An Assessment of the Causes of the Abandonment of the Gold Standard by the U.S. in 1933.

Paul Hallwood [*]

Ronald MacDonald [+]

Ian W. Marsh [++]

In this paper we present an investigation of the pressures on the United States to devalue the dollar against the franc and gold in the early 1930s. We calculate monthly time-series of realignment expectations and find that these are well explained by a set of fundamental economic variables. The implication is drawn that macroeconomic events were at least in part responsible for jolting the U.S. off the gold standard and that the Federal Reserve was constrained in its response to the Depression by the United States' commitment to gold.

1. Introduction

The indictment that the Fed played a causal role in the Great Depression in the United States is still of interest because it challenges the credibility of the Fed as an institution (Friedman and Schwartz 1963). Indeed, this matter has remained controversial, witness a recent conference where papers on material relevant to the issue by Temin (1998) and Sims (1998) attracted often contrary discussion. Moreover, whether the Great Depression in the United States was home-grown or was caused at least in part by the international economic contraction has been the subject of many earlier investigations including Eichengreen (1992), Kindleberger (1986), Temin (1989), and Wigmore (1987).

In this paper we investigate the role of the Fed in the Great Depression using ideas drawn from the recently developed theory of target zones (Krugman 1991) in combination with a reduced-form version of the monetary approach to the balance of payments. As money is endogenous in the monetary approach, any success of our modeling would indicate that the Fed was not a free agent in determining the American money base and interest rates. We in fact do find that the Fed managed money in a manner consistent with the United States' commitment to maintain the gold standard--a commitment that was widespread across the American political scene in the early 1930s (Eichengreen and Temin 1997).

We proceed to investigate the independence or otherwise of American monetary policy under the interwar gold standard in the following way. As we view the United States as being a part of an international system, section 2 describes relevant concurrent macroeconomic events in France--the other major country still on the gold standard after speculative pressures in foreign exchange markets had ejected the United Kingdom from it in 1931:09. Section 3 describes a method for calculating franc-dollar realignment expectations, using the theory of uncovered interest parity, for the period spanning the United States' and France's simultaneous adherence to the gold standard in the interwar period, 1926:12 to 1933:02. Empirical realignment expectations are then described. Section 4 shows how the macroeconomic determinants of realignment expectations may be modelled, and section 5 discusses our empirical evidence on the determination of expectations. Section 6 offers conclusions suggesting that the Fed was not an actor autonomous of the international gold standard.

2. France

The United States cannot be characterized as a colossus bestriding a rump of disconnected small open economies still on the gold standard after the United Kingdom's ejection. The population of the four countries that were later to form the gold bloc after the United States suspended gold in 1933:03--Belgium, France, Netherlands, and Switzerland--had a combined population that was almost exactly one-half that of the U.S. (League of Nations 1931/2, table 2). And, although their per capita incomes were lower than the United States', we know that this bloc was large enough to absorb copious quantities of American monetary gold.

In this study French macroeconomic data effectively proxy for that of the other gold bloc countries. This is justified as, outside the United States, the French economy was the largest still on gold. Secondly, for key macroeconomic variables--money, prices, and production--French monthly data are highly correlated with that of the other three countries. Thus, in the respective correlation matrixes for the four gold bloc countries over the period 1928:06 to 1933:02 most correlations are well over 0.90.

French monetary policy after the Great Crash in 1929:10 was more helpful to the Bank of England in its struggle to 1931:09 to stay on gold than it was after this date to the Federal Reserve in the run up to the American suspension of gold in 1933:03. Thus, the French nominal money base increased from F88bn in 1929:10 to F104bn in 1931:09--an 18% increase. [1] But in the crucial period for the Federal Reserve following the pound's departure from gold, 1931:10 to 1933:02, the French nominal money base contracted by 9%--from Fl15bn to Fl05bn. At the same time the U.S. money base increased by 16%--from $7.57bn to $8.8bn. This contraction of the French nominal money base was not due to falling French gold reserves because in these 17 months there were 15 straight months of bullion and specie inflows to France. The two odd months of outflows being the last two, possibly because increasing speculative pressure on the dollar eventually brought the franc under suspicion--Switzerland and the Netherlands receiving net bullion and specie inflows in these months.

This absolute and relative (to the United States) tightening of the French nominal money base was more unhelpful to the Federal Reserve than these figures indicate. This is because the demand for money and French interest rates were being supported by a smaller contraction in French industrial production compared to that in United States. Thus, French and U.S. industrial production fell by, respectively, 27% and 47% 1929:10 to 1933:02; and in the critical 17 months after the U.K. left gold by, respectively, 12% and 15% (i.e., 1931:10 to 1933:02).

