Discount rate policy under the Classical Gold Standard: Core versus periphery (1870s–1914)

Explorations in Economic History 50 (2013) 205–226

Contents lists available at SciVerse ScienceDirect

Explorations in Economic History journal homepage: www.elsevier.com/locate/eeh

Discount rate policy under the Classical Gold Standard: Core versus periphery (1870s–1914)☆ Matthias Morys ⁎ Department of Economics, University of York, York YO10 5DD, United Kingdom

a r t i c l e

i n f o

Article history: Received 19 March 2012 Available online 31 December 2012 JEL classification: E4 E5 E6 F3 N13 N23

Keywords: Gold standard Central bank reaction function Rules of the game Balance-of-payments adjustment Central banking

a b s t r a c t Drawing on a new data set of monthly observations, this paper investigates similarities and differences in the discount rate policy of 12 European countries under the Classical Gold Standard. It asks, in particular, whether the bank rate policy followed different patterns in core and peripheral countries. Based on OLS, ordered probit and pooled estimations of central bank discount rate behaviour, two main findings emerge: firstly, the discount rate decisions of core countries were motivated by a desire to keep the exchange-rate within the gold points. In stark contrast, the discount rate decisions of peripheral countries reflected changes in the domestic cover ratio. The main reason for the difference in behaviour was the limited effectiveness of the discount rate tool for peripheral countries, which resulted in more frequent gold point violations. Consequently, peripheral countries relied on high reserve levels and oriented their discount rate policy towards maintaining the reserve level. Secondly, interest rate decisions were influenced by Berlin and London to a similar degree, suggesting that the European branch of the Classical Gold Standard was less London-centred than had been hitherto assumed. In establishing general patterns of discount rate policy, this paper aims to contribute to the wider discussion on monetary policy under the gold standard and the core–periphery dichotomy. © 2013 Elsevier Inc. All rights reserved.

1. Introduction The Classical Gold Standard (1870s–1914) has attracted the interest of economists, economic historians and policy-makers ever since its foundation. The exchange-rate stability among most countries of the world for some forty years was unprecedented and remained an inspiration to policy-makers after both world wars. At the time, adherence to gold was not entirely uncontroversial, as the international bimetallic movement of the mid-1870s to mid-1890s demonstrates (Reti, 1998). However, the perspective soon changed as a result of monetary instability following World War I and high exchange-rate volatility in the 1930s. Policy-makers came to regard the pre-World War I gold standard as the benchmark against which any international monetary system should be measured — hence the label Classical Gold Standard. Economists and economic historians, aware of the costs and benefits of adhering to a system of fixed exchange-rates, have tended to avoid the eulogistic tone of policy-makers. They have contributed to the gold standard myth, however, by producing a highly stereotypical account of its workings. Some of the stereotypes have surely been overturned by more recent research. Following ☆ I owe a special thanks to the following people for helping me collect the data: Kath Begley (Bank of England), Sofia Lazaretou (Bank of Greece), Alfredo Gigliobianco (Bank of Italy), Kalina Dimitrova (Bulgarian National Bank), Ms Beex (De Nederlandsche Bank), Ivo Maes and Arnold de Schepper (National Bank of Belgium), Virgiliu Stoenescu, Elisabeta Blejan, Brandusa Costache and Adriana Aloman (National Bank of Romania), Milan Sojic and Ljiljana Djurdjevic (National Bank of Serbia), Walter Antonowicz, Clemens Jobst, Bernhard Mussak and Thomas Scheiber (Österreichische Nationalbank), Erik Buyst, Martin Ivanov, Jan Tore Klovland, Larry Neal, Giuseppe Tattara, Anders Ögren, Lars Fredrik Oksendal and Marc Weidenmier. ⁎ Fax: +44 1904 323759. E-mail address: [email protected]. 0014-4983/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.eeh.2012.12.003

206

M. Morys / Explorations in Economic History 50 (2013) 205–226

Hume's price-specie mechanism (1752), the textbook account of the gold standard had it that gold was physically shipped between countries to settle balance-of-payments disequilibria. Recent research, following earlier leads (Lindert, 1969), has demonstrated the importance and sophistication of foreign exchange policy (Jobst, 2009). Other scholars have provided the empirical basis to verify or reject some of the claims made in the earlier gold standard literature, as in the discussion of the benefits of gold standard adherence which are seen in improved access to global capital markets and reduced transaction costs with other gold standard countries (Bordo and Rockoff, 1996; Lόpez-Cόrdova and Meissner, 2003). Yet another strand of the recent literature has highlighted conditions crucial to the workings of the Classical Gold Standard which had been neglected so far, such as the importance of labour mobility and remittances in smoothing the adjustment mechanism (Esteves and Khoudour-Castéras, 2009; Khoudour-Castéras, 2005). While the gold standard myth has given way to a broader empirical analysis in some debates, in other areas it stubbornly persists. One of these areas is the alleged core–periphery dichotomy. It is argued that the adjustment process to balance-of-payments disequilibria was much smoother for the industrialised core countries of North-western Europe as opposed to the peripheral economies. Different factors have been emphasized in an effort to explain the alleged advantages of the core countries in the adjustment process. Drawing on the theory of optimum currency areas, one school of thought has argued that core countries were better suited to monetary integration (Martín Aceña and Reis, 2000). Others have argued that the central banks 1 of core countries helped each other in times of crisis, but did not help peripheral economies for the lack of self-interest (Eichengreen, 1995 2). More recently, differences in credibility have been emphasized (Hallwood et al., 1996; Bordo and MacDonald, 2005), whereas an older school of thought highlighted the peripheral countries' role as debtors in the global financial system, which made them vulnerable to sudden withdrawals of funds in times of financial strain (de Cecco, 1974). Although the existing literature alludes to the core–periphery dichotomy, it is surprising to see that little effort has gone into analysing what exactly these differences consist of. A number of publications in recent years on the experiences of individual countries have greatly expanded our knowledge of the European periphery under the Classical Gold Standard (Esteves et al., 2009; Jobst, 2009; Reis, 2007; Tattara, 2003; Ögren and Oksendal, 2012). However, case studies, by design, can never be a substitute for a cross-country study analysing the similarities and differences between countries based on the same methodology. Such comparative studies on different aspects of monetary policy under the Classical Gold Standard exist, but they are mostly confined to comparing core countries (Giovannini, 1986; Contamin and Denise, 1999). This paper aims to provide the first systematic comparison of discount rate policy under the Classical Gold Standard based on the concept of a central bank reaction function. The discount rate was the most important monetary policy instrument at the time; Bagehot's famous description of the London money market, for instance, is almost exclusively concerned with the discount rate (Bagehot, 1878). Modern research, going back to Bloomfield's ground-breaking 1959 study (Bloomfield, 1959, p. 27), has followed this approach. Drawing on a sample of 12 European countries (Austria-Hungary, Belgium, Bulgaria, England, France, Germany, Italy, the Netherlands, Norway, Romania, Serbia, and Sweden) and relying on monthly data (the highest frequency available for all countries) we will analyse the determinants of discount rate policy; in particular, we will ask whether core and peripheral countries followed different patterns and, if so, explain why this was the case. In the process of collecting the data required for this analysis, it became clear why a comparative study of similar size and data frequency had never been conducted before. With the exception of England, Italy and Norway, the central banks did not make their historical balance sheet data publicly available. Most of the data (though not all) could be found in the Annual Reports of the time, copies of which can nowadays only be found in the archives of the respective central banks. Hence intensive collaboration with the central banks' historical archives was needed to reconstruct the time series.3 Which countries do we view as core and which ones as periphery? This dichotomy is often used in the literature but rarely defined based on rigorous foundations. A classification based on GDP per capita appears problematic in this context, as some countries generally considered peripheral would need to be classified as core (Argentina comes to mind and Ford's famous comparison with the UK (Ford, 1962)). In our context, any definition should rather capture the position in the international economy and, in particular, the international financial system. Liquidity in the foreign exchange market, for instance, provides evidence of the ability to attract short-term capital. Another potential indicator might relate to the ability to attract long-term capital: raising long-term capital is more difficult for countries suffering from original sin than for those able to access global capital markets in domestic currency. Classifying countries as core and periphery using these two or similar criteria might lead to conflicting results; the example of the US comes to mind, which Morgenstern and Schwartz classify as core, while Bordo and Eichengreen view it as peripheral in the pre-WW I financial architecture.4 Fortunately, our sample of 12 countries poses little risk of unclear classification. Based on an analysis of foreign exchange-market liquidity, Flandreau and Jobst (2005, p. 997) classify England, France, Germany, Belgium and the Netherlands as “key countries”5 of the international monetary system in 1880, a year which conveniently coincides with the beginning of our estimation period. We would choose the same five countries if we looked at the second criterion alluded to above, that is, the ability to issue sovereign bonds in terms of their own currency. From the 12 countries in our sample, Bordo and Flandreau

1 We will use the word “central bank” in the following, even though the transition to modern central banking had not yet been completed and the terminology “banks of note issue” would be more appropriate. 2 For a sceptical view towards this argument see Flandreau (1998). 3 For a full acknowledgment see p. 1. 4 We thank Michael Bordo for the discussion on how the view of the US in the pre-World War I financial architecture has changed over time in historiography. 5 Flandreau and Jobst have a classification into key-intermediary-periphery in mind rather than a dichotomy between core and periphery; the implication for our research is that we merge the second and the third group into a single group labelled “periphery”.

M. Morys / Explorations in Economic History 50 (2013) 205–226

207

(2003, p. 439) show that it is the same five countries which were in a position to issue debt abroad in their own currency. In the following, we will therefore categorise Belgium, England, France, Germany and the Netherlands as core countries and the other seven countries as periphery. The remainder of the paper is organized as follows. In Section 2, we will estimate central bank reaction functions for core and peripheral countries. Having established that the two types of countries reacted to different variables in setting their discount rate, Section 3 aims to explain the differences. Based on published sources and internal documents of the time, we will show that peripheral countries adapted the “English” gold standard to suit their needs; a modification which also impinged on how exactly the discount rate was to be used. Section 4 summarises and concludes.

2. Determinants of discount rate decisions 2.1. Components of a central bank reaction function Adherence to the gold standard implied two main constraints. First, in order to maintain confidence in the domestic monetary system, banks of note issue were required to hold a certain minimum level of reserves against bank notes in circulation. If the ratio between the two, usually referred to as the cover ratio, fell below some statutory level (typically 30%–40%), some sort of sanction would be applied. In extremis, this could lead to the complete loss of the note issuing privilege granted by the government (Reichsbank, 1925). The second constraint relates to the international dimension of the gold standard, that is, the need to keep the exchange-rate within the gold points. As the vast majority of transactions in the late 19th century were settled by bills of exchange (Denzel, 2008), the exchange-rate (i.e., the price of a bill of exchange drawn on a foreign country) could deviate from mint parity (i.e., the “standard” exchange-rate implied by the respective mint ratios). Only over and above some level of deviation, known as the gold export point, would it be advantageous to settle debt with (physical) gold, transaction costs (mainly shipping and insurance) notwithstanding. As such a withdrawal of gold from domestic circulation led, directly or indirectly, to a loss in reserves, central banks used the discount rate to prevent the exchange-rate from breaking out of the gold points. Third, discount rate decisions might be explained by patterns of interest rate followership vis-à-vis England, France and Germany. In a world of fixed exchange-rates and capital mobility, the macroeconomic policy trilemma implies that “small countries” have to follow the interest rates set by “large countries” (Obstfeld et al., 2005). Ideally, real economic variables such as output, unemployment or whether the economy is in an expansionary or contractionary phase of the business cycle, would be included, for, after all, central bank reaction functions are meant to shed light on the trade-off between the external constraint (or, in today's setting, the inflation target) and the incentive to stimulate the domestic economy. This approach is well exemplified by Eichengreen and Watson and Grossman's (1985) study of the Bank of England interwar discount rate policy, where real economic conditions are captured by a monthly series on output and unemployment. Real economic data of such frequency are not even available for England, France and Germany for this period,6 let alone for the other 9 countries, many of which do not even have decent annual GDP estimates. We can therefore only point out the problem and move on.

