Letters to the editor
Polls and Comment
Library - an archive
Bibliography of EMU
The Euroland Economy
'One money. one market: estimating the effect of common currencies on trade' - Andrew K. Rose, CEPR discussion paper no 2329, December 1999.
The latest attempt to measure the effect of a common currency on trade is by Andrew Rose who has heroically looked across all the countries of the world in the period 1970-90 to see how far their trade has been influenced by different factors. (It should be made clear from the outset that the existence of a trade effect is not necessarily good for welfare; it could represent 'trade diversion' whereby the 'monetary union preference' switches trade from more efficient sources to protected union sources. However it is usually assumed that it would be 'trade creation', whereby trade is increased at the expense of less efficient but previously protected domestic sources - which is of course welfare-enhancing.) The basic model Rose uses is the gravity model which attributes the extent of bilateral trade to distance, size (GDP) and the level of income per head. On top of this are grafted other factors, of the usual sort, and in particular membership of a single currency. Rose has tried to allow for everything one could think of in this context - colonial status, common language, and much else. Then he enters the 'single currency' factor as well, and finds that it accounts for a tripling of the amount of bilateral trade. He performs a huge number of sensitivity exercises and still this tripling comes through. On the face of it is a remarkable endorsement of the importance of a single currency - since contrary to the previous research based on the experience of fixed exchange rates that trade is not increased by such fixity, this shows that the fixity from a monetary union does indeed dramatically increase trade. The Rose study confirms that fixed exchange rates alone does not much increase trade - the implication is that it requires the irreversibility of a union to do it.
This study seems extremely impressive in its thoroughness and professionalism. The doubts enter when one looks carefully at the sample of countries that are monetary unions. The list is appended. It can be seen that they are virtually all unions between developing countries and a mother country. This opens the study up to the serious underlying difficulty of such cross-section studies (i.e. studies that look across people or groups or countries at a point of time) - namely 'selection bias'. By this is meant that the countries in the monetary union (in this case) are not randomly selected- ideally one would wish to find a random selection of countries that were assigned to monetary union and another random selection that were not, for example as in a controlled experiment in the use of a medicine. In such an experiment a random sample of people is chosen and some are given the medicine, others a placebo; then one can check the differences using standard sample theory. The problem with the Rose sample is that the countries with monetary union were probably not randomly chosen; rather they were chosen because they had a mother country which imposed on them a mother currency. Those that had no monetary union were chosen because they did not have such a mother country. Hence those that had a monetary union and a mother country were naturally predisposed to have a lot of trade with the mother country. Those that did not have a mother country and no monetary union were predisposed not to have much special trade with other countries.
Another way of expressing this is to say that there is a cause of the relationship between monetary union and trade which is a common cause of both - in this case the existence of a mother country. Monetary union does not cause the trade it is the mother-son relationship that causes both higher trade and the monetary union.
To this Rose would no doubt reply that he tested for the effects of 'colonial status' and so on; so he does, with all sorts of controls. The trouble is the classic cross-section elephant trap. A country that decides to go independent of the mother country will also have an independent currency as part of that decision; that decision will also affect the closeness of trading relationships. We would like some random selection process that took some dependent countries and gave them both dependent and independent currencies and some independent countries and gave them both independent and common currencies (which is of course the situation with the EU and the euro). Unfortunately in this sample one can find no such random selection process. What we have is dependent countries with common currencies and independent countries with independent currencies.
To put it in concrete terms, is the fact that CFA countries have close trading relationships with France whereas Algeria does not, a tribute to a common currency or to a political decision about the degree of dependence? Or that Gibraltar has a close trade link with the UK whereas Malawi does not a tribute to its using the pound or just to its dependence? To conclude from the fact that there are significant differences in trade intensity between such groups that they are down to having a single currency is far from convincing. It is possible - the correlation is there - but the causation is simply unproven.
It may be that concealed in this data set is some random selection process for certain country episodes. For example, Ireland's relation to the UK seems to have been affected by random events - according to this study Ireland 'had a common currency with the UK before 1979', though this is technically incorrect (the punt has had a separate identity since independence though it is true that it was de facto joined to the pound by a fixed link up to 1979 - is that obviously-reversible link a 'union' rather than a fixed exchange rate?). However there have been degrees of linkage and these have varied for non-political reasons (the UK leaving the ERM in 1992, for example). Maybe there are other cases here which could be exploited. But as we have already seen, the list of countries that have changed to/from monetary union is short, consisting of only four, and provides no statistical significance in this sample as actually used.
To put it rather crudely, the UK is being invited to expect a tripling via monetary union of its trade with the EU, based on the experience of a few extremely small economies' trade with one another and their mother country. Mr. Rose, when interviewed by the Financial Times (report June 26, p.2, appended), had the honesty to admit that the experience of the small countries in his sample might not be a good guide to what the UK should expect. It is a pity he did not highlight this point more clearly in his paper; without the caveat he risks being a pawn in a ruthless power game.
Article from the Financial Times, 26th June 2000