The Mathematics of Foreign Exchange

Keir G. Armstrong

Carleton University, Ottawa, ON, K1S 5B6, Canada

[email protected]

April 29, 2021

It is well known that, since the supply of the currency of one country is the

demand for the currency of another and vice versa, either may be treated as the

quantity and the other as the price in a standard neoclassical model of the market

for foreign exchange between the two countries. What is less well known is how

the shape of the one curve is related to the shape of the equivalent other. Haberler

(1936; 1949) and Machlup (1939; 1950) articulated this relationship to some extent,

but neither they nor anyone else appears to have developed fully the particulars of

the mathematics behind it. The present paper attempts to fill that lacuna.

Key Words: foreign exchange; supply and demand; functional­form relationships.

JEL Classification Numbers: A2, C02, D41, F31.

Declarations of Interest: None.

1. Introduction

There have been two main approaches to the development of a theory of short­run

exchange­rate determination.1 The more recent is the asset­market approach, which was first

suggested in the 1970s and views exchange rates as being determined “in terms of stocks

of currencies relative to the willingness of people to hold these stocks. Several variants of

stock­based theories of exchange rates have been developed [over the] years, where these theories

differ primarily in the range of different assets that are considered, and in how quickly product

prices can adjust to changes in exchange rates” (Levi, 2009, p. 187).

The older traditional flow approach, also called the balance of payments view or the

exchange­market approach, begins with the (irrefutable) fact that real­world (flexible) exchange

rates are determined in foreign exchange markets subject to the forces of supply and demand

(Gandolfo, 2001, p. 226). These forces result from the various components of the balance of

payments, which account for flows of goods and services and financial capital and transfers

across national borders, as well as from speculation about future exchange­rate movements. It

is in specifying such supplies and demands that problems with the flow approach arise. The

simplest way to do so is to follow Levi (2009, p. 166) and assume that their slopes depend on the

1
A third approach, purchasing power parity (PPP) theory, in either its absolute or relative version, “is put forward

as a long­run theory of the equilibrium exchange rate, in the sense that in the short run there may be marked deviations

from PPP which, however, set into motion forces capable of bringing the exchange rate back to its PPP value in the long

term” (Gandolfo, 2001, p. 224).

2
effects of exchange rates on values of imports and exports, respectively, and that each of the other

components of the balance of payments as well as speculation “can be considered as shifting the

supply or demand curve.” In other words, ceteris paribus, the quantity supplied of a country’s

currency for foreign­exchange purposes depends on the values of its imports at different values

of the exchange rate measured in terms of foreign­currency units, the quantity demanded of the

country’s currency for foreign­exchange purposes depends on the values of its exports at different

values of the exchange rate measured in terms of foreign­currency units, and all other influences

on these quantities supplied or demanded are independent of the exchange rate.

“Since the supply of one currency constitutes the demand for the other and vice versa, we

may treat either of them as a commodity [i.e., quantity] and the other as money [i.e., price]”

(Haberler, 1936, p. 19). This is true because a willingness to buy a currency for foreign­exchange

purposes at a particular exchange rate must be accompanied by a willingness to sell the relevant

other currency for the same purposes at the reciprocal of the particular exchange rate, which is of

course measured in terms of units of the desired currency. There are therefore two equivalent ways

to view a given market for foreign exchange between two countries with different currencies.

And as Machlup (1939, p. 376) said, “It is not difficult, for example, to translate the demand for

dollars on the Paris market into a supply of francs on the combined foreign exchange market,

and likewise the supply of dollars on the Paris market into a demand for francs on the combined

foreign exchange market.” Amusingly I am sure to most current university instructors, he went on

to say:
It is a good undergraduate exercise to practice such a translation: starting from a demand
curve for dollars in terms of francs the amounts of dollars are shown by the horizontal axis
( ), the amounts of francs offered in exchange for these dollars are shown by the rectangle
( ); this gives a supply curve of francs for dollars where the abscissae correspond to the

3
 values of the rectangles in the original demand curve, while the ordinates on the new supply
curve, i.e. the prices of francs in terms of dollars, correspond to the quotient of the abscissae
divided by the values of the rectangles of the original demand curve. The analogous
calculation has to be done in order to transform the original supply curve of dollars in terms
of francs into a demand curve for francs in terms of dollars. This sounds complicated—yet
every sophomore ought to be able to do it, or he has never grasped the meaning of demand
and supply curves. [Ibid.] 2

Such an exercise of translating a demand or supply curve of one currency in terms of

another into the equivalent supply or demand curve of the second currency in terms of the first

suggests that the shapes of the two curves are related in a particular way. Haberler (1936, p. 20)

would seem to have been the first to suggest that there might “be a point at which the supply curve

curls backwards and slopes upwards to the left.” He also noted that “the supply and demand curve

are not symmetrical” and “[t]he point where the supply curve turns to the left corresponds to that

point in the demand curve when the area of the inscribed rectangle begins to diminish.” The latter

claim is demonstrated graphically in relation to a straight­line demand curve (for francs in terms

of dollars) in Haberler (1949; pp. 204–5). Machlup (1939, p. 384) also discussed the possibility

of a backward­bending supply curve and noted that it “cannot rise backward more sharply” than

shown in his “Figure 2” since “[t]he slope of the negative inclination must at every point be

greater than that of the rectangular hyperbola passing through that point.” This claim is illustrated

more clearly in relation to a gently curved (convex­to­the­origin) demand curve (for lire in terms

of dollars) in Machlup (1950; Fig. 1).