2.2. Modelling a central bank reaction function Our model emerges naturally from our previous discussion. The discount rate is modelled as a function of the exchange-rate, the cover ratio and interest rates set abroad. As for exchange-rate and interest rate, we choose England, France and Germany as the reference points, following widespread practice in the literature (Contamin and Denise, 1999; Tullio and Wolters, 2007). Two more decisions need to be taken: firstly, should specific discount rate changes or, more generally, discount rate behaviour be modelled? In the first case, we wish to explain why, for instance, the Bank of England increased its discount rate by 100 basis points on 31st October 1907 (the American banking crisis reached England). Some of the literature follows exclusively this approach (Tullio and Wolters, 2007), but in confining our sample to actual discount rate changes we miss an important dimension: if central banks leave the discount rate unchanged at their monthly meeting, this also constitutes a decision worth investigating. Consequently, we will model discount rate changes as well as monthly discount rate behaviour. The frequency of the second approach will be monthly; this is the highest frequency available (for all countries simultaneously), but it also seems justified, as our sources suggest that most central banks at the time had one (regular) meeting each month. Secondly, shall we pool the data or conduct country-specific regressions? The appeal of the latter option is to pursue a generalto-specific approach, thereby establishing which regressors mattered for which country. On the other hand, direct comparability of results is better ensured by pooling the data with country-specific (or group-specific) regressors. Moreover, country-specific regressions might constitute a case of seemingly unrelated regressions (SUR) and hence be econometrically questionable. As each approach has its own advantages, we decided to pursue both.

6 Bordo and MacDonald (2005) are faced with the same problem and are forced to rely on an interpolated series for England, France and Germany. As our sample of countries is larger, we cannot go down that road.

208

Table 1 Exchange rate performance (measured against mint parity) with respect to England, France and Germany. England

France Germany

Germany

France

Maximum deviation Standard deviation Max. dev. Avg. max. dev. Avg. st. dev.

1.0051 0.0026

Frequency (entire period, in %) of gold point violation xr >1.0060 xr >1.0100

2.7

3.7

1.4

0.0

4.7

0.0 0.0

0.0 0.0

0.0 0.0

0.0 0.0

Frequency (reduced period, in %) of gold point violation xr >1.0060 xr >1.0100

0.0

0.0

0.0

0.0 0.0

0.0 0.0

Time span of xr availability

1877m10 1914m6

1877m10 1914m6

1.0059 0.0024 1.0059 1.0055 0.0025

England

England

France

Germany

England

France

Germany

England

France

Germany

1.0068 0.0039 1.0068 1.0044 0.0029

1.0023 0.0024

1.0098 0.0029

1.0057 0.0016 1.0119 1.0091 0.0025

1.0119 0.0031

1.0116 0.0024

1.0111 0.0029 1.0116 1.0095 0.0022

1.0057 0.0012

6.6

0.0

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

0.3 0.0

0.5 0.0

0.0 0.0

0.3 0.0

0.0 0.0

10.3 7.1

0.0 0.0

11.8 12.1

4.2 0.5

4.2 0.5

0.0 0.0

0.0

1.7

2.9

0.0

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

0.0 0.0

0.0 0.0

0.0 0.0

0.3 0.0

0.0 0.0

0.3 0.0

0.0 0.0

0.0 0.0

0.0 0.0

0.3 0.3

0.3 0.0

1.1 0.0

0.0 0.0

1880m1 1913m12

1877m10 1914m6

1880m1 1913m12

1876m1 1914m6

1880m1 1913m12

1875m1 1914m7

1875m1 1914m7

1880m1 1913m12

1877m1 1913m12

1877m1 1913m12

1875m1 1914m6

1875m1 1914m6

1873m3 1914m6

1.0037 0.0050 1.0043 1.0040 0.0036

England 1.0063 0.0024

France

Sweden

1.0041 0.0023

1.0043 0.0022

Germany

Belgium

1.0062 0.0032 1.0063 1.0062 0.0028

Note: Calculations relate to the estimation period in Section 2 or, in the case of Bulgaria and Serbia, to the period of gold standard adherence (1/1906–9/1912 and 7/1905–9/1912, respectively). The reduced period excludes October 1899–July 1901 (Boer War), October 1906–February 1909 (crisis of 1907) and September 1912–December 1913 (Balkan Wars). The crisis of 1907 is often reduced to the American banking crisis in late 1907 and early 1908, even though it had begun the year before and its reverberations were felt until early 1909 (Kindleberger, 2005, pp. 119–120). Sources: Gold export point estimates: England/France/Germany: Morgenstern, Oskar. International Financial Transactions and Business Cycles. Princeton: Princeton University Press, 1959, pp. 178–81; England/The Netherlands and England/Austria-Hungary: Easton, H. T. Tate's Modern Cambist. London: Effingham Wilson, 1912, pp. 358–63. For exchange-rate data cf. appendix.

M. Morys / Explorations in Economic History 50 (2013) 205–226

Exchange rate w.r.t.

The Netherlands

Norway

Austria-Hungary

Italy

Romania

Bulgaria

Serbia

France

Germany

England

France

German.

England

France

German.

England

France

German.

England

France

German.

France

1.0116 0.0025 1.0125 1.0103 0.0024

1.0125 0.0034

1.0069 0.0014

1.0110 0.0035 1.0136 1.0110 0.0032

1.0136 0.0033

1.0085 0.0028

1.0277 0.0071 1.0283 1.0281 0.0070

1.0283 0.0064

1.0283 0.0074

1.0527 0.0066 1.0527 1.0509 0.0066

1.0512 0.0069

1.0490 0.0063

1.0115 0.0042 1.0128 1.0115 0.0040

1.0128 0.0036

1.0102 0.0044

1.0390 0.0094

n.a.

n.a.

n.a.

7.0

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

5.1 0.5

11.4 0.8

0.3 0.0

8.4 1.4

8.4 3.7

2.8 0.0

31.7 12.5

28.3. 14.2

21.7 10.8

28.4 14.5

33.4 16.3

30.0 13.1

12.3 2.5

12.3 2.5

4.9 1.2

40.0 23.8

n.a.

n.a.

n.a.

0.0

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

n.a.

0.9 0.0

3.7 0.0

0.0 0.0

0.0 0.0

0.7 0.0

0.0 0.0

30.7 5.3

24.0 8.0

14.7 0.0

17.7 5.0

23.1 6.8

22.2 7.2

17.7 3.9

13.7 3.9

7.8 2.0

40.0 14.0

1873m1 1914m6

1881m1 1914m6

1877m1 1914m6

1896m3 1914m6

1896m3 1914m6

1896m3 1914m6

1904m1 1913m12

1904m1 1913m12

1904m1 1913m12

1882m7 1913m12

1882m7 1913m12

1882m7 1913m12

1906m1 1912m9

1906m1 1912m9

1906m1 1912m9

1905m7 1912m9

M. Morys / Explorations in Economic History 50 (2013) 205–226

England

209

210

M. Morys / Explorations in Economic History 50 (2013) 205–226

This leads to four sets of equations, two of which we shall discuss first, namely the two country-specific equations. We propose the following equation for the actual discount rate changes: α0  constant 

Δi ¼ þ β1 xrem¼−1   þ γ1  ied¼−1 –ied¼−31

þ α1  crm¼−1 

þ β2 xrf m¼−1   þ γ2  if d¼−1 –if d¼−31

þ α2  ðcrm¼−1 –crm¼−2 Þ

þ

þ β3  xrgm¼−1   þ γ3  igd¼−1 –igd¼−31

þ

ð1Þ

þ ε:

The dependent variable Δi captures the interest rate change (in percent) on a given day of base rate change (which we arbitrarily set at day d = 0 in month m = 0). α-coefficients then relate to the cover ratio: (crm=−1 − crm =−2) measures its change, comparing the month prior to the discount rate change (m= −1) to the month preceding it (m= −2). crm = −1 measures the level of the cover ratio (in m = −1) and is included in addition to the change in the cover ratio for the following reason: a decline in the cover ratio is more likely to result in a discount rate increase, if the central bank is faced with already low reserve levels (Davutyan and Parke, 1995). Including a constant is a consequence: while the variables for the discount rate change, the exchange-rate deviation and the interest rate changes (as well as the dependent variable itself) all have an expected value of zero, crm=−1 is always positive; which means that we require the constant as an offset. Assuming that the central bank targets a specific reserve level, we would expect this level to be close to −(α1 / α0). We will return to this issue later. β-coefficients relate to the exchange-rates vis-à-vis England, France and Germany. They measure deviations from mint parity, with a positive value implying a depreciated exchange-rate; for instance, a value of 0.01 for xrEngland means that the exchange-rate vis-à-vis England was depreciated by 1% with respect to mint parity. Finally, γ-coefficients relate to potential patterns of interest rate followership vis-à-vis England, France and Germany. While data on cover ratio and exchange-rate are available only on a monthly frequency, we exploit the daily character of our discount rate data by measuring the (potential) discount rate change of England, France and Germany as the difference between the discount rate level the day before the discount rate decision of the country in question (d= −1) and 30 days earlier (d= −31). What signs do we expect? A declining cover ratio should lead to a discount rate increase. Similarly, a discount rate increase is more likely at low reserve levels. Hence, α1 and α2 are expected to be negative; α0 as an offset should then be positive. β-coefficients are expected to be positive, as depreciation puts upward pressure on the bank rate; similarly, the concept of interest rate followership implies that γ-coefficients should be positive. Modelling monthly discount rate behaviour entails two changes: firstly, we estimate the equation as an ordered probit with three entries: comparing the last day of the current month (m= 0) with the last day of the previous month (m= −1), the discount rate was either increased (Δi = +1), decreased (Δi = −1) or remained at the same level (Δi =0). Secondly, adaptation is required for the interest rate. As the dependent variable looks at a one-month horizon (as opposed to a specific day in the first set of equations), i_e, i_f and i_g refer to the potential interest rate changes between the last day of the previous month (m= −1) and the last day of the month before (m= −2). Sign expectations remain unchanged but refer to marginal effects (or, in the case of interest rate changes, differential effects). 

xrem¼−1 Δi ¼ þε1   ζ1  iem¼−1 –iem¼−2

δ1  crm¼−1 þε2  xrf m¼−1   þζ2  if m¼−1 –if m¼−2

þδ2 ðcrm−1 –crm−2 Þ þε3  xrgm¼−1   þζ3  igm¼−1 –igm¼−2

þ þ þε:

ð2Þ

How do Eqs. (1) and (2) (results in Tables 4 and 5) compare to central bank reaction functions estimated in earlier research? For all their minor differences, most estimations include the same variables as we do, with the possible exception of the exchange-rate deviation, which is only captured by Tullio and Wolters (2003a,b,c, 2007). The main difference lies in the high frequency of our data as well as the large sample. Most earlier studies do not venture beyond England, France and Germany (Contamin and Denise, 1999) and most tend to involve only one (Goodhart, 1972; Dutton, 1984; Pippenger, 1984; Davutyan and Parke, 1995) or two (Giovannini, 1986) of them; there is really only one comparable study dedicated to a peripheral country, namely Austria-Hungary (Tullio and Wolters, 2007). The pooled estimations follow the same principle with one modification: the exchange-rate data on England, France and Germany was highly multi-collinear so that it seemed prudent to replace the individual series by a principal component series. We first estimate a separate pool for core countries and for peripheral countries (estimations (3A) and (3B) for actual discount rate changes and estimations (4A) and (4B) for monthly discount rate behaviour). In another estimation (estimations (3C) and (4C)), we then merge both pools and use group-specific regressors for those variables which delivered substantially different results in estimations (3A) vs. (3B) and estimations (4A) vs. (4B). As for Eqs. (1) and (2), we follow a general-to-specific approach by eliminating successively all variables that are not statistically significant at the 10 percent level. This often implied the exclusion of multi-collinear regressors. Low variance inflation factors for the final results indicate that multi-collinearity no longer poses a problem. As for estimations (3) and (4) (shown in Tables 6 and 7), we report results for different weights (pooled least squares versus pooled EGLS) as well as for different coefficient covariances (standard, cross-section SUR and period SUR). All equations pass conventional residual tests for white noise, normal distribution and heteroskedasticity as well as tests against misspecification.

