The Changing Regime of the Pound/Guilder
Exchange Rate 1600-1912
Peter Kugler
WWZ/Universität Basel
Abstract
Before the “short” 20th century the exchange rate regime was determined by the internal monetary standard:
private arbitrage led to fixed equilibrium exchange rate for currencies with the same metallic standard,
whereas different standard led to flexible rates. However, the strong decrease in transaction costs during the
19th century led to more “fixity” of exchange rates as it is illustrate by application of a TAR model to
historical data for pound-guilder rate: under the 17th and 18th century silver standard, we have deviations of
2.8% from parity until private arbitrage operations with only 28% annual adjustment are triggered. In the
gold standard period this bandwidth is tightened to only 0.25% with full adjustment within one year. The
intermediate period from 1817 to 1874 with a gold pound and silver guilder the no-arbitrage band around a
changing parity according to the gold/silver ratio is 0.84% and 60% adjustment within one year. During the
inconvertibility regime for the pound (1797-1816) we have highly volatile flexible exchange rate as no
parity rate exists and arbitrage operations were not possible.
Key Words: Foreign exchange market arbitrage, parity, transaction costs, threshold
autoregressive model, guilder, pound
JEL: N23
First Draft July 2013
1. Introduction
Before the wide spread creation of central banks at the end of the 19th and beginning of
the 20th century the exchange rate regime between two currencies was determined
spontaneously by the internal monetary regime selected by the two countries: firstly, for
currencies with the same metallic standard (say silver or gold) with full convertibility and
free international movement of money and the monetary metal private arbitrage operation
resulted in an equilibrium exchange rate corresponding to the relative metallic content of
the currencies being exchanged. Secondly, for currencies with different metallic
monetary standard the arbitrage-equilibrium exchange rate evolved according to the
relative price of the two metals. Thirdly, when at least one of the currencies was on a
paper standard no arbitrage was possible and the exchange rate was fully flexible and was
determined by the volume of fiat money issued. However, even with the same metallic
standard and a fixed equilibrium exchange rate actual rates may have been rather volatile
in pre-modern times as these arbitrage operations were subject to rather high transaction
costs. In particular, in preindustrial times hazardous and costly transport routes for
goods and information may often have prevented the arbitrage operations between the
foreign exchange and metal market. Thus, we conjecture a rather broad band of exchange
rate movements around the metallic parity and a corresponding “flexibility” until the
information and transport revolution triggered by the introduction of telegraphy,
steamships and railways in the 19th century.
In this paper the relationship between internal monetary standard and exchange rate
regime is analysed for the pound/guilder exchange rate and the period 1600 to 1912.
These two currencies are of major interest in our context. Firstly, both countries showed
strong financial developments with a Dutch lead during the 17th and 18th century. Indeed,
Amsterdam was considered to be the most important financial and trade centre worldwide
until the late 18th century when London took over this position. Secondly, Amsterdam
and London are relatively close cities and they were well connected by see transport,
which was much less costly than land transport in pre-industrial times. Thirdly, both
2
currencies were based for most of the time period under consideration on the same stable
metallic standard, but there were periods which different monetary standards which
provide illustrations for all the possible exchanger rate outcomes sketched above.
The paper is organized as follows: Section 2 presents the annual data for the Amsterdam
and London foreign exchange markets taken from Denzel (2010) and gives a brief
account of the history of the pound/guilder exchange rate. The Threshold Autoregressive
model which is used to estimate the non-linear adjustment patter to the metallic parity
with transaction costs is described in Section 3. The results are reported in Section 4 and
section 5 concludes.
2. The Pound/Guilder Exchange Rate 1600–1912
In order to analyze the relationship between internal monetary standard and exchange rate
we will consider the historical development of the pound/guilder exchange rate from
1600 to 1912. The historical development of this exchange rate over more than three
centuries is interesting for two reasons: Firstly, during the 17th century the Netherlands
(Dutch Republic) became the economically most advanced country and dominant trade
nation with Amsterdam as the leading financial center. The country had a high saving
rate and an advanced financial system was developed. Moreover, sound government
finance provided a stable currency (Flemish pound = 6 guilder) and low interest rates.
During the 18th century Britain followed the development in the Dutch Republic and
became the economically most advanced country with London as the leading financial
center and the pound as leading currency in the 19th century. Therefore, the exchange rate
between the pound and the guilder assumed a function of major historical importance.
