The Making of Global Finance, 1880-1913

Published by EH.NET (February 2005)

Marc Flandreau and Frederic Zumer, The Making of Global Finance, 1880-1913. Paris: OECD, 2004. 144 pp. $30 (paperback), ISBN: 92-64-01534-5.

Reviewed for EH.NET by Hugh Rockoff, Department of Economics, Rutgers University.

The authors of this careful and vigorously argued monograph enter the debate over the role of the gold standard in the international economic system at the end of the nineteenth century. One issue is whether or not a country’s adherence to the gold standard mattered to international financial markets. One school of thought holds that it mattered a great deal, that financial markets approved of countries that joined the gold standard. The literature that supports this view includes Michael Bordo and Hugh Rockoff (1996), Nathan Sussman and Yishay Yafeh (2000), and Maurice Obtsfeld and Alan M. Taylor (2003). Flandreau and Zumer, however, are firmly on the other side. As far as they can see, international capital markets were supremely indifferent to whether or not a country adhered to gold.

(There is, I have to confess, a certain satisfaction in knowing that I have at least helped stirred things up. A famous economic historian once told me, however, that it wasn’t enough to have papers and books attacking you — you hadn’t really arrived in economic history until an entire conference was held to refute your ideas!)

The methodology that Flandreau and Zumer use is straightforward and in some ways is simply a further development of what has gone before. They run a regression in which the dependent variable is the difference between the yield for a particular country’s bond and the yield on U.K. bonds. The independent variables include a dummy variable for adherence to the gold standard and other variables that they classify as structural variables, reputational variables, and political variables. The structural variables include debt burden (defined as government interest payments relative to government revenues), exports relative to population, bank reserves relative to banknotes, exports relative to population, government deficits relative to government revenue, and exchange rate volatility. The reputational variables include whether or not a country has defaulted on its bonds and a memory variable to allow for the slow recovery of its reputation. The political variables include the enfranchised share of the population and political crises.

The great strength of the monograph is the care that Flandreau and Zumer have put into measuring their variables and running alternative specifications of the regression to test the robustness of their results. The sample consists of annual data for seventeen countries from 1880 to 1913. The countries were chosen mainly, it seems, because Flandreau and Zumer found data for them in the archives of the Credit Lyonnais that they could use to double check their variables and to be sure that they were looking at things in the way that a major participant in the market looked at things. For this reason some countries that you might want to include, such as the United States, Canada, and Australia, are omitted. Nevertheless, there is something to be said for having authors work with data they are sure about. Their data is reported in appendices, and available at http://eh.net/databases/finance/.

The main finding is that the coefficient on the gold standard dummy turns out to be small and insignificant in most of the regressions and the coefficient on the debt burden variable turns out to be large and significant. Flandreau and Zumer conclude that adherence to the gold standard did not matter, but that the debt burden did. The dragon has been slayed. One can’t help but be impressed with the effort and care that has gone into the monograph. Still, I have some reservations about the test that I have grouped under five headings. Some of these reservations might be regarded more as ideas for future research than as direct criticisms of what Flandreau and Zumer have done.

1. Left-hand and right-hand variables The variable that Flandreau and Zumer tout as the scourge of the gold standard dummy is the debt burden: interest payments divided by revenue. This is the real determinant of interest rates, they claim, because this is what the Credit Lyonnais looked at, and because it just makes sense as an indicator of the likelihood of bankruptcy. Just as a bank lending to a homebuyer would want to know the debt burden of the potential buyer relative to his or her ability to pay, bankers at the turn of the century wanted to look at the debt burden of the governments to which they lent relative to the ability of the governments to pay. One problem with using this variable in a linear regression, however, is that the variable on the left hand side, the yield for a particular country less the U.K. yield, is closely related by construction to the variable on the right hand side, interest payments divided by revenues. Consider the following example. You have two countries identical in every way except A borrows at a high rate of interest and B at a low rate of interest. If they both borrow the same amount, interest payments will be higher for A than for B — interest rate and debt burden will be positively correlated. Or consider what happens if the rate on a government’s bond falls, and the government refinances some of its debt at the lower rate. Again, the dependent variable and independent variable move together by construction.

2. Upstream and downstream variables A second, perhaps more important, problem is that the regression strategy combines what I like to call upstream and downstream variables. An extreme example will make the distinction I have in mind clearer. Suppose someone claims that prohibiting construction near a lake increases the amount of fish caught in a lake. A regression of fish caught per acre of water on a dummy variable for whether or not construction was permitted near a lake might show that prohibiting construction increases the catch. Add the stock of fish per acre in each lake to the regression and the significance of the prohibition of construction might disappear. The number of fish in the lake determines how many will be caught. One could then jump to the conclusion that prohibiting construction near lakes has no effect. Presumably, with enough data one could tease out the effects of limiting construction and other variables that might affect the amount of fish in the lake. In practice, however, this might be difficult if there are a small number of lakes in the sample.