M. Morys / Explorations in Economic History 50 (2013) 205–226

211

It was not feasible to include Bulgaria and Serbia in the country-specific regressions due to the degree of freedom constraints. As they adhered to the gold standard only for short periods of time (Bulgaria: 1906–1912; Serbia: 1905–1912), their discount rates were changed only two and four times, respectively. Both countries are, however, included in the pooled regressions. Last but not least, we performed several tests to establish whether central bank reaction functions were asymmetric or not, that is, whether individual countries behaved differently under exchange rate pressure or continued to follow their general discount rate setting pattern also under adverse conditions. Two approaches are followed: first, we estimate separately the three periods which we identify as periods of wide-spread gold point violations (cf. Table 1 and the more detailed discussion in Section 3): October 1899–July 1901 (Boer War), October 1906–February 1909 (crisis of 1907) and September 1912–December 1913 (Balkan Wars). Second, we estimate regressions 3A, B, and C and 4A, B, and C as threshold regressions, differentiating between “crisis periods” – i.e., situations in which the exchange rate to England, France or Germany deviated sufficiently from mint parity to raise concerns over gold exports – and “normal periods”, that is, all other periods. Parameter stability suggests that the discount rate policy of individual countries did not differ between crisis periods and normal periods; for details see the appendix.

2.3. Summary statistics on the independent and the dependent variables Before presenting the estimation results (Tables 4–7), it seems worthwhile exploring the raw data on the exchange-rate performance, the discount rate behaviour and the cover ratio. Some summary statistics of these three indicators alone are suggestive of substantial differences in monetary policy between core and peripheral countries.

2.3.1. Exchange-rate performance Available gold point estimates 7 mainly cover intra-core country pairs and range between 0.367% (Germany to England) and 0.645% (Austria-Hungary to England).8 For core–periphery pairs, gold points are expected to be further away from mint parity, as transportation costs for gold shipping between, say, Bucharest and Berlin were higher than between Paris and Berlin. Fortunately, we can avoid calculating gold points, as peripheral countries did normally not introduce specie convertibility; this is well documented for Austria-Hungary and Italy, but recent research on Bulgaria, Romania, Serbia and Sweden suggests the same for other peripheral economies (Morys, 2006; Fratianni and Spinelli, 2001; Tattara, 2000; Monetary Time Series of South-Eastern Europe).9 Stabilising the exchange-rate with respect to England, France and Germany is often referred to as shadowing the gold standard and is conventionally seen as another form of gold standard adherence. 10 Crucial in this context is that if the peripheral countries wanted to stabilize their exchange-rate, they had to set themselves “virtual” gold points which would trigger some kind of central bank reaction. We have therefore assumed two such points at 0.6% and 1.0%. Table 1 provides the relevant statistics, based on exchange-rate data with respect to England, France and Germany. For each of the three exchange-rates, the maximum deviation and the standard deviation are provided. The information is then “condensed” into three indicators: the maximum deviation with respect to all three countries, the average maximum deviation and the average standard deviation. The lower part of Table 1 shows how often (in percent) the gold export point was violated with respect to England, France, and Germany, as well as how often the exchange-rate was depreciated by more than 0.6% and 1.0%, respectively. An additional statistic is provided in which gold export point violations are analysed without taking into account periods of global financial strain before World War I. These periods were identified as the onset of the Boer War (1899–1902), the crisis of 1907, and the Balkan Wars (1912/13). These crises – and their importance in understanding the core-periphery dichotomy – will be addressed in Section 3. England, Germany, France and the Netherlands are the only countries whose exchange-rate never depreciated by more than 1%. Belgium was slightly below this at 1.19%. At the other end of the spectrum, Italy and Romania stand out with maximum deviations of 2.83% and 5.27%, respectively. Sweden, Norway, and Austria-Hungary occupy the middle ground (in that order). The average standard deviation provides a similar picture, even though Sweden, Norway and Austria-Hungary (again in that order) are more similar to the core countries according to this measure. The five core countries and Sweden, Norway and Austria-Hungary all have average standard deviations between 0.22% and 0.36%. This contrasts strongly with Italy and Romania with average standard deviations of approximately double this amount. Focusing on the middle and lower parts of Table 1, we first consider the export point violations of those countries for which we have gold export point estimates, namely England, Germany, France, the Netherlands and Austria-Hungary. Such violations were rare but did occur. The only country without any such violation in our sample is the Netherlands. However, England and Germany, for instance, had more than the occasional gold point violation. In their bilateral relationship, they had, on average, one violation every year. This might not sound much; however, the data relies on monthly averages; consequently, the number of gold point violations reported in Table 1 only constitutes a lower bound of the actual number of violations if data of higher frequency were used. A similar picture emerges for the German exchange-rate vis-à-vis France. 7 It should be emphasized that gold points were not time-invariant; they grew smaller over time due to declining transportation costs and they could also change during financial crises, as shown by Canjels et al. (2004). For the argument advanced here, however, it seems justified to operate with an average gold point taken from contemporary calculations. 8 Cf. sources of Table 1. 9 We thank Anders Ogren for providing similar insight for the Swedish case. 10 Inter alia Obstfeld et al. (2005), who do not even differentiate between de jure and de facto adherence to gold.

212

M. Morys / Explorations in Economic History 50 (2013) 205–226

Austria-Hungary is the only peripheral country for which contemporary gold point estimates exist, and we find that gold export point violations were more frequent (7.0%) than for England, France, Germany, and the Netherlands (2.7% on average). We now turn to Sweden, Norway, Italy and Romania for which we do not have gold point estimates. When applying the 0.6 percent benchmark (a criterion less stringent than any of the gold point estimates that exist bar the one for Austria-Hungary), only Sweden's exchange-rate performance comes close to England, France, Germany and the Netherlands (2.8%). Next comes Norway (5.6%) and Austria-Hungary (6.5%), to be followed – with a wider margin – by Italy (27.2%) and Romania (30.6%). If a 1 percent depreciation is allowed, fewer gold point violations occur but the broad pattern remains unchanged: Sweden, Norway, and Austria-Hungary had substantially fewer gold point violations than Italy and Romania. A closer observation of the exchange-rate data shows that most gold point violations happened during three well-defined periods: the early phase of the Boer War (1899–1902), the crisis of 1907, and the Balkan Wars (1912/13). For the purposes of this section, we merely point out how our statistics change when the three episodes are left out of consideration (more on the implications for discount rate policy in Section 3). The exchange-rate performance of both core and periphery improves but improvements are more pronounced for the periphery (lower part of Table 1). Sweden, Norway and Austria-Hungary have no (on the 1 percent benchmark) or virtually no (on the 0.6 percent benchmark) gold point violations, and Italy's and Romania's exchange-rate performance also appears much improved; Italy, for instance, does not exhibit a single gold point violation vis-à-vis Germany using the 1.0 percent benchmark. The findings on the exchange-rate performance can be summarised as follows: firstly, peripheral countries exhibited higher exchange-rate volatility, which, in turn, led to more frequent and more sizeable violations of the gold points. Secondly, differences in the ability to keep the exchange-rate close to mint parity were particularly acute in times of global financial strain. 2.3.2. Discount rate behaviour Table 2 provides data on the average discount rate and the frequency of discount rate changes. As Austria-Hungary and Italy only joined the gold standard in 1896 and 1904, respectively, we distinguish the average discount rate between three periods: 1883–1913, 1896–1913, and 1904–1913. For some countries, we were able to collect data for the private discount rate. This relates to the interest rate applied to bills of exchange discounted on the private market, which usually meant discounting at banks specialised in discounting bills (this was especially true in the English case which we understand best thanks to Bagehot). We failed to locate such data for Bulgaria, Italy, Norway, Romania, Serbia and Sweden, which suggests that the money market in these countries was less mature and most, if not all, of the discounting took place at the central bank. France enjoyed both the lowest bank rate and the lowest private discount rate. In terms of bank rate, the Netherlands rather than England comes in second place. However, England does take second place in the private discount rate. The spread among core countries was low, but it is interesting to note that Germany had the highest discount rate of the five core countries. Peripheral countries had substantially higher interest rates. Sweden, Norway, and Romania, all of which adhered to gold for the entire period, exhibit a discount rate spread of more than 150 basis points (4.97% versus 3.41%). If Austria-Hungary, Italy, Bulgaria and Serbia are included, all of which adhered to gold only for parts of the 1883–1913 period, the interest rate spread increases to more than 200 basis points (5.48% versus 3.41%). Turning to the frequency of discount rate changes, Table 2 shows that core countries had a more active discount rate policy. Core countries changed their discount rate 2.9 times per annum, compared to 1.6 times per annum for the periphery. We will be able to Table 2 Discount rate statistics. Discount rate 1883–1913

Core countries England France Germany The Netherlands Belgium Average Peripheral countries Sweden Norway Austria-Hungary Italy Romania Bulgaria Serbia Average periphery on gold Average periphery

Discount rate 1896–1913

Discount rate 1904–1913

Discount rate changes

Official

Private

Official

Private

Official

Private

Total

Per annum

3.36 2.92 4.15 3.21 3.40 3.41

2.71 2.46 3.17 2.82 2.83 2.80

3.58 2.98 4.50 3.53 3.66 3.65

3.09 2.58 3.60 3.15 3.04 3.09

3.73 3.20 4.75 3.76 3.93 3.88

3.27 2.65 3.78 3.29 3.37 3.27

194 19 115 67 98

5.7 0.8 3.4 2.0 2.7 2.9

62 60 25 43 26 2 4

1.7 1.8 1.4 4.3 0.8 0.3 0.6 1.6

4.69 4.75 4.25 4.75 5.47 7.61 6.87 4.97 5.48

3.85

5.02 5.07 4.30 4.39 5.55 7.35 6.71 4.99 5.48

3.93

5.07 5.01 4.44 4.46 5.21 6.95 6.08 4.84 5.32

4.08

Note: The number of discount rate changes relates to the estimation period in Section 2 or, in the case of Bulgaria and Serbia, to the period of gold standard adherence (1/1906–9/1912 and 7/1905–9/1912, respectively). Entries in italics refer to countries that were not on gold during the entire period. Sources: Cf. data appendix.

M. Morys / Explorations in Economic History 50 (2013) 205–226

213

CENTRAL BANK BALANCE SHEET Assets

Liabilities

International assets Gold (bullion and specie) Silver (bullion and specie) Foreign exchange and other international assets

Liquid liabilities Bank notes in circulation Bank deposits Other liabilities payable on demand

Domestic assets Bills of exchange Cash advances Other assets (real estate etc.)

Other liabilities

Fig. 1. Central bank balance sheet.

give a full explanation only in Section 3, but two factors should be mentioned at this stage: peripheral countries were, on balance, under more government pressure, and hence more reluctant to change the discount rate (particularly in cases where formal government approval was required). This argument, which is often found in the literature (Reis, 2007, pp. 712–22), probably goes some way towards explaining the differences but fairly independent central banks could also come under intense pressure. According to Bagehot (1878, p. 72), “[t]he Bank directors now fear public opinion exceedingly; probably no kind of persons are so sensitive to newspaper criticism.” Presumably more important was how dominant the central bank's position was in the money market. If faced with heavy competition from discount houses, the central bank had to adjust its bank rate frequently to the market rate so as to obtain its share of the discount market. If the central bank dominated the money market, which was presumably the case in the periphery, fewer discount rate changes were needed.