Secondly, the two currencies experienced different monetary standards. The guilder was
based on a silver standard from 1600 to 1839, followed by a brief bimetallistic period
(with a de facto silver standard) and finally switched over to a gold standard in 1875. The
pound was based on a bimetallic standard from 1600 to 1797, on an inconvertible paper
3
standard from 1797 to 1816, and subsequently on the gold standard which was kept until
1914 (Denzel, 2010, pp. 3–4, 57–59) 1.
Figure 1 presents two series, namely the London and Amsterdam market price of the
guilder expressed in pounds, respectively2. The data is based on bills of exchange
between Amsterdam and London drawn in both directions provided by Denzel (2010,
xxii-xlix, 3-101). The data are averages of monthly observations which were, however,
not always available for all 12 months in particular before the 19th century. Thus, the
annual data are based on a varying number of exchange bills. The two prices are highly
correlated, i.e. the simple correlation coefficient is 0.98. Nevertheless the series are not
equal as in a frictionless world and the difference we observe is mainly brought about by
transaction costs preventing arbitrage operations between the two markets. In addition,
we note that we have some missing data until the end of the Napoleonic wars. In
particular this is true for the London market with 39 missing observations. For the
Amsterdam market we note only 15 missing values indicating the leading role of
Amsterdam as financial center in the 17th and 18th century. This position of Amsterdam is
supported by the median number of exchange bills annually available of 12 compared to
only 6 for London. An econometric analysis provided by Kugler (2013) shows that these
two exchange rates could deviate by approximately 2.5% until this gap was closed by
arbitrage operations in the 17th and 18th century. This bandwidth of oscillation was
strongly reduced in the 19th century when it became considerably smaller; namely,
reaching approximately 0.4%. Before the 19th century, larger deviations led to an
adjustment of the London market, whereas thereafter the gap of the two exchange rates
was closed by changes in the Amsterdam market. This pattern of adjustment is consistent
with the leading role of the Amsterdam (London) market in the 17th and 18th (19th)
century.
1
Silver remained the relevant monetary metal for bills of exchange and the convertibility of bank notes of
the Bank of England, even when gold coins crowded out silver coins in domestic circulation and Britain
moved de facto to a domestic gold standard during the 18th century.
4
Figure 1: Pound/Dutch Guilder Exchange Rate in Amsterdam and London 1600–
1912
Guilder: silver (1600–1838), bimetallic (1839–1875), gold (1875–1914)
Pound sterling: bimetallic (1600–1797), paper (1797–1819), gold (1819–1914),
Data: Denzel (2010)
.12
.11
.10
.09
.08
.07
1600
1650
1700
1750
1800
1850
1900
£/Guilder London
£/Guilder Amsterdam
Figure 1 suggests a long run stability of the exchange rate from 1600 to 1785. In the 19th
century after the Napoleonic wars we note again long run exchange rate stability, but now
at a lower level. Indeed, the unit root hypothesis is clearly rejected by the Phillips-Perron
test for the both sub-periods 1960 -1815 and 1816 – 1912 3.
For the international gold standard in the period 1875–1914 we note very low exchange
rate volatility after the information and transport revolution in the second half of the 19th
century. This was brought about by private arbitrage operations at low transaction costs.
The gold content of the guilder and the pound was 0.605 and 7.3 gr., respectively, and the
3
The PP-statistic is -3.23 and -10.12 for London as well as -4.07 and -10.66 in the two periods,
respectively.
5
corresponding equilibrium exchange rate was therefore 0.0829 £/G. Now assume the
exchange rate would be .08£/G. With 1 pound we can buy 12.5 guilder convert them into
7.56 gr. gold in the Netherlands and convert this in England into 7.56/7.3= £1.035, i.e.,
we would make an arbitrage profit of 3.5%.