U.S. short-term interest rates (90-day prime bankers' acceptances) during these 17 critical months fell sharply from 2.25% to 0.44%; while French short-rates (private discount rate) fell only from 1.8% to 1.75%. (But they had been only 0.91% in 1932:12 before almost doubling by 1933:02, presumably due to the disturbances on foreign exchange markets.)

3. Calculating Realignment Expectations

Countries on the gold standard defined a price of gold at which domestic currency was convertible. In turn these commitments set the mint par parity between a pair of currencies. Owing to the cost of arbitraging gold between countries, the gold import and export points implicitly defined the edges of an exchange rate fluctuation band or target zone. If financial markets thought that national commitments to convertibility at unchanged prices were credible, the expectation for mint par parity would be for no change and the exchange rate would not be expected to move outside of the established gold points. However, if for some reason a country's commitment to a fixed gold price was questioned, there would be some finite probability of a change in mint par parity (or abandonment of the gold standard altogether) and the exchange rate would deviate outside of the gold points.

To calculate franc--dollar realignment expectations during the interwar gold standard we use the method of Svensson (1991, 1993) and Bertola and Svensson (1993). A target zone as defined by Krugman (1991) is credible in the sense that the market believes that the exchange rate will be contained within the zone, if it is believed that the authorities are committed to managing economic fundamentals such as interest rates or the money supply to this end. [2]

In natural logarithms, define the current exchange rate, [s.sub.t], as

[s.sub.t] = [x.sub.t] + [c.sub.t] (1)

where [c.sub.t] is central rate [3] and [x.sub.t] is the proportionate deviation from it. Taking time derivatives

[E.sub.t][d[s.sub.t]]/dt = [E.sub.t][d[x.sub.t]]/dt + [E.sub.t][d[c.sub.t]]/dt (2)

where [E.sub.t] is the rational expectations operator. Thus, the rationally expected rate of change of the exchange rate can be divided into the expected movement within the band, ([E.sub.t][d[x.sub.t]]/dt), plus the expected rate of change of the central parity, ([E.sub.t][d[c.sub.t]]/dt).

On rearranging Equation 2, a statement of the rationally expected realignment expectation is obtained:

[E.sub.t][d[c.sub.t]]/dt = [E.sub.t][d[s.sub.t]]/dt - [E.sub.t][d[x.sub.t]]/dt. (3)

This realignment expectation can be calculated if we know both the expected change in the exchange rate, [E.sub.t]/[ds]/dt; which is calculated from the forward premium assuming uncovered interest parity, and the expected movement of the exchange rate within the band, [E.sub.t][dx]/dt.

The expected movement of the exchange rate within the band may be estimated in the following way:

[x.sub.t+m] - [x.sub.t] = [a.sub.0] + [a.sub.1][x.sub.t] + [u.sub.r] (4)

That is, the expected movement in the band depends on the current deviation from the center of the band. If [a.sub.1] is significantly less than zero, mean reversion is occurring. [4]

Table 1 shows the results for estimating Equation 4. As the t-statistic on [a.sub.1] is large enough on an ADF test to reject the null of a unit root we are confident that the franc--dollar exchange rate was mean reverting within the gold points.

The final step in implementing Equation 3 is to take our 95% confidence interval for mean reversion (calculated using Eqn. 4) and combine it with the forward premium data (the proxy for [E.sub.t][ds]/dt) to calculate the 95% confidence intervals for realignment expectations.

Figure 1 shows realignment expectations for the franc--dollar exchange rate, the upper and lower bounds to the 95% confidence interval denoted with dotted lines. Lower values signify an expected depreciation of the U.S. dollar. The expected rate of realignment can be interpreted as the expected devaluation size multiplied by the frequency of realignment. Suppose that, conditional on there being a devaluation, the devaluation will be 10%. An expected rate of realignment of 5% implies that the expected frequency of realignment is 0.5 per annum. That is, the market expects a 10% devaluation within the year to happen with a 50% probability. Thus, though the average expected rate of realignment may appear to be small, it can be consistent with quite substantial devaluation expectations. Of course, when the confidence interval spans zero we cannot reject the hypothesis that the expected probability of a devaluation of any magnitude is zero.