2.3.3. Cover ratio The cover ratio represents the fraction of reserves to liquid liabilities and served as an indicator of the central bank's liquidity. The legal requirement to publish the cover ratio on a monthly or even weekly basis was aimed at retaining confidence in the domestic monetary system. A standardized central bank balance sheet illustrates which items might potentially constitute “reserves” and “liquid liabilities” (Fig. 1). Rules as to what exactly defined the cover ratio differed across countries and over time, but the broad pattern is as follows: initially, the reserve ratio was defined as gold (bullion or specie) over bank notes. 11 Gold was the quintessential store of value, and the convertibility requirement referred only to bank notes but not deposits. As time went on, the cover ratio matured from this somewhat legalistic perspective into a more economic concept. This was largely the result of foreign exchange holdings becoming more important relative to gold holdings. Consequently, countries amended their bank acts in order to include foreign exchange in the note cover (Morys, 2006, p. 136). It appears that a similar process did not take place on the liability side. While deposits grew as a share of liquid liabilities, bank acts were not changed to provide cover for both bank notes as well as deposits. This is probably explained by the fact that the share of deposits never grew larger than 15 to 20% of the liquid liabilities. Table 3 shows the average, the minimum, the maximum and the standard deviation of the cover ratio. Core countries had a substantially lower average reserve ratio than the periphery (56.0% versus 69.6%). The minimum reserve ratio shows a similar dichotomy. England, Germany, and Belgium, for instance, let their reserve ratio decline to 30% and below, which contrasts strongly with a minimum cover ratio of 66.1% for Italy. So while neither core nor periphery violated their bank acts, which normally stipulated a 30%–40% minimum cover (Reichsbank, 1925), the findings suggest that peripheral countries felt the need to have substantially higher reserve levels. To summarise: core countries had few gold point violations (which allowed them to introduce specie convertibility), changed their discount rate frequently and had a low cover ratio. Peripheral countries, in contrast, had a worse exchange-rate performance (especially in periods of global financial strain), changed their discount rate infrequently and relied on a high cover ratio. The question arises of whether these pronounced differences in independent and dependent variables translate into different central bank reaction functions for core and peripheral countries. This is what we turn to now.

11 The cover ratio of the Bank of England was defined differently as a result of the 1844 Bank Act. The Issue Department only issued notes. There was a limited fiduciary note issue, but any additional note issue had to be fully backed by gold. The cover ratio – often referred to as the “proportion” in the English case – refers to the Banking Department and represents its reserves (largely bank notes) to its liquid liabilities.

214

M. Morys / Explorations in Economic History 50 (2013) 205–226

Table 3 Cover ratios. Cover ratio

Core countries England France Germany The Netherlands Belgium Average core Peripheral countries Sweden Norway Austria-Hungary Italy Romania Bulgaria Serbia Average periphery

Average (%)

Minimum (%)

Maximum (%)

St. deviation

46.7 74.1 56.5 62.9 39.7 56.0

28.0 58.4 30.1 45.8 29.9 38.4

71.2 86.1 75.4 80.4 49.5 72.5

0.073 0.051 0.097 0.073 0.036 0.066

74.6 69.4 79.2 75.0 51.8 68.8 68.4 69.6

50.0 51.4 54.5 66.1 38.3 45.0 50.0 50.8

137.7 87.1 98.0 84.3 76.7 99.7 88.4 96.0

0.145 0.072 0.098 0.041 0.077 0.149 0.093 0.096

Note: Calculations relate to the estimation period in Section 2 or, in the case of Bulgaria and Serbia, to the period of gold standard adherence (1/1906–9/1912 and 7/1905–9/1912, respectively). Sources: Cf. data appendix.

2.4. Results of country-specific estimations (Eqs. (1) and (2); Tables 4 and 5) 2.4.1. Cover ratio Results for Eq. (1) (actual discount rate changes) show that England, Germany, the Netherlands and Belgium did not respond to changes in the cover ratio by adjusting their discount rate. In contrast, for all peripheral countries (with the exception of Sweden), we

Table 4 Determinants of actual discount rate changes (estimation 1). Sources: Own calculations based on sources as described in the appendix. England α0 Intercept α1 Cover ratio: level in m−1 α2 Cover ratio: change m−2 to m−1 β1 β2 β3

γ1 γ2 γ3

R2 N

xr deviation in m−1 vis-à-vis England xr deviation in m−1 vis-à-vis France xr deviation in m−1 vis-à-vis Germany English bank rate: change d−31 to d−1 French bank rate: change d−31 to d−1 German bank rate: change d−31 to d−1

France

Germany The Belgium Netherlands

1.82⁎⁎⁎ 2.54⁎⁎⁎ −2.98⁎⁎⁎ −3.52⁎⁎⁎

Sweden

−0.20⁎⁎

−7.85⁎⁎

Austria-Hungary Italy

69.70⁎⁎⁎

57.17⁎⁎⁎

Romania

3.19⁎ −4.28⁎

1.00 −1.74⁎ −10.84⁎⁎⁎ −11.89⁎⁎⁎

70.15⁎⁎⁎

−4.30⁎⁎

−11.27⁎⁎⁎

42.8⁎⁎ 77.35⁎

44.79⁎⁎⁎

65.78⁎⁎⁎

0.31⁎⁎⁎

88.78⁎⁎⁎

0.56⁎⁎⁎

0.27⁎⁎⁎ 0.71⁎⁎⁎

0.54⁎⁎⁎

0.34 194

Norway

0.30⁎⁎⁎

0.77 16

0.42 115

0.57 65

0.20 98

0.63⁎⁎⁎ 0.29⁎

0.24⁎⁎⁎

0.37⁎⁎⁎

0.48 62

0.47 60

Notes: We report Newey–West heteroskedasticity and autocorrelation consistent standard errors. ⁎ Significant at the 10 percent level. ⁎⁎ Significant at the 5 percent level. ⁎⁎⁎ Significant at the 1 percent level.

0.33⁎⁎⁎

0.42 25

0.49 43

0.54 26

M. Morys / Explorations in Economic History 50 (2013) 205–226

215

Table 5 Determinants of monthly discount rate behaviour, marginal effectsa (estimation 2). Sources: Own calculations based on sources as described in the appendix. England δ1 Cover ratio: level in m−1 −0.46⁎⁎ δ2 Cover ratio: change m−2 −0.60⁎ to m−1 ε1 xr deviation in m−1 vis-à-vis England ε2 xr deviation in m−1 vis-à-vis France ε3 xr deviation in m−1 vis-à-vis Germany ζ1 English bank rate: change m−2 to m−1 ζ2 French bank rate: change m−2 to m−1 ζ3 German bank rate: change m−2 to m−1 Pseudo-R2 DWb N

France

Germany

−0.56⁎⁎⁎ −0.34⁎⁎ 8.55⁎

The Belgium Sweden Netherlands

Austria-Hungary Italy

−1.88⁎⁎⁎ −1.99⁎⁎⁎

Romania

−3.42⁎⁎⁎ −0.45⁎⁎⁎

9.84⁎⁎⁎

14.37⁎⁎

9.90⁎

12.21⁎⁎

9.40⁎⁎⁎

0.01⁎⁎

0.09⁎⁎⁎

0.12⁎⁎⁎

0.05 1.92 406

−0.11⁎⁎

−0.52⁎

Norway

0.10 2.44 298

0.04 1.96 406

0.04⁎

0.06⁎⁎⁎

0.07⁎⁎⁎

0.05⁎⁎

0.04

0.08⁎⁎⁎

0.10⁎⁎⁎

0.12⁎⁎⁎

0.22⁎⁎

0.16 1.98 398

0.03 2.18 440

0.09 2.15 430

0.12 2.18 394

0.13 1.91 214

0.12 1.98 120

0.11 2.18 375

Notes: We report Huber–White quasi-maximum likelihood (QML) standard errors. a Differential effects (to +1%) for the bank rate as a discontinuous variable. b Values taken from the corresponding OLS regression. ⁎ Significant at the 10 percent level. ⁎⁎ Significant at the 5 percent level. ⁎⁎⁎ Significant at the 1 percent level.

find the expected negative sign. For some peripheral countries (Italy, Norway and Sweden), we can detect an additional role for the level of the cover ratio.12 Our findings for Eq. (2) (monthly discount rate behaviour) are supportive. Given that more observations are now available for each equation, we do find a role for the cover ratio in all equations (except for Belgium). Crucially, however, marginal effects for core countries are much smaller than for peripheral countries and are in some cases only marginally significant. Germany, for instance, has a marginal effect that is ten times smaller than Italy (−0.34 versus −3.42). 2.4.2. Exchange-rates β- and ε-coefficients on the exchange-rate reveal another core-periphery divide. Counteracting exchange-rate deviations by means of discount rate changes was crucial for England, Germany, the Netherlands, and Belgium, but not for peripheral countries. In Eq. (2), which should allow more precise estimation given the higher number of observations, the exchange-rate does not come out as significant for any of the five peripheral countries. Only in Eq. (1) do we find a role for the exchange-rate in the cases of Sweden and Norway. This intra-periphery difference is in line with our findings for the exchange-rate performance, where the two Scandinavian countries performed better than Austria-Hungary, Italy and Romania. 2.4.3. Interest rates Eqs. (1) and (2) demonstrate that all countries track the interest rate of at least one of the large core countries. While differences prevail between core and periphery regarding cover ratio and exchange-rate, similarities are strong as far as the interest rate is concerned. Most countries follow either England or Germany and coefficients/differential effects are of comparable magnitude. While interest rate followership can clearly be detected, the size of the coefficients/differential effects is probably not as large as would be expected given the macroeconomic policy trilemma. In Eq. (1), none of the coefficients comes even close to unity (which would mean perfect interest rate pass through). Similarly, differential effects in Eq. (2) are low: for example, when the Bank of England increases its discount rate by 1%, the probability of the Reichsbank increasing its discount rate increases only by 9%; even the highest value in this context (a 22%-increase in probability for Italy in response to Germany) is a long way from an “automatic” interest rate response. Thus, γ- and ζ-coefficients suggest that a considerable amount of monetary autonomy was retained under the Classical Gold Standard, even for peripheral countries. 12 In these cases, −(α0 / α1) ≈ average cover ratio, vindicating our earlier interpretation of the intercept as an offset to the level of the cover ratio; e.g., Italy: −(α0 /α1) = 74.53% ≈ 75.0% (cf. Table 3).

216

M. Morys / Explorations in Economic History 50 (2013) 205–226

Table 6 Determinants of actual discount rate changes (pooled). Method (weights)

Pooled least squares (no weights)

Coefficient covariances Estimation 3A: core countries pooled (country fixed-effects, N = 482) Intercept α0 Cover ratio: level in m−1 α1 Cover ratio: change m−2 to m−1 α2 β

xr dev. in m−1 vis-à-vis England, France and Germany (pca-series)

γ1 γ2 γ3

English bank rate: change d−31 to d−1 French bank rate: change d−31 to d−1 German bank rate: change d−31 to d−1

R2

Std.

⁎⁎⁎

⁎⁎⁎

58.14

⁎⁎⁎

0.34 0.10 0.27

Period SUR

β

xr dev. in m−1 vis-à-vis England, France and Germany (pca-series)

−0.33

γ1 γ2 γ3

English bank rate: change d−31 to d−1 French bank rate: change d−31 to d−1 German bank rate: change d−31 to d−1

2

R 1 p-Value: 15.0%; 2p-value: 12.3%; 3p-value: 13.1%. Estimation 3C: all countries pooled (country fixed-effects, N = 696) Intercept α0 α1 Cover ratio: level in m−1 Group-specific regressor core Group-specific regressor periphery Cover ratio: change m−2 to m−1 α2 xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) Group-specific regressor core Group-specific regressor periphery English bank rate: change d−31 to d−1 French bank rate: change d−31 to d−1 German bank rate: change d−31 to d−1

R2

0.09 0.07 0.37

***

***

⁎⁎⁎

***

59.21

***

⁎⁎⁎

⁎⁎⁎

***

***

⁎⁎⁎

⁎⁎⁎

*

0.34 0.09 0.20

***

0.38

⁎ ⁎ ⁎⁎

⁎ ⁎ ⁎⁎

*** ***

0.58 −0.86 −1.88

** ** *

3.36 1

2

⁎⁎⁎

⁎⁎⁎

***

0.36

0.09 0.00 0.37

3

***

0.39

2

0.23 −0.19 −0.74 −2.51

Std.

