) Indeed, in recent years, American researchers have favoured a so-called equilibrium model of

migration' which

assumes perfect mobility of people and that all regional wage differentials are |compensating

differentials', compensating

for different amenities etc. Such differentials would not directly encourage migration. See,

for example, Graves and

Linneman (1979) and Schachter and Althaus (1989). This contracts with the traditional

assumption that migration

equations represent disequilibrium adjustment functions (see Greenwood, 1985). (30) The South, or Mezzogiorno, consists of the regions of Abruzzi-Molise, Campania, Puglia,

Calabria, Sicily and Sardinia.

The labour supply elasticity estimate is based on the migration equations for the entire South

in Salvatore's (1977) Table

2. (31) This estimates is based on the same assumption as that used to derive estimates from Walsh's

models, as noted in

footnote 24 above. (32) Adams (1980a, 1980b) and Harrigan, Jenkins and McGregor (1986). (33) When the net migration equation takes the following form:

[Mathematical Expressions Omitted]

there are |compensating wage differentials' between countries or regions. To see this, note

that in equilibrium [M.sub.x] = 0, so

that [C.sub.x] = - b.log[W.sub.xt] + [delta]([U.sub.xt]/[U.sub.t]). Thus, workers in regions/

countries favoured by migrants (high [C.sub.x], say because of a

favourable climate) receive lower pay and/or experience higher equilibrium unemployment than

average. It also

implicitly assumes that the regional labour supply is perfectly elastic with respect to

relative pay. In order to see this

clearly, let [delta] = 0. Then in equilibrium - [C.sub.x/b] = log[W.sub.xt], and as long as

log[W.sub.xt] exceeds - [C.sub.x]/b, there will be net inward

migration.