In financial markets the U.S. dollar was deemed to be securely on the gold standard until the month following the pound's ejection from it in 1931:09. From then until the United States suspended the gold standard in 1933:03, the dollar as often as not experienced statistically significant expectations of being devalued. Notice that the dollar was not under threat during the first banking crisis beginning in 1930:10 nor during the second beginning in 193l:03. [5] So in both periods the Fed most probably had room to lower interest rates, and this would have been helpful to the United Kingdom in its struggle to stay on gold. But how much room is another matter. When the Fed did try open market purchases during the first-half of 1932 in. an effort to stimulate the economy, realignment expectations turned sharply against the dollar. Also, notice that the dollar was under pressure from 1933:1, concurrent with the third banking crisis.

We draw the conclusion that financial markets anticipated the United States' suspension of the gold standard well in advance of the event. It could not have been a great surprise in informed financial circles when Roosevelt gave the order for suspension.

4. A Model of the Determinants of Realignment Expectations

In this section we investigate whether realignment expectations ([E.sub.t][d[c.sub.t]]/dt), for the franc--dollar exchange rate can be rationally explained by the behavior of economic fundamentals. The variables that we use in our statistical test have been chosen for a number of reasons: first, because they are consistent with the monetary approach to the exchange rate and the exchange market pressure monetary model of Girton and Roper (1977); secondly, because following Caramazza (1993) and Chen and Giovannini (1994) it seems reasonable to include them as affecting expectations in a reduced form estimating equation; and thirdly, because monthly time-series data on them are available.

Thus, we include as determinants of franc--dollar realignment expectations in Equation 5 a measure of the balance of trade position, money growth, changes in reserves, real income growth, and international competitiveness as proxied by changes in the exchange rate. Our estimating equation is

[E.sub.t][d[c.sub.t]]/dt = [[beta].sub.0] + [[beta].sub.1][BoT.sub.t-1] + [[beta].sub.2][BoT.sub.t-2] + [[beta].sub.3][delta][M.sub.t-1] + [[beta].sub.4][delta][M.sub.t-2] + [[beta].sub.5][delta][Res.sub.t-1] + [[beta].sub.6][delta][Res.sub.t-2]

+ [[beta].sub.7][delta][IP.sub.t-1] + [[beta].sub.8][delta][P.sub.t-2] + [[beta].sub.9][delta][P.sub.t-1] + [[beta].sub.10][delta][P.sub.t-2] + [[beta].sub.11][delta][s.sub.t] + [[beta].sub.12][delta][s.sub.t-1]

+ [[beta].sub.13][delta][s.sub.t-2] + [[beta].sub.14][D.sub.t] (5)

where BoT denotes the relative balance of trade ratio (French exports/imports -- U.S. exports/imports), M the relative money supply (France - U.S.) discussed further below, Res denotes relative reserves, IP relative income (proxied by industrial production), P relative wholesale price levels, and s denotes the nominal exchange rate. A [delta] denotes first differences, and because all transformed variables are in log levels, differences become percent changes. Two lags of each fundamental determinant are included in the regression to allow for information delays. D is a dummy variable taking the value of unity once the United Kingdom leaves the Gold Standard in 1931:09. All independent variables except the spot exchange rate are seasonally adjusted. Seasonal dummies are also included in the regressions to account for residual seasonality or, conceivably, some seasonality in devaluation expectations. A data appendix lists sources.

The expected signs on the coefficients of Equation 5 are such that a stronger U.S. balance of trade relative to France increases confidence in the dollar and reduces pressure against it; an increase in U.S. money stock induces dis-hoarding and increases pressure against the dollar; an increase in U.S. reserves reduces pressure against the dollar; an increase in U.S. income induces hoarding, thus strengthening the dollar; and a rise in relative U.S. inflation represents a loss in competitiveness and is expected to increase dollar devaluation expectations. The other coefficients have less obvious interpretations, and we remain agnostic at this stage over their expected signs.

5. Empirical Evidence on the Proximate Causes of Realignment Expectations

Because a focus of this paper is the actions of the Fed, we have experimented with different definitions of the money supply. The results below relate to regressions in which money is defined relatively broadly as M2. However the conclusions are unchanged if we instead use the money base as our measure.

The regression results for Equation 5 are shown in the first colunm of Table 2. The fully parameterized model suggests that the variables are capable of capturing some of the movements in realignment expectations. The [R.sup.2] is an encouraging 0.75, and a test of the joint insignificance of the economic variables is strongly rejected. Nevertheless, some of the variables are wrongly signed and several have unstable signs across lags, both of which comments are applicable to the money supply terms.