0.12 −0.22 −2.58

0.34

0.58 −0.86 −2.58

γ1 γ2 γ3

Cross-section SUR

0.11 −0.20 −2.40

Estimation 3B: peripheral countries pooled (country fixed-effects, N = 214) Intercept α0 Cover ratio: level in m−1 α1 α2 Cover ratio: change m−2 to m−1

β

Pooled EGLS (cross-section weights)

1

0.25 −0.24 −0.76 −2.34

3

⁎⁎⁎

⁎ ⁎⁎⁎

** ***

58.64 1.88

⁎⁎⁎

⁎⁎⁎

***

60.03 5.12

***

0.26 0.08 0.27

⁎⁎⁎

⁎⁎⁎

***

***

⁎⁎⁎

⁎⁎⁎

**

0.25 0.04 0.24

0.33

* ***

***

0.37

1

p-Value: 14.9%; 2p-value: 19.5%; 3p-value: 17.7%. ⁎ Significant at the 10 percent level. ⁎⁎ Significant at the 5 percent level. ⁎⁎⁎ Significant at the 1 percent level.

Taken together, Eqs. (1) and (2) show the similarities and differences between core and periphery. An element of interest rate followership was common to both, though not as strong as would be expected under fixed exchange-rates and capital mobility. Core and periphery differ, however, in the importance they attach to cover ratio and exchange-rate deviations. Core countries use their discount rate changes to target specifically the exchange-rates and they do so with great success, as evidenced by the superior exchange-rate performance compared with peripheral countries. In turn, they pay little attention to the cover ratio, which was allowed to fall close to the minimum level required by the bank act. Peripheral countries pursue the opposite strategy: paying little attention to exchange-rate deviations (hence their poor performance on this account), they target the cover ratio and successfully manage to keep it at levels approximately twice as high as required by the bank act. The only country which escapes this clear-cut core–periphery dichotomy seems to be France. Neither Eq. (1) nor Eq. (2) suggests that discount rate policy was motivated by exchange-rate deviations. Similarly, the high cover ratio and low frequency of discount rate changes set it apart somewhat from England, Germany, Belgium and the Netherlands. Yet its exchange-rate performance was stellar. This seeming contradiction is probably best reconciled by a well-established body of research which has

M. Morys / Explorations in Economic History 50 (2013) 205–226

217

Table 7 Determinants of monthly discount rate behaviour (pooled). Method (weights)

Pooled least squares (no weights)

Coefficient covariances Estimation 4A: core countries pooled (country fixed-effects, N = 1941) Intercept δ0 Cover ratio: level in m−1 δ1 Cover ratio: change m−2 to m−1 δ2 ε

xr dev. in m−1 vis-à-vis England, France and Germany (pca-series)

ζ1 ζ2 ζ3

English bank rate: change m−2 to m−1 French bank rate: change m−2 to m−1 German bank rate: change m−2 to m−1

R2 DW 1 p-Value: 11.6%; 2p-value: 18.9%; 3p-value: 19.9%.

Std.

⁎⁎⁎

⁎⁎⁎

17.86

⁎⁎⁎

0.06 −0.05 0.04

ε

xr dev. in m−1 vis-à-vis England, France and Germany (pca-series)

−0.52

ζ1 ζ2 ζ3

English bank rate: change m−2 to m−1 French bank rate: change m−2 to m−1 German bank rate: change m−2 to m−1

ε

ζ1 ζ2 ζ3

Period SUR

0.04 0.01 0.07

Std.

⁎⁎⁎

−0.04 0.08 −1.06

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

9.64

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎

2

3

0.05 −0.03 0.00

⁎⁎⁎

1

0.04 1.91

0.09 −0.12 −0.71

Estimation 4C: all countries pooled (country fixed-effects, N = 3459) Intercept δ0 Cover ratio: level in m−1 δ1 Group-specific regressor core Group-specific regressor periphery Cover ratio: change m−2 to m−1 δ2

Cross-section SUR

−0.05 0.09 −0.99

Estimation 4B: peripheral countries pooled (country fixed-effects, N = 1518) Intercept δ0 δ1 Cover ratio: level in m−1 Cover ratio: change m−2 to m−1 δ2

R2 DW

Pooled EGLS (cross-section weights)

0.04 1.99

⁎ ⁎ ⁎⁎⁎

⁎⁎ ⁎⁎ ⁎⁎⁎

⁎⁎ ⁎ ⁎

0.09 −0.12 −0.60 −0.46

⁎⁎⁎

⁎⁎⁎

⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎

0.03 0.01 0.07

0.05 1.92

0.05 1.95

0.02

0.02

0.07 −0.11 −0.90

⁎⁎ ⁎⁎ ⁎⁎⁎

1

2

⁎⁎⁎

⁎ ⁎⁎⁎

⁎⁎⁎

0.04 −0.11 −0.79

⁎⁎⁎ ⁎⁎⁎

⁎ ⁎⁎⁎

xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) Group-specific regressor core Group-specific regressor periphery

17.92 −0.48

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

9.36 −0.62

⁎⁎⁎

English bank rate: change m−2 to m−1 French bank rate: change m−2 to m−1 German bank rate: change m−2 to m−1

0.05 −0.02 0.05

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎

⁎⁎⁎

0.04 0.00 0.04

2

R DW

0.04 1.92

⁎⁎⁎

0.04 1.97

1

p-Value: 22.3%; 2p-value: 21.7%. ⁎ Significant at the 10 percent level. ⁎⁎ Significant at the 5 percent level. ⁎⁎⁎ Significant at the 1 percent level.

stressed the sophistication and effectiveness of French gold devices, thereby substantially reducing reliance on the discount rate tool (inter alia Contamin, 2003).

2.5. Results of pooled estimations (estimations (3A), (3B), and (3C) and estimations (4A), (4B), and (4C); Tables 6 and 7) The results obtained from country-specific regressions are confirmed by pooled estimations. We first discuss the pooled estimates for core countries vs. peripheral countries (3A vs. 3B and 4A vs. 4B) and then proceed to a common pool with group-specific regressors.

218

M. Morys / Explorations in Economic History 50 (2013) 205–226

2.5.1. Cover ratio Changes in the cover ratio are important for the core and periphery but peripheral countries react more strongly in estimation (3). Moreover, the level of reserves is significant for peripheral countries but not for the core (the same is true for the constant, see our discussion above on the constant as an offset for the level of reserves). Both findings are in line with Eqs. (1) and (2) and underline the specific importance attached to the cover ratio, both in levels and first differences, by peripheral countries. 2.5.2. Exchange-rates Core countries target the exchange-rate but peripheral countries do not, confirming earlier results. 2.5.3. Interest rates Estimations (3A), (3B), (4A), and (4B) all show patterns of interest rate followership vis-à-vis England and Germany but there is some difference between core and periphery. Core countries follow the English interest rate more closely, as evidenced by a higher coefficient.13 By contrast, peripheral countries follow Germany more closely, with some specifications not even indicating any role for the London bank rate. It is worth exploring this difference in more detail by referring back to the country-specific equations. Leaving England and Germany aside for obvious reasons, Eqs. (1) and (2) suggest the same interest rate setter in seven out of eight cases. Sweden follows both England and Germany, with the size of the coefficient/differential effect suggesting similar dependence on London and Berlin. The remaining six countries either follow the Bank of England or the Reichsbank, with France and Romania following England and the Netherlands, Norway, Austria-Hungary and Italy following Germany. Only Belgium seems to follow the German bank rate in Eq. (1) and the French bank rate in Eq. (2). Combining our findings of Eqs. (1)–(2) and estimations (3)–(4), we find that the Reichsbank had a much bigger role in setting interest rates under the Classical Gold Standard, at least vis-à-vis countries on the European periphery, than is conventionally acknowledged (Eichengreen, 1987; Tullio and Walters, 1996). 2.6. Pooling core and peripheral countries with group-specific regressors (estimations (3C) and (4C)) Estimations (3A), (3B), (4A) and (4B) have shown that the main difference lies in the importance attached to the reserve level and the exchange-rate. Core and peripheral countries were therefore pooled and group-specific regressors were formed for the level of the cover ratio and the exchange-rate. The fundamental difference is confirmed in estimations (3C) and (4C). While all other coefficients can easily be pooled for all countries (which then take on an intermediate value compared to the separate equations reported before), the coefficient for the exchange-rate is only significant for core countries, while the reserve level is significant only for peripheral countries.14 In conclusion, we found similarities and differences between core and periphery. A strong element of interest rate followership was common to both. They differ, however, in the importance they attach to cover ratio and exchange-rate deviations. Core countries used their frequent discount rate changes to target the exchange-rate. Little attention was paid to the cover ratio which occasionally fell close to the minimum level required by the bank act. Peripheral countries, by contrast, targeted the cover ratio and successfully managed to keep it at levels approximately twice as high as required by the bank act. But why, then, did peripheral countries not attach more importance to exchange-rate deviations in setting the discount rate? Was this out of choice or out of necessity? We recall that the exchange-rate performance of peripheral countries was not much worse than that of core countries for most of the time, but deteriorated quickly during periods of global financial strain. Understanding central bank behaviour during these periods will enable us to give an answer to these questions. This is what we turn to now. 3. Central bank behaviour during the periods of global financial strain As indicated earlier, there were three periods of sustained gold point violations by peripheral countries: the Boer War, starting in 1899; the crisis of 1907; and the Balkan Wars of 1912/13. We briefly explain each episode and then analyse the discount rate behaviour of core and peripheral countries during these crises. Subsequently, we will support our analysis of the data by a study of the internal protocols and the Annual Reports of the Austro-Hungarian bank. (a) The Boer War began in October 1899. Early successes of the Boers culminated in the so-called “Black Week” (10th–15th December 1899). As the British Empire grew increasingly determined to win the war, major reinforcements were sent. By January 1900 it became clear that England would, at least, not lose the war (even though hostilities continued for a long time and peace was only achieved in May 1902). For the purpose of this study it is important to keep in mind that South-Africa at the time was the world's largest supplier of gold for a rapidly growing world economy connected to the gold standard. The chronology of military events is mirrored by three discount rate increases of the Bank of England from 3.5% to 6% (3.10., 5.10. 13

Moreover, in the case of estimation (4A), the coefficient on Germany is not significant at conventional levels. In some specifications, the reserve level is not significant at the 10%-level for the peripheral countries, as opposed to estimations (3B) and (4B). This is owed to the fact that we cannot have a group-specific constant (which would allow the constant to assume the function as an offset to the reserve level as argued above) in a country-fixed-effects estimation. The reason for this is that country-fixed effects imply adding dummy-variables for each cross-sectional unit; a group-specific constant means adding (in our case) two dummy variables. If both types of dummy variables are added simultaneously, each group-specific dummy variable can be written as a linear combination of some of the country-specific dummy variables, thereby resulting in a non-full-rank matrix (which can hence not be inverted). 14

General discount rate spread to England and Germany (basis points)

England Germany Sweden 94 Norway 100 Austria-Hungary 26 Italy 22 Romania 172 Average 83 periphery

Episode 1: crisis of 1899–1900 (Boer War) 1.10.1899–10.1.1900

Episode 2: crisis of 1906–1907 (American banking crisis) 11.9.1906–16.1.1907

Episode 3: crisis of 1907–1908 (American banking crisis) 27.10.1907–1.1.1908

Episode 4: crisis of 1912–1913 (Balkan Wars) 27.8.1912–15.4.1913

Discount rate increase during crisis (%)

Maximum discount rate during crisis (%)

Discount rate increase during crisis (%)

Maximum discount rate during crisis (%)

Discount rate increase during crisis (%)

Maximum discount rate during crisis (%)

Discount rate increase during crisis (%)

Maximum discount rate during crisis (%)

2.5 2.0 0.0 0.5 1.0 n.a. 2.0 0.9

6.0 7.0 6.0 6.5 6.0 n.a. 9.0 6.9

2.5 2.5 1.0 0.5 0.5 n.a. 0.0 0.5

6.0 7.0 6.0 5.5 4.5 n.a. 5.0 5.3

2.5 2.0 1.0 1.0 1.0 n.a. 1.0 1.0

7.0 7.5 7.0 6.0 6.0 n.a. 8.0 6.8

2.0 1.5 1.0 0.0 1.0 0.5 1.0 0.7

5.0 6.0 5.5 5.5 6.0 6.0 6.0 5.8

Discount rate spread to England and Germany (bp)

21 70 −1 n.a. 229 80

Discount rate spread to England and Germany (bp)

−7 −31 −125 n.a. −68 −58

Notes: No entries for Italy in episodes 1–3 as Italy was either not yet on gold or not in violation of the gold export point. Sources: Cf. data appendix.

