96 EICHENGREEN and FRANKEL with one another in 1928,which is plausibly attributable to the fact that many of these countries were formerly constituents of Prussia and the Austro-Hungarian Empire,regions with long-standing economic links.There is also evidence that these (largely)Central European countries traded less with the rest of the world than would be predicted by their other characteristics.Countries in the Western Hemisphere,in contrast,showed no ten- dency to trade disproportionately with one another but displayed a significant tendency to trade less than expected with the rest of the world. Particularly interesting from the standpoint of present concerns is how these coefficients change between 1928 and 1935-1938.There is an increase in the tendency for Common- wealth members to trade disportionately with one another,as would be expected on the basis of the system of imperial preference established following Britain's adoption of the General Tariff in 1932.This is consistent with our findings in Table 3.The coefficient on trade between Commonwealth countries and the rest of the world switches its sign,from positive to negative,as if the system of Commonwealth preferences was discriminatory, but the effect is small and weak. The literature on the bilateral agreements used by Germany to secure supplies of raw materials and increase the self-sufficiency of its sphere of influence does not yield clear predictions for the volume of trade (discussed further below).While the coefficient for trade between two members of the bloc increases modestly between 1928 and 1935,it falls by a still larger amount in 1938;there is little evidence here that the trade policies of the members of the German bloc stimulated trade among them.Again,this confirms the results of Table 3.This is not surprising insofar as the goal of these policies was to promote self- sufficiency,not to encourage trade.Indeed,there is evidence that trade with the rest of the world was discouraged.There is statistically significant evidence for 1935 and 1938 that the members of the German bloc traded less with the rest of the world than their other char- acteristics and the average behavior of countries in the sample would predict,unlike 1928, when the coefficient in question differs insignificantly from zero at conventional levels of confidence. The econometric evidence provides little support for the existence of a dollar bloc in the 1930s.While the coefficient on Dollar2,denoting that both trading partners are members of the dollar bloc,is positive in all three years.in no year does its coefficient differ signif- icantly from zero at standard confidence levels.Again,the multivariate results do not differ from the simple tabulations of Table 3 in this respect.Nor is there much evidence of trade diversion:the coefficients on Dollary,while uniformly negative,also differ insignificantly from zero. Thus,in contrast to the evidence for the 1980s,which suggests that regional trade arrangements tended to be trade creating more often than trade diverting,the evidence for the 1930s points to the existence of both benign and malign regionalism.It suggests that the Commonwealth bloc was largely trade creating,whilc Germany's bilateral arrange- ments discouraged trade with countries outside its sphere of influence but without encour- aging trade within it,especially in the years leading up to World War II. WHAT IT MEANS To interpret the implications of these patterns of emerging regionalism,it is necessary to analyze their political and economic origins.Our argument is that the roots of regionalism were very different in the two periods,pointing to different implications.96 EICHENGREEN and FRANKEL with one another in 1928, which is plausibly attributable to the fact that many of these countries were formerly constituents of Prussia and the Austro-Hungarian Empire, regions with long-standing economic links5 There is also evidence that these (largely) Central European countries traded less with the rest of the world than would be predicted by their other characteristics. Countries in the Western Hemisphere, in contrast, showed no ten￾dency to trade disproportionately with one another but displayed a significant tendency to trade less than expected with the rest of the world. Particularly interesting from the standpoint of present concerns is how these coefficients change between 1928 and 1935-1938. There is an increase in the tendency for Common￾wealth members to trade disportionately with one another, as would be expected on the basis of the system of imperial preference established following Britain’s adoption of the General Tariff in 1932. This is consistent with our findings in Table 3. The coefficient on trade between Commonwealth countries and the rest of the world switches its sign, from positive to negative, as if the system of Commonwealth preferences was discriminatory, but the effect is small and weak. The literature on the bilateral agreements used by Germany to secure supplies of raw materials and increase the self-sufficiency of its sphere of influence does not yield clear predictions for the volume of trade (discussed further below). While the coefficient for trade between two members of the bloc increases modestly between 1928 and 1935, it falls by a still larger amount in 1938; there is little evidence here that the trade policies of the members of the German bloc stimulated trade among them. Again, this confirms the results of Table 3. This is not surprising insofar as the goal of these policies was to promote self￾sufficiency, not to encourage trade. Indeed, there is evidence that trade with the rest of the world was discouraged. There is statistically significant evidence for 1935 and 1938 that the members of the German bloc traded less with the rest of the world than their other char￾acteristics and the average behavior of countries in the sample would predict, unlike 1928, when the coefficient in question differs insignificantly from zero at conventional levels of confidence. The econometric evidence provides little support for the existence of a dollar bloc in the 1930s. While the coefficient on Dollurz, denoting that both trading partners are members of the dollar bloc, is positive in all three years, in no year does its coefficient differ signif￾icantly from zero at standard confidence levels. Again, the multivariate results do not differ from the simple tabulations of Table 3 in this respect. Nor is there much evidence of trade diversion: the coefftcients on DollurN, while uniformly negative, also differ insignificantly from zero. Thus, in contrast to the evidence for the 198Os, which suggests that regional trade arrangements tended to be trade creating more often than trade diverting, the evidence for the 1930s points to the existence of both benign and malign regionalism. It suggests that the Commonwealth bloc was largely trade creating, while Germany’s bilateral arrange￾ments discouraged trade with countries outside its sphere of influence but without encour￾aging trade within it, especially in the years leading up to World War II. WHAT IT MEANS To interpret the implications of these patterns of emerging regionalism, it is necessary to analyze their political and economic origins. Our argument is that the roots of regionalism were very different in the two periods, pointing to different implications