A more parsimonious model was sought, using a general-to-specific methodology. The model is derived by sequentially dropping variables not significant at the 10% level. The resulting equation is given in the second column of Table 2. The dummy variable raises dollar devaluation expectations by almost 3% over the final 17 months of the sample. We discuss the significance of this variable in more detail below. Changes in the spot exchange rate are also significant and are such that an appreciation of the dollar reduces dollar devaluation expectations, consistent with intuition. Several researchers have found that the level of the exchange rate is significant in explaining expectations (Caramazza 1993; Mizrach 1995). Our data suggested that changes in the exchange rate are more relevant--including two lagged levels simply resulted in equal and opposite coefficient signs.

Of the four significant fundamental determinants of realignment expectations three are correctly signed--the condition of relative reserves (because the sum of the coefficients is negative), balances of trade, and industrial production. The inflation differential bears a large coefficient but of theoretically incorrect sign. Money is not a significant determinant of realignment expectations but is included in the model given its importance in the debate about the driving forces behind realignment expectations. It is also insignificant if a more narrow definition is applied.

Figure 2 plots the time series of realignment expectations derived in section 3, together with the fitted values from the parsimonious version of Equation 5. In the relatively tranquil period up to late 1931 the fitted series follows the derived expectations tolerably well. The sharp movement to significant dollar devaluation expectations in late 1931 is captured by the model. For eight months following the United Kingdom's move off gold, dollar devaluation was expected. Two large deteriorations in the dollar's position are clear, one of which is captured by the fundamental determinants. The spike in 1932 is not well accounted for, but the subsequent return to expected stability in late 1932 is explained by developments in the fundamental variables, as is the final downturn just before the dollar floated.

Which of the variables are responsible for the significant changes in realignment expectations? Figure 3a and b plots the contributions to expectations made by various determinants retained in the parsimonious model over the period 1931:01-1933:02. These are calculated simply by setting all variables apart from the one under discussion to zero, and then calculating the fitted value from the parsimonious model. The absolute level of the contribution has no clear interpretation because a constant is included in the model, but the path of each can be used to explain the evolution of devaluation expectations.

The shift to high pressure on the dollar in late 1931 is almost entirely explained by the dummy variable D (denoted dummy in Figure 3a). Unfortunately, it is not clear which factor this dummy variable is capturing. One interpretation is that it is related to the wave of banking crises affecting the United States at this time (Friedman and Schwartz 1963). In light of the recent experiences in Asia, an alternative is that it represents a contagion effect, with the U.S. coming under speculative pressure following sterling's move off gold (although why it was attacked rather than France is an open question). We remain agnostic as to the correct interpretation but note that after sterling left gold/the U.S. banking crisis worsens, and the U.S. dollar was fighting against a 3% realignment expectation.

The slight downward trend from 1931:10 to 1932:5 appears to be due to a deterioration in the relative balance of trade position of the United States (Figure 3b). Similarly, recovery in the same variable drives expectations back toward zero in mid-1932. The effects of changes in reserves follow a similar pattern, but the impact on expectations is not so large. The contribution of inflation developments is negligible during the crisis years (Figure 3a). This is encouraging given that inflation effects bear the theoretically wrong sign in regression Equation 5. The contribution of relative money growth over this period was statistically insignificant in the model. It is also insignificant in a practical sense, barely registering in Figure 3b. Finally, note that much of the volatility in devaluation expectations is driven by changes in the spot exchange rate (Figure 3a). The slight recovery in 1931:11 and first downward spike in 1931:12 are due to an appreciation and subsequent depreciation of the dollar. Anothe r rise in its value helped the mid-1932 recovery. But the 1% drop in the dollar in 1933:01 appears to be the sole determinant of the final increase in pressure just before Roosevelt gave the order for suspension.

6. Conclusions

We have used a two-stage procedure to provide an assessment of the causes of the United States' abandonment of the gold standard in March 1933. In the first stage a procedure due to Krugman (1991) and Bertola and Svensson (1993) was used to calculate dollar realignment expectations. This showed that while financial markets thought that the dollar was securely on the gold standard until the United Kingdom abandoned it in 1931:09, after that date financial markets became skeptical of the United States' willingness or ability to do so.

The second step of our approach was to discover the determinants of dollar realignment expectations. A set of explanatory variables--gold reserves, money stock, real GDP (proxied by industrial production), and the balance of trade--was chosen, and our time series of franc-dollar realignment expectations regressed on it. By the standards of exchange rate models we obtained good explanatory power and all significant variables bore the correct theoretical signs.