Discount rate spread to England and Germany (bp)

−44 −120 −123 n.a. 80 −52

Discount rate spread to England and Germany (bp)

6 35 54 59 56 42

M. Morys / Explorations in Economic History 50 (2013) 205–226

Table 8 Discount rate behaviour during financial crises.

219

220

M. Morys / Explorations in Economic History 50 (2013) 205–226

and 30.11.) before it fell to 5% on 11th January 1900. We observe a similar pattern for the Reichsbank which raised its discount rate by 2% to 7% (3.10. and 15.11.) before reducing it to 6% on 12th January 1900. (b) The crisis of 1907 is often erroneously reduced to the American banking crisis of the same year but in fact it was a much wider phenomenon (Kindleberger, 2005, pp. 119–120). It followed the business cycle upswing of the first years of the new century in which many countries had participated. Bubbles burst in different places at different times. European discount rate data suggests two waves. The first one began in September 1906 and ran until January 1907 when the first central banks start lowering their discount rates. This was then followed by a second wave in late 1907. The failure of the Knickerbocker Trust Company in New York on 22nd October 1907 soon led to a general suspension of cash payments by the entire American banking system. A week later, Germany (29th October) and England (31st October) started raising their discount rate from 5.5% and 4.5%, respectively, to 7.5% and 7%. These were the highest values for both countries during the Classical Gold Standard era which underlines the severity of the crisis. On 2nd January 1908 the Bank of England became the first major central bank to decrease its discount rate again. (c) The third episode relates to the Balkan Wars of 1912/13. Great power rivalries had manifested themselves on the Balkans since the congress of Berlin (1878), which explains why every new crisis led to increased anxiety worldwide. In the event, the so-called Balkan League was established in early 1912 as an alliance between Greece, Serbia, Bulgaria, and Montenegro with the aim of conquering the European lands of the Ottoman Empire. The outbreak of war was likely as early as summer 1912, eventually breaking out on 8th October 1912. The Bank of England was the first central bank to raise its discount rate, increasing it by 2% to 5% (29.8. and 17.10.) before decreasing it on the 17th April. Germany also increased its discount rate by 1.5% to 6% (24.10. and 14.11.). Table 8 compares the discount rate policy of England and Germany with peripheral countries. The time window is defined as two days in advance of the first central bank raising the discount rate and ends the day prior to the first central bank lowering it again. Table 8 shows by which margin discount rate changes were made and to which level they were raised. We also report the discount rate differentials to England and Germany (represented by the arithmetic average of their respective discount rates) during the time window as opposed to the normal discount rate spread. England and Germany engaged in competitive discount rate increases early on but peripheral countries reacted late and increased their discount rates by substantially lower margins. Consequently, the discount rate differential between core and periphery actually declined during crises. The peripheral economies had an average discount rate spread of 83 basis points. The spread declined during all four crisis episodes and even turned negative during the crisis of 1907. We are hence confronted with a paradox: in crisis situations with prolonged violations of the gold export point, peripheral countries decreased rather than increased their discount rate spread to the core countries. This paradox might be explained as follows: as the general discount rate level was higher for peripheral countries, there was limited room for manoeuvre during periods of global financial strain. Moreover, a signalling problem might have prevented peripheral economies from raising the discount rate by a wide margin. Sizeable discount rate increases could be interpreted as signs of weakness and hence deter rather than encourage the inflow of much-needed short-term capital. Last but not least, the discount rate increases could have been a less effective tool in peripheral countries. In order to provide some qualitative evidence, we studied the annual reports of the Austro-Hungarian Bank and the protocols of the general council. The Annual Reports were intended for the shareholders of the Austro-Hungarian Bank, whereas the protocols of the general council were internal documents. The internal discussions in particular provide an answer as to why peripheral countries used the discount rate tool so sparingly during crises. First, there was a general sense that discount rate increases would be unpopular and to the detriment of domestic business, an argument particularly often advanced by government representatives on the general council.15 The Austro-Hungarian Bank could be very explicit about this; when describing discount rate policy during the Boer War, for instance, they write that “… we should not forget that the bank's duties do not only consist of defending mint parity. It is of no less importance to protect and promote all the other interests of our national economy which is beset with so many difficulties …”. 16 Secondly, members of the governing council doubted the effectiveness of discount rate increases. From a general council meeting held at the height of the American banking crisis (28th November 1907), we learn that Austria-Hungary did contemplate increasing the discount rate further (at this point it was 100 basis points below England and 150 basis points below Germany). They failed to do so because “even a higher interest rate would not have made a difference”, 17 in other words, the exchange-rate would have remained substantially depreciated. If raising the discount rate was difficult and, in times of global financial strain, potentially not even effective, what was the alternative? The annual reports and protocols give indications which help explain the different discount rate policies established in Section 2. Firstly, accumulating large reserves in “good times” and returning them slowly to the market when necessary were seen as a good way of keeping interest rates low and stable: “The enormous increase of our metallic holdings … and, more importantly, the vast stock of foreign bills of exchange and foreign deposits has proven beneficial to the domestic economy. As a result, we could offer relatively low interest rates throughout the year despite adverse interest rates abroad ….”18 The concept that a policy of infrequent 15 General council meeting #523, held 27th June 1907, pp. 5–6. The Austro-Hungarian Bank had two government representatives, one for Austria and one for Hungary. 16 Report to the 22nd General Meeting of the Austro-Hungarian Bank (1900), p. 11. A very similar statement can be found in the report to the 32nd General Meeting of the Austro-Hungarian Bank (1910). 17 General council meeting #528, held 28th November 1907, p. 4. 18 Report to the 29th General Meeting of the Austro-Hungarian Bank (1907), pp. 10–11.

M. Morys / Explorations in Economic History 50 (2013) 205–226

221

discount rate changes required high reserve levels can be found elsewhere 19 and provides qualitative evidence in support of our econometric findings. Secondly, not introducing specie convertibility provided important “breathing space” during periods of prolonged exchange-rate depreciations. As indicated earlier, one of the key institutional differences between core and peripheral countries was that the latter only “shadowed” the gold standard, in other words, they were not constrained by the rather narrow gold points implied by specie convertibility. Austria-Hungary was a case in point, where, despite a long and intense discussion, specie convertibility remained suspended.20 It is therefore interesting to note that during the 1907 crisis, one of the council members stated how helpful it proved in the current financial crisis that Austria-Hungary had not introduced specie convertibility, thus giving more flexibility to the monetary authority.21 Thirdly, “moral suasion” on behalf of the central bank probably played a much bigger role in peripheral money markets. The London money market, due to its size and anonymity,22 reacted to the price of money. In the much smaller and less developed money markets of the periphery (in which, probably for the lack of it, we could not even find data for open market discount rates for most countries, cf. above), by contrast, it was potentially feasible to reduce, if only temporarily, demand for foreign currency by other means. Due to the informal nature of “moral suasion”, it is difficult to provide an abundance of hard evidence to support this claim, but from the sources available it seems that the Austro-Hungarian bank did make use of it.23 In conclusion, the qualitative evidence is supportive of our statistical analysis. Peripheral countries adapted the “English” gold standard in two crucial aspects to suit their needs: firstly, they did not introduce specie convertibility, thus widening the exchangerate bands and hence providing more flexibility; secondly, peripheral countries came to rely on high reserve levels and oriented their discount rate policy towards maintaining the reserve level rather than targeting more narrowly the exchange-rate.

4. Conclusion In this paper we have provided the first systematic comparison of discount rate policy under the Classical Gold Standard based on the concept of a central bank reaction function. Drawing on a new data set of monthly observations for 12 European countries, we analysed the determinants of discount rate policy; in particular, we asked whether core and peripheral countries followed different patterns and why this was the case. Two key findings emerged: firstly, core and periphery differed in the importance they attached to cover ratio and exchange-rate deviations. Core countries used their discount rate changes to target specifically the exchange-rate; and they did so with great success, as evidenced by the superior exchange-rate performance compared with peripheral countries. In turn, they paid little attention to the cover ratio which was allowed to fall close to the minimum level required by the bank act. Peripheral countries pursued the opposite strategy: paying little attention to exchange-rate deviations (hence their poor performance on this account), they targeted the cover ratio and successfully managed to keep it at levels approximately twice as high as required by the bank act. Secondly, an element of interest rate followership vis-à-vis England and Germany was common to core and periphery, though not as strong as one would expect under fixed exchange-rates and capital mobility. Core countries followed more strongly the bank rate set in London, whereas countries on the European periphery were more influenced by the Reichsbank. Thus, our findings suggest that the European branch of the Classical Gold Standard was less London-centred than has hitherto been assumed. We then turned to explaining the differences between core and peripheral countries. The key difference was the effectiveness of the discount rate tool. In the case of core countries, a discount rate increase quickly led to short-term capital inflows and hence an improvement in the exchange-rate. This mechanism did not operate as smoothly for peripheral countries, resulting in more frequent violations of the gold export point. This core–periphery dichotomy was particularly pronounced in periods of global financial strain, when peripheral countries had to live with prolonged periods of unfavourable exchange-rates. As a result, peripheral countries adapted the gold standard in two crucial aspects to suit their needs. Firstly, they did, for the most part, not introduce specie convertibility, thus widening the exchange-rate bands and hence providing more flexibility. Secondly, peripheral countries came to rely on high reserve levels and they oriented their discount rate policy towards maintaining the reserve level rather than targeting more narrowly the exchange-rate. If peripheral countries modified the “English” gold standard to suit their needs, this probably entails a wider lesson for the functioning of the Classical Gold Standard. There was not only one gold standard but a variety of gold standards. Peripheral countries apparently followed a version different from the one pioneered by England. Perhaps it is precisely this institutional flexibility which explains why the Classical Gold Standard remains to this day the longest-ever system of fixed exchange-rates.