The same problem, arises, I believe, in the current context. One example is the inclusion of both exchange rate volatility and the gold standard dummy in the same equation. Exchange rate stability (which turns out to have a large and significant coefficient in many of the regressions) is downstream from the gold standard. We might substitute it for the gold standard dummy, but to include both exchange rate volatility and the gold standard dummy obscures the effect of the choice of exchange rate regime.

The debt burden variable is also, to some extent, a downstream variable. The point of attaching one’s currency to gold was not to fool investors, while going about one’s old spendthrift ways. The point was to achieve long-run discipline. A country could not run continual deficits and inflate them away and remain permanently on the gold standard. Defending the currency was the principle that allowed central banks and governments to follow more conservative policies than they otherwise would. Ultimately, of course, the goal was to have modest deficits and a debt well within the capacity of the country to service. Putting all of these variables into the right hand side of the regression tends to obscure the effect of adhering to gold on the achievement of these long-term goals.

3. The gold standard dummies While the authors have put a great deal of energy into perfecting the other variables in the equation, they have put no energy into improving the gold standard variables. Indeed, it seems at points as if they can hardly be bothered to look at such nonsense. It is clear to me, however, that the current gold-standard variables are rather primitive because they fail to reflect the credibility of the commitment to the gold standard. True, credibility is not easily measured, but it is none-the-less crucial. The United States is a good example. Flandreau and Zumer, as I noted, exclude the United States. But it is the case I know best, and it makes the point. The dollar was convertible into gold, except during financial crises, from the time that convertibility was established after the Civil War until the Great Depression. One could represent this by simply assigning a dummy variable of one (on the gold standard) in every year in the period that Flandreau and Zumer examine. Yet the credibility of the U.S. commitment to the gold standard varied. When the Free Silver movement was at its height in the 1890s the fear that the United States would leave the gold standard was real. Milton Friedman and Anna Schwartz, looking at short-term interest rates, conclude that the resolution of the fear that the United States would leave gold explains a sharp drop in the level of the short-term U.S.-U.K. differential between 1874 and 1896. According to Friedman and Schwartz (1982, 515) peaks in the differential in 1893 and 1896 are consistent with this interpretation.

“The peak in 1893 is connected to the banking panic in that year. The initial banking difficulties reinforced fears, endemic before 1896 because of silver politics, that the United States would go off gold and the dollar would depreciate. … The peak in 1896 is connected with the capital flight of that year accelerated by Bryan’s nomination, which greatly strengthened fears that the United States would leave gold. In both cases, fear of devaluation meant that owners of United Kingdom capital were reluctant to participate in the United States short-term market except at a substantial premium. The election of McKinley changed the situation drastically. It made United States’ retention of the gold standard secure for the time being, and the subsequent flood of gold from South Africa, Alaska, and Colorado removed all doubts.”

Charles Calomiris (1992), similarly, thought that the threat of free silver affected the capital market, although he argued that the markets simply feared a temporary suspension of convertibility and post-suspension devaluation, rather than a permanent abandonment of gold. The point, however, is that even in the case of the United States which in the end remained solidly committed to gold, credibility varied, and that a dummy variable that simply looks at whether convertibility was maintained during a particular year is insufficient to capture the credibility of the commitment.

4. The changing credibility of the gold standard The credibility of the gold standard itself changed over the period 1880 to 1913. The correlation between the gold standard dummies and the defaults shows why. In 1880, the first year in the sample, 10 of the 17 countries in the sample were adhering to the gold standard. Up to 1913 only one, Portugal, would default. One country, Spain, was off gold and in default in 1880. The remaining six countries in the sample were off gold in 1880, but paying their debts. Three of the six would later default. Being on or off gold in 1880, in other words, turned out to be a good predictor of which countries would pay their debts. I suppose that participants in international financial markets might have ignored this information on the grounds that it was irrelevant. The ratio of interest payments to government revenues was equally good as a predictor of default. Nevertheless, adhering to gold was a device for achieving long-term stability. Perhaps recognition of the connection between being off gold and defaulting is why countries made an effort to stay on or get on gold. Of the 10 countries on the gold standard in 1880 only one was off in 1913. Of the seven countries off the gold standard in 1880, six were on by 1913 including all those that defaulted. In short, of the 17 countries in the sample, 15 had made the decision by 1913 to adhere to the gold standard. Given this scenario, it is possible that the credibility of the gold standard itself, as a means of achieving and as a symbol of financial rectitude was increasing over the period 1880 to 1913. The regression strategy as far as I can see simply assumes that being on gold in 1880 meant the same thing to participants as being on gold in 1913.