During the period immediately before (1816–74) when the guilder was based on a silver
or bimetallic (de facto silver) standard and the pound on a gold standard, we see more
flexibility in the exchange rate, and, in particular, we observe very high volatility in the
period of the pound’s inconvertibility during the Napoleonic wars (1797–1816) where no
arbitrage was possible. Arbitrage possibilities exists but are more complicated for the
silver-gold case than the same metallic standard: if the exchange rate of the Dutch (silver) guilder against the (gold-) pound was higher than the silver-gold ratio, we could buy
guilder for pounds, convert them into silver, sell the silver to buy gold, which we
subsequently convert into pounds, and thereby make an arbitrage gain. Thus in the long
run, the exchange rate was flexible but followed the relative price developments of silver
and gold. In Figure 2, the London pound sterling/guilder exchange rate and the
corresponding silver/gold parity rate is displayed for the period 1820–1874. The parity
rate is calculated by dividing the silver content of the guilder (9.61 g, 1816–1838, 9.45 g,
1839–1874) by the silver-market value of the 7.3 g of gold contained in the pound
sterling. In order to gain the silver value of the pound, we multiplied 7.3 by the London
market price ratio of gold to silver, which is displayed for the period 1687–1874 in Figure
3. Figure 2 suggests a rather high correlation between the market and parity exchange
rate. Indeed, the simple correlation coefficient between the two series is 0.86.
6
Figure 2: Log Pound Sterling/Dutch Guilder Exchange Rate in London and the
Varying Silver/Gold Parity, 1820-1873
.090
.088
.086
.084
.082
.080
.078
20
25
30
35
40
45
50
55
60
65
70
£/Guilder London Market
£/Guilder Varying Silver/Gold Parity
Figure 3: Gold/Silver Market Price Ratio, 1687–1873
Gold/Silver Price Ratio London Market
17.0
16.5
16.0
15.5
15.0
14.5
14.0
1700
1725
1750
1775
1800
1825
1850
1875
Data source: Officer and Williamson (2013).
7
The remarkable long-run stability of the exchange rate during most of the 17th and 18th
century (before the French revolution) is explained by a stable link to silver of both
currencies. Indeed, the silver content of the guilder was 10.4 g and that of the pound, 111
g (Denzel, 2010, pp. 3–4, 57–59), which implies a silver parity exchange rate of 0.094
pounds/guilder, which is close to the mean of the exchange rate until the end of the 18th
century. In the 17th century, we note some strong deviations from parity caused by
difficult political conditions (civil war in England and three Dutch-English wars, the Nine
Years’ War) with temporary restrictions on the free movement of silver.
Table 1 collects a couple of descriptive statistics of the deviation from parity for the
silver standard (1600–1785) and gold standard (1876–1912), as well as the silver/gold
interlude (1817–1874). We use the exchange rate data from the leading market, namely
Amsterdam (1600–1785) and London (1817–1912), respectively. All these statistics
indicate a strong tendency towards narrower bounds around the parity exchange rate: the
standard deviation is reduced from 3.57% (1600-1785) to 0.19% (1975-1912). Equally
impressive are the observed ranges: in the 17th and 18th century, we have values ranging
from -12% to 18%, whereas in the gold standard period these figures range from -0.9% to
0.08%. Note that in the first and second period, we have a slight average premium on the
guilder (mean = 0.32/0.38%), whereas in the gold standard, we have a slight average
discount on the guilder (mean = -0.17%).
Table 1: Deviations from Parity Pound/Guilder Exchange Rate, 1600–1912, in
Percent
Period
Pound/Guilder
Mean
Standard
1600–1785
Standard
Min
Max
deviation
0.32
3.57
-12.2
18.1
0.38
1.44
-2.91
6.50
-0.17
0.19
-0.93
0.082
Silver/silver
1816–1874
Gold/silver
1875–1912
Gold/gold
8
3. Econometric Model
Now let us turn to the econometric model used in this study. Let y be the log
pound/guilder exchange rate in London or Amsterdam and ypt the metallic parity rate,
respectively. We apply the following threshold autoregressive model for these series 4:
∆yt = a1 + λ1 ( yt −1 − ypt ) + ε t , if abs ( yt −1 − xpt ) < τ
∆yt = a2 + λ2 ( yt −1 − ypt ) + ε t , if abs ( yt −1 − xpt ) ≥ τ
λ1 = 0, λ2 ≤ 0;
(1)
Var (ε t ) = σ 2 / nt
The model above is slightly different from the standard TAR as we take into account the
heteroskedasticity of the error term. As we noted in section 2 we have different numbers
of observations used to calculate the mean annual exchange rate in the two markets, in
particular in the 17th and 18th century. This time varying precision of the exchange rates
calculated is taken into account by dividing a time invariant error variance by the number
of observations used, n, respectively. This reflects the formula for the standard deviation
of the arithmetic mean. This time varying variance is incorporated by a GLS-approach
involving simply the division of the observations entering equation (1) by n0.5.