19

Report to the 30th General Meeting of the Austro-Hungarian Bank (1908). Hemetsberger-Koller (2001). 21 General council meeting #527, held 9th November 1907, p. 9. 22 A factor emphasized by Capie (1999) in the functioning of the London money market. 23 We thank Clemens Jobst for this insight and sharing the following newspaper entry with us: “Budapest, 29th September (foreign exchange and money markets), Today's developments on the foreign exchange market were tranquil. Acquiescing in the wishes of the Austro-Hungarian bank, several Vienna commercial banks no longer accepted orders from abroad to purchase gold and foreign exchange. The English and French currencies remained sought-after but transactions were less extensive than in previous days. [Demand for] German currency weakened somewhat.” (“Pester Lloyd”, 30th September 1911, p. 11, author's translation from German). The idea that credit rationing must have played a role in peripheral economies can also be found in Neil and Weidenmier, 2003. 20

222

M. Morys / Explorations in Economic History 50 (2013) 205–226

Acknowledgment Earlier versions of this paper were presented at the AEA, EHES, EHS and IEHA meetings; at seminars in Antwerp, Berkeley, Glasgow, Leuven, Moscow (ICEF), NYU, Oxford, Paris, Rutgers, York; and at the EABH Young Scholars Workshop in Rotterdam and the SEEMHN Annual Conference in Bucharest. I am grateful to all participants for their spirited discussion and helpful suggestions, and in particular to Dion Bongaerts, Michael Bordo, Carsten Burhop, Barry Eichengreen, Sergey Gelman, Clemens Jobst, Joost Jonker, Chris Meissner, Larry Neil, Jaime Reis, Max Schulze, Alan Taylor, Marc Weidenmier and two anonymous referees. Appendix A Austria-Hungary, Bulgaria, Romania and Serbia All data except for private discount rates (cf. below) from Monetary Time Series of South-Eastern Europe from the 1870s to 1914. Belgium – Exchange rates: Neal-Weidenmier-Gold Standard data base (England), Schneider et al., 1991, Europäische und nordamerikanische Devisenkurse 1777–1914, vol. 2, pp. 239–240 & vol. 3, pp. 354–356 (France, Germany). – Bank rate: Kauch, La Banque Nationale de Belgique, pp. 148–152. – Reserves and monetary base: “Assemblée Générale des Actionnaires de la Banque Nationale. Rapport fait par le Gouverneur au nom du Conseil d'Administration”, section “Extrait des situations publiées au moniteur belge en …”, Brussels 1878–1914. England – Exchange rates: Neal-Weidenmier-Gold Standard data base (Germany), NBER Macrohistory database #14107 (France). – Bank rate: Hawtrey, A Century of Bank Rate, pp. 281–296. – Reserves and monetary base: Capie & Webber, A Monetary History of the United Kingdom, pp. 408–431. France – Exchange rates: Neal-Weidenmier-Gold Standard data base (England), Schneider et al., 1991, Europäische und nordamerikanische Devisenkurse 1777–1914, vol. 2, pp. 351–352 (Germany). – Bank rate: Hawtrey, A Century of Bank Rate, p. 302. – Reserves and monetary base: “Compte rendu des operations de la Banque de France et de ses succursales pendant l'année 1889” etc., Paris 1890–1914, section “Situation hebdomadaire des principaux comptes de la Banque”. Germany – Exchange rates: Neal-Weidenmier-Gold Standard data base (England), NBER Macrohistory database #14071 (France). – Bank rate: Reichsbank, Vergleichende Notenbankstatistik, pp. 186–189. – Reserves and monetary base: “Verwaltungs-Bericht der Reichsbank fuer das Jahr 1876” etc., Berlin 1876–1914, section “Zusammenstellung der … veroeffentlichten Wochen-Uebersichten”. Italy – Exchange rates: Spinelli, Per la storia monetaria dell'Italia, vol. 2, pp. 45–94 (England), Schneider et al., 1991, Europäische und nordamerikanische Devisenkurse 1777–1914, vol. 3, pp. 22–23 & pp. 69–71 (France, Germany). – Bank rate: kindly communicated by Alfredo Gigliobianco, Historical Archive of the Bank of Italy. – Reserves and monetary base: de Mattia, I bilanci degli istituti di emissione italiani dal 1845 al 1936, vol. 2, pp. 619–753 and pp. 446–454. The Netherlands – Exchange rates: Neal-Weidenmier-Gold Standard data base (England), Schneider et al., 1991, Europäische und nordamerikanische Devisenkurse 1777–1914, vol. 2, pp. 122–123, 126–127, 188–189 (France, Germany). – Bank rate: de Jong, Geschiedenis van de Nederlandsche Bank, vol. 3, pp. 537–543. – Reserves and monetary base: “Verkorte Balans der Nederlandsche Bank”, 1875–1913 (Nationaal Archief, The Hague). Norway – Bank rate: Annual Report of Norges Bank 1979, p. E10. – All other data: downloaded from www.norges-bank.no. Sweden – Exchange rates: Schneider et al. (1993), Statistik der Gold- und Wechselkurse in Deutschland und im Ostseeraum, 18. und 19. Jahrhundert, pp. 299–300, 337–338, 318–320, and 348 (England, France, Germany).

M. Morys / Explorations in Economic History 50 (2013) 205–226

223

– Bank rate: Sveriges Riksbank, Sveriges Riksbank 1668–1924–1931, pp. 136–138. – Reserves and monetary base: “Sammandrag af Bankernas Uppgifter”, Stockholm 1878–1912 and “Sveriges Riksbank Årsbook”, Stockholm 1913–1915. Private discount rates for Austria-Hungary, Belgium, England, France, Germany, the Netherlands Reichsbank, Vergleichende Notenbankstatistik, pp. 212–231. Appendix B. Threshold regressions for crisis periods In defining the precise exchange-rate threshold which might give rise to different discount rate policies, we operate with four different percentages of the gold export point (0%, 33%, 50% and 66%). In the case of a country with a gold export point 1% above the mint ratio, the 50%-threshold, for instance, implies estimating different coefficients for all observations with a monthly exchange rate≥ 1.005 of mint parity. The first case of a 0%-threshold is a pure test of asymmetry (did the central bank react differently to positive as opposed to negative deviations from mint parity?), while the other threshold values refer to increasing levels of exchange-rate pressure. As core countries violated gold points infrequently (Table 1), raising the threshold above 66% reduces the number of observations significantly. All threshold levels suggest parameter stability between “crisis periods” and “normal periods”. We report below the coefficients (for the crisis periods only) obtained for the intermediate level of 33%, a level sufficient to raise concerns over gold outflows for two reasons. Firstly, as we rely on monthly data, an average depreciation of a third of the gold export point will often involve a large number of daily observations that deviated considerably more from mint parity. Second, recent research suggests that the gold points might have been much narrower than conventional estimates suggest. Canjels et al. (2004, p. 880, figures 7 and 8) calculate the so-called revealed preference gold points based on differences in exchange-rate behaviour inside and outside of the exchange-rate bands; their bands have only approximately half the width of the bands calculated by the transaction cost approach to calculating gold points favoured by Officer (1996). The fact that a full 50% of gold exports in the country pair they investigate took place above the revealed preference gold export point but below Officer's export point lends considerable credence to the new much-reduced estimates of Canjels et al. (2004). For the gold export points we use the sources quoted in Table 1; where gold export point estimates are not available for individual countries, we rely on the average value for countries for which we have data. While this inevitably introduces some inaccuracy, this is compensated for by the fact that we are establishing the robustness of our results through the use of different threshold levels, all of which suggest parameter stability. Determinants of actual discount rate changes during crisis periods.

Method (weights)

Pooled least squares (no weights)

Coefficient covariances Estimation 3A: coefficients for crisis periods: core countries pooled α0 Intercept Cover ratio: level in m−1 α1 Cover ratio: change m−2 to m−1 α2 β xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) English bank rate: change d−31 to d−1 γ1 French bank rate: change d−31 to d−1 γ2 German bank rate: change d−31 to d−1 γ3 2 R Estimation 3B: coefficients for crisis periods: peripheral countries pooled Intercept α0 α1 Cover ratio: level in m−1 Cover ratio: change m−2 to m−1 α2 β xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) English bank rate: change d−31 to d−1 γ1 γ2 French bank rate: change d−31 to d−1 German bank rate: change d−31 to d−1 γ3 R2 1 p-Value: 12.3%; 2p-value: 15.7%; 3p-value: 11.4%; 4p-value: 12.3%. Estimation 3C: coefficients for crisis periods: all countries pooled Intercept α0 Cover ratio: level in m−1 α1 Group-specific regressor core Group-specific regressor periphery

Std. −0.13 0.21 −4.24 62.73 0.30 0.14 0.30 0.33

0.66 −0.96 −2.51 −1.47 0.12 0.09 0.33 0.33

Pooled EGLS (cross-section weights) Cross-section SUR

Period SUR

⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎

⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎

⁎⁎ ⁎⁎⁎ ⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎

⁎ ⁎ ⁎⁎

⁎ ⁎ ⁎

⁎⁎⁎ ⁎⁎⁎

1

⁎

2

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

0.17 −0.22 −0.86

Std. −0.24 0.45 −4.51 67.83 0.31 0.09 0.27 0.36

0.72 −1.06 −1.88 2.02 0.11 0.01 0.33 0.35

⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎

⁎⁎ ⁎⁎ 3

4

⁎⁎⁎

0.13

1

⁎

⁎⁎⁎

0.46 −1.00

⁎⁎

(continued on next page)

224

M. Morys / Explorations in Economic History 50 (2013) 205–226

Appendix B (continued) Appendix A (continued) Method (weights)

Pooled least squares (no weights)

Coefficient covariances α2 β

Cover ratio: change m−2 to m−1 xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) Group-specific regressor core Group-specific regressor periphery English bank rate: change d−31 to d−1 French bank rate: change d−31 to d−1 German bank rate: change d−31 to d−1

γ1 γ2 γ3 R2 1 p-value: 13.9%.

Pooled EGLS (cross-section weights)

Std.

Cross-section SUR

Period SUR

−3.46

⁎⁎⁎

⁎⁎⁎

⁎

62.80 0.63 0.22 0.10 0.28 0.32

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

Std. −3.14

⁎⁎⁎

68.09 4.96 0.21 0.04 0.27 0.35

⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎

⁎ Significant at the 10 percent level. ⁎⁎ Significant at the 5 percent level. ⁎⁎⁎ Significant at the 1 percent level.

Determinants of monthly discount rate behaviour during crisis periods.

Method (weights)

Pooled least squares (no weights)

Coefficient covariances Estimation 4A: coefficients for crisis periods: core countries pooled δ0 Intercept Cover ratio: level in m−1 δ1 Cover ratio: change m−2 to m−1 δ2 ε xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) English bank rate: change m−2 to m−1 ζ1 French bank rate: change m−2 to m−1 ζ2 German bank rate: change m−2 to m−1 ζ3 R2 DW 1 p-Value: 11.7%; 2p-value: 19.7%; 3p-value: 12.1%. Estimation 4B: coefficients for crisis periods: peripheral countries pooled Intercept δ0 Cover ratio: level in m−1 δ1 Cover ratio: change m−2 to m−1 δ2 ε xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) English bank rate: change m−2 to m−1 ζ1 ζ2 French bank rate: change m−2 to m−1 German bank rate: change m−2 to m−1 ζ3 2 R DW Estimation 4C: coefficients for crisis periods: all countries pooled Intercept δ0 δ1 Cover ratio: level in m−1 Group-specific regressor core Group-specific regressor periphery Cover ratio: change m−2 to m−1 δ2 ε xr dev. in m−1 vis-à-vis England, France and Germany (pca-series) Group-specific regressor core Group-specific regressor periphery English bank rate: change m−2 to m−1 ζ1 ζ2 French bank rate: change m−2 to m−1 German bank rate: change m−2 to m−1 ζ3 2 R DW 1 p-Value: 27.0%; 2p-value: 25.1%.

Std.

−0.03 0.08 −1.21 14.31 0.05 −0.08 0.08 0.03 1.84

0.10 −0.14 −0.76 −1.48 0.04 0.03 0.05 0.05 1.93

⁎⁎⁎ ⁎⁎

Pooled EGLS (cross-section weights) Cross-section SUR

Period SUR

⁎⁎⁎ ⁎⁎ ⁎

⁎⁎⁎

⁎

⁎

3

⁎ ⁎ ⁎⁎⁎

⁎⁎ ⁎ ⁎⁎⁎

⁎⁎ ⁎ ⁎

⁎⁎

⁎⁎

⁎⁎

⁎⁎⁎

⁎⁎

⁎

1

2

⁎⁎

0.05 0.04 −0.12 −1.00 13.79 −1.36 0.04 −0.01 0.06 0.04 1.80

Std.