5. Policy variables and non-policy variables One reason why (some) economists and economic historians focus on adherence to the gold standard is because this was a policy variable, a choice actually being made by many countries at the turn of the century. Again, this was clearly true in the United States. Free silver (bimetallism at a ratio of 16 units of silver to one of gold) was a genuine alternative to the gold standard. It was promoted by one of the two major political parties, and might have been adopted by the United States. Fiscal policies, perhaps to a lesser degree, fall in the same category. It is important to see what might have been driving these debates, and what effects the choices made had on the U.S. economy.

There are other potential fundamentals that might have been more important than adherence to the gold standard in shaping the flow of capital, but were not policy variables. It may have been extremely important to British lenders that the population of a country be predominantly white, Anglo-Saxon, and Protestant. Being a colony of Britain may have been important as well. And having a high level of education may have impressed potential investors. Of course, if these factors were important they must be given their due in order to see what impact policy choices actually had. But the focus on policy variables, even when their potential contribution is marginal, is important if history is to provide lessons for today.

To be sure, what is or is not a policy variable is to some extent a matter of costs and benefits. Educational levels, colonial status, even religion can be, and in some cases have been deliberately changed. And it is always possible to look at a more fundamental analysis in which what look to be policy choices are really predetermined outcomes. There are variables, in other words, that explain why a country “chose” to adhere or not adhere to gold. When the election of 1896 began it seemed as if Bryan might win, and so Americans debated his policies at length. It seems as if Americans could make a choice. Yet it may also be true that there exists some model of the political process that takes into account various political fundamentals (the ethnic mix of the population, the rate of growth of the economy, and so on) that would show us why Bryan was destined to lose. Hillel is right: “all is foreseen, and freewill is given.”

The lack of clear distinctions between policy and non-policy variables affects the usefulness of the concluding section of the book. Flandreau and Zumer feel that the choice of exchange rate regime didn’t matter. This is useful advice — don’t waste your time worrying about gold or bimetallism. Perhaps today the advice would be don’t worry about fixed or flexible exchange rates. The debate over free silver in the United States and many other countries was, evidently, a lot of sound and fury signifying nothing. But what should people have worried about? Here Flandreau and Zumer make a number of assertions, but few of them qualify as practical advice. They note that “The adoption of ‘good’ [their quotation marks] domestic policies expedited the globalization of capital much more decisively than did the removal of legal barriers to financial exchange.” They also note that “debt burden was the one key factor that determined market access.” And they note that “Political crises such as wars or domestic unrest were detrimental to a country’s credit.” And that “?the reduction of public debts was achieved not through fiscal balance but via [their emphasis] economic growth,” which leads them to conclude that this “shows the importance of development policies in fostering international financial integration …” But what are we to make of all this? Surely every politician and government official knew that debt burdens and political crises were bad and that economic growth was good. The question was not where they wanted to go, but what policies would help them get there.

If I were starting to work on the “Good Housekeeping” paper again I would certainly write it differently. I would certainly want to take into account the important work by Flandreau and Zumer based on the archives of the Credit Lyonnais to better describe the channels through which adherence to gold mattered. But I am not persuaded that the gold standard was a matter of no concern, and that all of the debate over it at the end of the nineteenth century was a waste of time. I am persuaded, however, that Flandreau and Zumer are fine scholars, and have written an important book that future research in this area will need to take into account.

References:

Michael Bordo and Hugh Rockoff, “The Gold Standard as a Good Housekeeping Seal of Approval.” Journal of Economic History 56 (June 1996): 389-428.

Charles Calomiris, “Greenback Resumption and Silver Risk: The Economics and Politics of Monetary Regime Change in the United States, 1862-1900.” NBER Working Paper, w4166, September 1992.

Milton Friedman and Anna J. Schwartz, Monetary Trends in the United States and the United Kingdom: Their Relation to Income, Prices, and Interest Rates, 1867-1975. Chicago: University of Chicago Press, 1982.

Maurice Obstfeld and Alan M. Taylor, “Sovereign Risk, Credibility and the Gold Standard: 1870-1913 versus 1925-31.” Economic Journal (April 2003): 241-75.

Nathan Sussman and Yishay Yafeh. “Institutions, Reforms, and Country Risk: Lessons from Japanese Government Debt in the Meiji Era.” Journal of Economic History 60 (June 2000): 442-67.