If transaction costs are relevant, we expect λ1 = 0 and the other error correction
coefficient to have a negative sign. An intercept term, which may differ across the two
regimes, is included in the model. As the two series have no drift, the constant term can
only arise from an intercept term in the equilibrium relationship ( yt −1 − ypt − c) indicating
a non-zero long-run equilibrium value for the parity deviation. Correspondingly,
a1 = 0 should hold, too.
4
This model can be directly applied to stationary series (as indicated to be the case) when the equilibrium
coefficients are known to be one, as in our case.
9
Defining zt −1 = yt −1 − ypt and a dummy variable d, which is one when abs ( zt −1 =) < τ
holds and otherwise 0, we can rewrite our model in one equation:
∆yt = (a1 + λ1 zt −1 )d t −1 + (a2 + λ2 zt −1 )(1 − d t −1 ) + ε t
(2)
These equation can be estimated by GLS when τ is known. In order to determine this
parameter, we use a grid search over the possible interval 5 to get the value which
minimizes the residual sum of squares of equation (2). The parameter τ is only identified
if this non-linearity is present. This condition appears to be fulfilled as the non-linearity
tests reported for the exchange rate on the Amsterdam and London market clearly reject
the linearity hypothesis (Kugler, 2013).
4. Empirical Results
Table 1 reports the results for the TAR-model for three subsamples, namely 1600-1785
and 1817-1874, and 1876-1912. For the first Period we have a constant silver parity for
both currencies. From 1797-1816 the pound was on a paper standard and there is no
defined parity rate 6. For the period 1817-1874 the metallic parity is determined by the
silver/gold price ratio as displayed in Figure 2 and the last sub sample is the gold standard
period with a fixed gold parity. From the two exchange rate series available we selected
always that of the leading market, namely Amsterdam (1600-1785) and London (18171912), respectively.
Table 2 displays the estimation for a restricted variant of the TAR model, namely the
error correction coefficient and the intercept in the inner band is set to zero in both
equations ( λ1 = 0 , a3 = 0 ). Of course, this restriction is motivated by transaction costs.
Allowing these coefficients to be different from zero does, however, not have an essential
influence on our findings as the corresponding estimates are statistically insignificant.
The range of τ is restricted to values between 10% and 90% quantiles of the absolute values of z.
The years of, and immediately before the French Revolution are excluded as they are a period of
transition with a relatively high number of missing observations.
5
6
10
Table 2 shows a threshold estimate of 2.8% for the first sub-period. This threshold
corresponds nicely with the value estimated for the adjustment between the Amsterdam
and London market which is obtained as 2.4% (Kugler, 2013). The estimated adjustment
coefficient (-0.24) is highly statistically significant but has relatively low absolute value:
only one fourth of the deviation from parity is corrected within one year. The a-estimate
is statistically highly significant and indicates that we have a premium on the
internationally leading currency of approximately 4% 7.
The estimates for the 1817-1872 period shows three major breaks. Firstly, the threshold
estimate is now 0.87%, which is only one third of the earlier value. This reflects the
increased arbitrage possibilities and increased international financial integration triggered
by the information and transportation revolution of the 19th century. Secondly, the
adjustment pattern is much faster: the EC-Coefficient estimate is now -0.604 and
statistically highly significant. Thirdly, the estimate of the constant term is statistically
significant wit a reversed sign indicating that we have now a premium on the pound of
approximately 1.3%. Finally this premium disappears in the gold standard period and we
have now nearly complete adjustment (0.923) with a band of only 0.25%. These results
confirm with the findings for the integration of the Amsterdam and London market for
the entire period 1816-1912 obtained by Kugler (2013): the estimated band was 0.44%
and the EC coefficient in the Amsterdam market was 0.7 in absolute value. Moreover, the
gold standard results are consistent with the findings of Caniels et al. (2004) for the
pound-dollar exchange rate. The strong move to lower band and faster adjustment from
the gold/silver regime to the gold/gold regime in the 19th century is explained by
technological progress and by fact the former regime has ceteris paribus higher
transaction costs as the arbitrage operation involving both metals is more complicated.
7
Remember that constant term in the EC equation stems from an intercept term in the equilibrium
deviation. Correspondingly, the a-coefficient is equal to -λc and for non-zero λ values we get c=-a/ λ which
is equal to 0.0397 with our estimates.