−0.07 0.19 −1.15 6.04 0.05 −0.04 0.02 0.03 1.88

0.09 −0.13 −0.62 −1.48 0.03 0.02 0.06 0.05 1.94

⁎⁎⁎ ⁎⁎

⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎⁎

0.04

1

2

⁎⁎⁎

⁎ ⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎

⁎

⁎⁎⁎

⁎⁎

⁎

⁎⁎⁎

⁎⁎⁎

⁎⁎⁎

0.11 −0.13 −0.78 5.48 −1.48 0.03 0.01 0.05 0.04 1.97

⁎ ⁎⁎⁎

⁎⁎⁎ ⁎⁎⁎

M. Morys / Explorations in Economic History 50 (2013) 205–226

225

References Monetary time series of Southeastern Europe from the 1870s to 1914. Bank of Greece Working Paper 2009, Nr. 94. Athens. Bagehot, W., 1878. Lombard Street. A Description of the Money Market, 6th ed. Kegan Paul, London. Bloomfield, A.I., 1959. Monetary Policy under the International Gold Standard, 1880–1914. Federal Reserve Bank of New York, New York. Bordo, M.D., Flandreau, M., 2003. Core, periphery, exchange rate regimes, and globalization. In: Bordo, M.D., Taylor, A.M., Williamson, J.G. (Eds.), Globalization in Historical Perspective. University of Chicago Press, Chicago, pp. 417–472. Bordo, M.D., MacDonald, R., 2005. Interest rate interactions in the Classical Gold Standard, 1880–1914: was there any monetary independence? Journal of Monetary Economics 52, 307–327. Bordo, M.D., Rockoff, H., 1996. The gold standard as a ‘good housekeeping seal of approval’. Journal of Economic History 56, 389–428. Canjels, E., Prakash-Canjels, G., Taylor, A.M., 2004. Measuring market integration: foreign exchange-arbitrage and the gold standard, 1879–1913. The Review of Economics and Statistics 86, 868–882. Capie, F., 1999. Banking in Europe in the nineteenth and twentieth centuries: the role of the central bank. In: Sylla, R., Tilly, R., Tortella, G. (Eds.), The State, the Financial System and Economic Modernization. Cambridge University Press, Cambridge, pp. 118–133. Capie, F.H., Webber, A., 1995. A Monetary History of the United Kingdom. 1870–1982. Volume 1. Data, Sources, and Methods. Routledge, London, New York. Contamin, R., 2003. Interdépendences financières et dilemma de politique monétaire: la Banque de France entre 1880 et 1913. Revue Economique 54, 157–180. Contamin, R., Denise, C., 1999. Quelle autonomie pour les politiques monétaires sous l'étalon-or, 1880–1913? Economie Internationale. La Revue di CEPII 78, pp. 59–84. Davutyan, N., Parke, W., 1995. The operations of the Bank of England, 1890–1908: a dynamic probit approach. Journal of Money, Credit, and Banking 27, 1099–1112. de Cecco, M., 1974. Money and Empire: the International Gold Standard, 1890–1914. Blackwell, Oxford. de Jong, A.M., 1967. Geschiedenes van de Nederlandsche Bank, 2nd ed. (Haarlem). de Mattia, R., 1967. I bilanci degli istituti di emissione italiani dal 1845 al 1936, altre serie storiche di interesse monetario e fonti. Banca d'Italia: Studi e ricerche sulla moneta, vol. 2. Rome. Denzel, M.A., 2008. The European Bill of Exchange. Its development from the Middle Ages to 1914. In: Chaudhuri, S., Denzel, M.A. (Eds.), Cashless Payments and Transactions from the Antiquity to 1914. Franz Steiner Verlag, Stuttgart, pp. 153–194. Dutton, J., 1984. The Bank of England and the rules of the game under the international gold standard: new evidence. In: Bordo, M.D., Schwartz, A.J. (Eds.), A Retrospective on the Classical Gold Standard, 1821–1931. Chicago University Press, Chicago, pp. 173–202. Eichengreen, B., 1987. Conducting the international orchestra: Bank of England leadership under the Classical Gold Standard. Journal of International Money and Finance 6, 5–29. Eichengreen, B., 1995. Central bank cooperation and exchange rate commitments: the classical and interwar gold standards compared. Financial History Review 2, 99–117. Eichengreen, B., Watson, M., Grossman, R.S., 1985. Bank rate policy under the interwar gold standard: a dynamic probit model. The Economic Journal 95, 725–745. Esteves, R., Khoudour-Castéras, D., 2009. A fantastic rain of gold: European migrants' remittances and balance of payments adjustment during the gold standard period. Journal of Economic History 69, 951–985. Esteves, R., Reis, J., Ferramosca, F., 2009. Market integration in the golden periphery: the Lisbon/London exchange, 1854–1891. Explorations in Economic History 46, 324–345. Flandreau, M., 1998. Central bank cooperation in historical perspective: a sceptical view. The Economic History Review 50. Flandreau, M., Jobst, C., 2005. The ties that divide: a network analysis of the international monetary system, 1890–1910. Journal of Economic History 65, 977–1007. Ford, A.G., 1962. Britain and Argentina. Clarendon Press, Oxford. Fratianni, M., Spinelli, F., 2001. Storia Monetaria d'Italia. Lira e politica monetaria dall'Unità all'Unione Europea, 2nd ed. Mondadori, Milan. Giovannini, A., 1986. ‘Rules of the Game’ during the international gold standard: England and Germany. Journal of International Money and Finance 5, 467–483. Goodhart, C.A.E., 1972. The business of banking, 1891–1914. London. Hallwood, C.P., MacDonald, R., Marsh, I.W., 1996. Credibility and fundamentals: were the classical and interwar gold standards well-behaved target zones? In: Bayoumi, T., Eichengreen, B., Taylor, M.P. (Eds.), Modern Perspectives on the Gold Standard. Cambridge University Press, Cambridge, pp. 129–161. Hawtrey, R.G., 1962. A Century of Bank Rate, 2nd ed. Longmans Green & Co., London. Hemetsberger-Koller, H., 2001. Die suspendierte Goldkonvertibilität. Barzahlungskrise in Österreich-Ungarn zu Beginn des 20. Jahrhunderts. In: Bachinger, K., Stiefel, D. (Eds.), Auf Heller und Cent. Beiträge zur Finanz- und Währungsgeschichte. Überreuter, Frankfurt, Vienna, pp. 185–219. Jobst, C., 2009. Market leader: the Austro-Hungarian Bank and the making of foreign exchange intervention, 1896–1913. European Review of Economic History 13, 287–318. Kauch, P., 1950. La Banque Nationale de Belgique. 1850–1918. Banque Nationale de Belgique, Brussels. Kindleberger, C.P., 2005. Manias, panics and crashes. A history of financial crises, Palgrave MacMillan (5th edition). Khoudour-Castéras, D., 2005. International adjustment during the classical gold standard: the migration nexus. Working Paper Chaire Finances Internationales.Science Po, Paris. Lindert, P.H., 1969. Key currencies and gold. 1900–1913. Princeton Studies in International Finance, 24 (Princeton). Lόpez-Cόrdova, J.E., Meissner, C.M., 2003. Exchange-rate regimes and international trade: evidence from the classical gold standard era. American Economic Review 93, 344–353. Martín Aceña, P., Reis, J. (Eds.), 2000. Monetary Standards in the Periphery: Paper, Silver, and Gold. 1854–1933. MacMillan Press, St. Martin's Press, London, New York. Morys, M. “The Classical Gold Standard in the European Periphery: a case study of Austria-Hungary and Italy, 1870–1913.” PhD Diss., London School of Economics, 2006. Neal, L., Weidenmier, M.D., 2003. Crises in the global economy from tulips to today: contagion and consequences. In: Bordo, M.D., Taylor, A.M., Williamson, J.G. (Eds.), Globalization in Historical Perspective, pp. 473–514. Obstfeld, M., Shambaugh, J.C., Taylor, A.M., 2005. The trilemma in history: tradeoffs among exchange rates, monetary policies, and capital mobility. The Review of Economics and Statistics 87, 423–438. Officer, L.H., 1996. Between the dollar-sterling gold points: exchange rates parity, and market behavior. Cambridge University Press, Cambridge. Ögren, A., Oksendal, L.F. (Eds.), 2012. The Gold Standard Peripheries: Monetary Policy, Adjustment and Flexibility in a Global Setting. Palgrave Macmillan, London. Pippenger, J., 1984. Bank of England operations, 1893–1913. In: Bordo, M.D., Schwartz, A.J. (Eds.), A Retrospective on the Classical Gold Standard, 1821–1931. Chicago University Press, Chicago, pp. 203–232. Reichsbank, 1925. Vergleichende Notenbank-Statistik. Organisation und Geschäftsverkehr wichtiger europäischer Notenbanken 1876–1913 statistisch dargestellt. Reichsdruckerei, Berlin. Reis, J., 2007. An “art” and not a “science”? Central bank management in Portugal under the gold standard, 1863–1887. The Economic History Review 60, 712–741. Reti, S.P., 1998. Silver and Gold: the Political Economy of International Monetary Conferences. 1867–1892. Greenwood, Westport (CT), London. Schneider, J., Schwarzer, O., Zellfelder, F., 1991. Europäische und nordamerikanische Devisenkurse 1777–1914. 3 vols. Steiner, Stuttgart. Schneider, J., Schwarzer, O., Schelzer, P., 1993. Statistik der Gold- und Wechselkurse in Deutschland und im Ostseeraum, 18. und 19. Jahrhundert. Quellen und Forschungen zur historischen Statistik von Deutschland, vol. 12. St. Katharinen. Spinelli, F., 1990. Per la storia monetaria dell'Italia, vol. 2 (Turin). Tattara, G., 2000. Was Italy ever on gold? In: Martín Acena, P., Reis, J. (Eds.), Monetary Standards in the Periphery: Paper, Silver, and Gold, 1854–1933. London, New York, MacMillan Press, St. Martin's Press, pp. 18–68. Tattara, G., 2003. Paper money but a gold debt: Italy on the gold standard. Explorations in Economic History 40, 122–142. Tullio, G., Walters, J., 1996. Was London the conductor of the international orchestra, the second violinist or the triangle player? A survey of the literature on asymmetries in balance of payments adjustment during the Classical Gold Standard. Scottish Journal of Political Economy 43, 419–443. Tullio, G., Wolters, J., 2003a. The objectives of British monetary policy during the Classical Gold Standard, 1876–1913: an econometric analysis of domestic and foreign determinants of bank rate. Diskussionsbeiträge des Fachbereichs Wirtschaftswissenschaft der Freien Universität Berlin, Nr. 13.

226

M. Morys / Explorations in Economic History 50 (2013) 205–226

Tullio, G., Wolters, J., 2003b. The objectives of French monetary policy during the Classical Gold Standard, 1876–1913: an econometric analysis of the determinants of the Banque de France's official discount rate. Diskussionsbeiträge des Fachbereichs Wirtschaftswissenschaft der Freien Universität Berlin, Nr. 12. Tullio, G., Wolters, J., 2003c. The objectives of German monetary policy during the Classical Gold Standard, 1876–1913: an econometric analysis of the determinants of the Reichsbank's discount rate. Diskussionsbeiträge des Fachbereichs Wirtschaftswissenschaft der Freien Universität Berlin, Nr. 14. Tullio, G., Wolters, J., 2007. Monetary policy in Austria-Hungary, 1876–1913: an econometric analysis of the determinants of the central bank's discount rate and the liquidity ratio. Open Economy Review 18, 521–537.