The nonsignificance (but with the correct theoretical sign) of the U.S. monetary variable has an interesting interpretation: namely, that the Federal Reserve was managing money in accordance with the United States' commitment to the gold standard. Certainly, adherence to a pegged exchange rate regime such as the gold standard requires that monetary policy be disciplined by this commitment. We are skeptical therefore of claims that the Federal Reserve was somehow responsible for bringing the Depression on the United States because it mismanaged the money supply. What is apparent from our discussion is that if it had chosen to target money on U.S. real economic variables, it would most probably have increased pressure on the dollar by increasing realignment expectations against it.

Finally, just as the position of the United Kingdom under the classical gold standard, circa 1875-1914, has been reappraised by applying the monetary approach to the balance of payments (McCloskey and Zecker 1976), so downgrading it from being "the conductor of the international orchestra" to the "triangle player," [6] so must the position of the U.S. during the interwar period be reappraised. The United States was ejected from the gold standard because its macroeconomic fundamentals got out of line with those of other members of the system. The main problem for the United States was that French interest rates increased relative to American rates, and gold flowed from the latter to the former country, eventually requiring macroeconomic adjustment in the United States, namely, a reduction in the demand for money via any combination of higher U.S. interest rates, lower prices, or lower production. It is not a coincidence that gold was suspended amidst the third banking panic as Roosevelt moved decisively to sa ve the banks and stimulate the economy by lowering interest rates. We think that this would not have been possible had the United States continued to adhere to the gold standard because realignment expectations would have gone even more strongly against the dollar.

(*.) Department of Economics, University of Connecticut, U-63, Storrs, CT 06289, USA; E-mail [e-mail]; corresponding author.

(+.) Department of Economics, Strathelyde University, 100 Cathedral Street, Glasgow G4 0LN, UK; E-mail [e-mail].

(++.) Department of Banking and Finance, City University Business School, Barbican Centre, London EC2Y 8HB, UK, and CEPR, Goswell Road, London EC1V 7RR, UK; E-mail [e-mail].

We thank the coeditor, Kent Kimbrough, and two anonymous referees for comments on earlier drafts. The usual disclaimer applies.

Received May 1998; accepted November 1999.

(1.) This compares with only a 2% increase in the United States over the same period (Friedman and Schwartz 1963, table B3). Note that quoting real values will not make much difference as French and U.S. price levels moved closely together.

(2.) A critique of the target zone theory is found in Flood, Rose, and Mathieson (1991). Target zone theory can also be compared to empirical examinations of breakdowns of fixed exchange rate regimes using speculative attack models. Typically in these papers a time-varying shadow exchange rate, [[s.sup.*].sub.t], is computed based on some fundamental determinants. A devaluation will occur if [[s.sup.*].sub.t] exceeds the fixed parity. However, in many cases [[s.sup.*].sub.t] is substantially below the fixed parity level, and yet there is little explanation of why capital does not flow into the country prompting a revaluation (e.g., Blanco and Garber 1986, Figure 1). While capital controls and risk premia might be valid arguments for a small (developing) country case, this would appear to be a major weakness when considering large countries as we do. Perhaps it is because of oddities such as this that Agenor, Bhandari, and Flood call for speculative attack modeling to be extended to include the target zone rea lity (1992, p. 383).

(3.) The central rate, [c.sub.t], is not necessarily mint par parity for technical reasons due to asymmetric costs depending on the direction in which gold was being arbitraged between France and the United States.

(4.) Strictly speaking, in the Krugman model [x.sub.t+m] - [x.sub.t] is a nonlinear function of [x.sub.t]. However, Bertola and Svensson (1993) and Caramazza (1993) use a linear function as a good approximation.

(5.) This dating of banking crises is taken from Friedman and Schwartz (1963) chart 30.

(6.) The reference here is to Tullio and Wolters (1996).

References

Agenor, Pierre-Richard, Jagdeep S. Bhandari, and Robert P. Flood. 1992. Speculative attack models and balance of payments crises. IMF Staff Papers 39:357-94.

Bertola, Giuseppe, and Lars E. 0. Svensson. 1993. Stochastic devaluation risk and the empirical fit of target-zone models. Review of Economic Studies 60:689-712.

Blanco, Herminio, and Peter M. Garber. 1986. Recurrent devaluation and speculative attacks on the Mexican peso. Journal of Political Economy 94:148-66.

Caramazza, Francesco. 1993. French--German interest rate differentials and time-varying realignment risk. IMF Staff Papers 40:567-83.