11
Table 2: Estimation Results Threshold Autoregressive Model Log Pound/Guilder
Exchange Rate, 1600-1912
∆yt = a1 + λ1 ( yt −1 − ypt ) + ε t , if abs ( yt −1 − xpt ) < τ
∆yt = a2 + λ2 ( yt −1 − ypt ) + ε t , if abs ( yt −1 − xpt ) ≥ τ
λ1 = 0, λ2 ≤ 0;
Var (ε t ) = σ 2 / nt
a2
λ2
τ
R-squared
DW
1600-1785
0.00956
(0.0033)
-0.241
(0.0443)
0.028
0.263
1.77
1817-1872
-0.0078
(0.0028)
-0.604
(0.110)
0.0087
0.364
1.43
1875-1912
0.00003
-0.923
(0.00024) (0.0629)
0.0025
0.859
1.99
Standard errors in parentheses
12
5. Conclusion
Before the wide spread creation of central banks at the end of the 19th and beginning of
the 20th century the exchange rate regime between two currencies was determined
spontaneously by the internal monetary regime selected by the two countries. For
currencies with the same metallic standard with full convertibility and free international
movement of money and the monetary metal private arbitrage operation resulted in an
equilibrium exchange rate corresponding to the relative metallic content of the currencies
being exchanged. When currencies had different metallic monetary standard, the
arbitrage-equilibrium exchange rate evolved according to the relative price of the two
metals. Finally, the exchange rate was fully flexible and was determined by the volume of
fiat money issued when at least one of the currencies was on a paper standard. However,
even with the same metallic standard and a fixed equilibrium exchange rate actual rates
may have been rather volatile in pre-modern times as these arbitrage operations were
subject to rather high transaction costs. In particular, in preindustrial times hazardous
and costly transport routes for goods and information may often have prevented the
arbitrage operations between the foreign exchange and metal market. Thus, we conjecture
a rather broad band of exchange rate movements around the metallic parity and a
corresponding “flexibility” until the information and transport revolution triggered by the
introduction of telegraphy, steamships and railways in the 19th century.
The relationship between internal monetary standard and exchange rate regime is
analysed for the pound/guilder exchange rate and the period 1600 to 1912. These two
currencies are of major interest in our context as they played an important international
role and were based for most of the time period under consideration on the same and
rather stable metallic standard but there were periods which different monetary standards
(paper standard of the pound 1797-1816, silver guilder and gold pound 1817-1874).
The estimation of a TAR model for the pound-guilder exchange rate show indicate a
strong tendency towards narrower bounds around the parity exchange rate: In the 17th and
18th century with silver standard, we have deviations of 2.8% from parity until arbitrage
operations are triggered, which lead to a rather slow adjustment of 28% within one year.
13
In the gold standard period, this bandwidth is tightened to only 0.25%, where a near full
adjustment is achieved within one year. The intermediate period from 1817 to 1874 with
a silver guilder and a gold pound is within these two extremes with a parity deviation
bandwidth of 0.84% and 60% of the adjustments achieved within one year. The parity
rate is determined by the gold silver market price ratio. During the inconvertibility regime
for the pound (1797-1816) we have high volatility of the exchange rate as no parity rate
exists and arbitrage operations were not possible.
6. References
Bernholz and Kugler, P.: Financial Market Integration in Early Modern Spain: Results
from a Threshold Error Correction Model, Economics Letters 110, 2011, 93-96.
Canjels, E., Prakash-Canjels G. and Taylor, A. M.: Measuring Market Integration:
Foreign Exchange Arbitrage and the Gold Standard, 1879-1913, The Review of
Economics and Statistics 86 (4), 2004, 868-882.
Denzel, M. A.: Handbook of World Exchange Rates, 1590-1914, Ashgate, Burlington,
2010
Kugler, P.: Financial Market Integration in Late Medieval Europe: Results from a
Threshold Error Correction Model for the Rhinegulden and Basle Pound, 1365-1429,
February 2011, Swiss Journal of Economics and Statistics 147, 3, 337-357.
Kugler, P: The Changing Pattern between the Foreign Exchange Markets in Amsterdam and
London 1600-1912, mimeo, University of Basel, February, 2013.
Prakash G. and Taylor, A. M.: Measuring Market Integration: A Model of Arbitrage with
an Econometric Application to the Gold Standard, 1879-1913, NBER Working Paper
6073, June 1997.
14