Chen, Zhaohui, and Alberto Giovannini. 1994. The determinants of realignment expectations under the EMS: Some empirical regularities. London School of Economics Financial Markets Group Discussion Paper No. 184.

Eichengreen, Barry. 1992. Golden fetters: The gold standard and the great depression, 1919-1939. Oxford: Oxford University Press.

Eichengreen, Barry, and Peter Temin. 1997. The gold standard and the great depression. NBER Working Paper No. 6060.

Einzig, Paul. 1937. The theory of forward exchange. London: MacMillan.

Federal Reserve Board, 1943. Banking and monetary statistics, 1914-1941. Washington, D.C.

Flood, Robert P., Andrew K. Rose, and Donald J. Mathieson. 1991. An empirical exploration of exchange rate target zones. Carnegie--Rochester Series on Public Policy 35:7-66.

Friedman, Milton, and Anna J. Schwartz. 1963. Monetary history of the United States, 1867-1960. Princeton, NJ: Princeton University Press.

Girton, Lance, and Don Roper. 1977. A monetary model of exchange market pressure applied to the postwar Canadian experience. American Economic Review 67:537-48.

Kindleberger, Charles P. 1986. The world in depression, 1929-1939. Berkeley, CA: University of California Press.

Krugman, Paul R. 1991. Target zones and exchange rate dynamics. Quarterly Journal of Economics 106:669-82.

League of Nations. 1931/2. international statistical yearbook. Geneva.

McCloskey, Donald N., and J. Richard Zecker. 1976. How the gold standard worked, 1880-1913. In The monetary approach to the balance of payments, edited by J. A. Frenkel and H. G. Johnson. Toronto: University of Toronto Press, pp. 357-85.

Mizrach, Bruce. 1995. Target zone models with stochastic realignments: An econometric evaluation. Journal of international Money and Finance 14:641-57.

Patat, Jean-Pierre, and Michel Luffall. 1990. A monetary history of France in the twentieth century. New York: St. Martin's Press.

Sims, Christopher A. 1998. The role of interest rate policy in the generation and propagation of business cycles: What has changed since the 1930s. In Beyond shocks: What causes business cycles? Federal Reserve Bank of Boston conference series No. 42. pp. 121-60.

Svensson, Lars E. 0. 1991. The simplest test of target zone credibility. IMF Staff Papers 38:655-65.

Svensson, Lars E. 0. 1993. Assessing target zone credibility: Mean reversion and devaluation expectations in the ERM: 1979-1992. European Economic Review 37:763-802.

Temin, Peter. 1989. Lessons from the great depression. Cambridge, MA: MIT Press.

Temin, Peter. 1998. The causes of American business cycles: An essay in economic historiography. In Beyond shocks: What causes business cycles? Federal Reserve Bank of Boston conference series No. 42. pp. 37-59.

Tullio, Giuseppe, and Jurgen Wolters. 1996. Was London the conductor of the international orchestra or just the triangle player? An empirical analysis of asymmetries in interest rate behaviour during the classical gold standard, 1875- 1913. Scottish Journal of Political Economy 43:419-43.

Wigmore, Barrie A. 1987. Was the bank holiday of 1933 caused by a run on the dollar? Journal of Economic History 47:139-55.

Appendix: Data Sources

Exchange Rates

Spot and 3-month forward rates on last Saturday of the month (Einzig 1937).

Reserves

United States: Gold reserves of Federal Reserve Banks, Federal Reserve Board (1943), NBER series 14062.

France: Gold reserves, League of Nations, Monthly Bulletin of Statistics.

Money Supply

United States: M2 and high-powered money, Friedman and Schwartz (1963).

France: M2 and liabilities of the Banque de France, Patat and Luffall (1990).

Price Level

United States: Wholesale prices, 1926 = 100, Federal Reserve Board (1943).

France: Wholesale prices, all commodities, 1913 = 100, Federal Reserve Board (1943).

Industrial Production

United States: Industrial production, 1923/25 = 100, Federal Reserve Board (1943).

France: Industrial production, 1913 = 100, League of Nations, Monthly Bulletin of Statistics.

Balance of Trade

United States: Total value of imports, Federal reserve Board (1943), NBER series 07028.

Total value of exports, Federal Reserve Board (1943), NBER series 07023.

France: Total value of imports, League of Nations, Monthly Bulletin of Statistics, NBER series 07032.

Total value of exports, League of Nations, Monthly Bulletin of Statistics, NBER series 07027.