More recent treatments of the traditional flow approach discuss in elasticity terms the

shapes of one pair of supply and demand curves, but make no reference to the equivalent other.

2
“Professor Machlup . . . says that every undergraduate ought to know how that is to be done. But experience

shows that he is too optimistic in making that assumption” (Haberler, 1949, p. 204).

4
Gandolfo (2001, §7.3) and Levi (2009, Ch. 8) in particular also discuss the possibility of a

downward­sloping supply curve with justifications rooted in the fact that “for currencies we

plot values (price quantity) on the horizontal axis, whereas we normally [for non­currency

goods] plot just physical quantities” (Levi, 2009, p. 178). Both provide diagrams illustrating

the stability­of­equilibrium implications of downward­sloping supply with Gandolfo (2001,

Fig. 7.1b) replicating (without attribution) the aforementioned backward­bending form derived by

Haberler (1949).

An exception to the no­reference­to­the­equivalent­other norm is OpenStax College (2018),

which asserts that, “In foreign exchange markets, demand and supply become closely interrelated,

because a person or firm who demands one currency must at the same time supply another

currency—and vice versa” (ibid., §29.1). Subsequently, however, there is no mention of how

the assumed shape of one curve affects the shape of the other. And although the two­graph

diagram illustrating “Demand and Supply for the U.S. Dollar and Mexican Peso Exchange Rate”

(ibid., §29.2) portrays the corresponding equilibrium points correctly, it errs in its portrayal of

corresponding non­equilibrium points.3

What has been missing from the literature, then, is a thorough analysis of the relationship

between an assumed functional form of a country’s foreign­exchange supply or demand and that

of another country’s foreign­exchange demand or supply. Section 2 specifies this relationship in

general mathematical terms and then analyzes and provides illustrations of five different pairs of

closed­form specifications of supply­and­demand or demand­and­supply. Section 3 concludes

3
For instance, since is very clearly a point on the supply curve of U.S. dollars, = together with

should be a point on the demand curve for Mexican pesos, which it most definitely is not!

5
after arguing that such specifications provide a more feasible and expeditious basis for empirical

estimations than do those that might be derived from the underlying foreign­exchange demand

behaviour of domestic and foreign agents (mostly large firms) for the purposes of international

trade and foreign direct and portfolio investment abroad.

2. Mathematical Analysis

For the sake of simplicity, consider a world comprising just two countries, each with its

own currency, and a single, competitive, free market within which one currency can be exchanged

for the other. Taking the perspective of one country or the other, the market quantity supplied

or demanded of the domestic currency is some non­decreasing or non­increasing function

: ++ ++ of the exchange rate measured in units of foreign currency per unit of domestic

currency.4 The inverse correspondence ¡1

( ) can be used as the price5 in converting a market

quantity of domestic currency units into the equivalent quantity of foreign currency units ¤
:

¤ ¡1
= = ( ) =: ( ) . (1)

¡1
Clearly, the inverse correspondence ( ) converts the quantity ¤
into the quantity and must

be decomposable into the inverse (or reciprocal) exchange rate ¡1

times ¤
:

¡1 ¤ ¡1 ¤
= ( )= . (2)

4
That is, the market for foreign exchange is assumed to be characterized by an upward­sloping supply curve

together with a downward­sloping demand curve.

5
That is, willingness to accept or pay for the marginal unit on the part of sellers or buyers, respectively.

6
 ¡1
The inverse exchange rate is given by a correspondence ( ) defined by
¡1 ¤
¡1 ( )
= ¤
, (3)

which means that the market quantity demanded or supplied of foreign currency is given by

( ¡1
), which we would like to be a function. Whether it is or not depends on the assumed

functional form of ( ) as we will see below. Note that

¡1 ¡1 ¤ ¡1
( ) ( )= 1. (4)

Assuming that ¡1
( ) is a differentiable function, so is ( ) and the elasticity of supply of or

demand for domestic currency is

¡1 ¡1
( ) ( )
:= = ¡1 0 ( )
= 0 ¡1 (
. (5)
( ) )
¡1 ¡1
Assuming that ( ) is a differentiable function, so is ( ) and the elasticity of demand for or

supply of foreign currency is

¡1 ¡1 ¤ ¡1 ¤
( ) ( )
:= ¡1 = ¤ ¡1 0 ¤)
= ¡1 0 ¤) ¡1 ¤)
. (6)
¤ ( ( (
¤

¡1 0 1 ¡1 1
Since ( ¤
)= 0
( )
and ( ¤
)= ¡1 ( )
, we have
1 0
¡1 ( ) ( )
= 1 1 = ¡1 ( 0 . (7)
0
( ) ¡1 ( ) ) ( )

Summing the two elasticities as functions of yields

+ = 1. (8)

Note that this result is the same as that inferred by Haberler (1949, p. 205) from a geometric

analysis of the affine demand case.6 The foregoing constitutes a more rigorous proof of it.

6
Since Haberler (1949, n. 1) “follow[s] the usual procedure of taking as positive (although the slope of the demand

curve is conventionally called negative, [he felt the need] to use the same convention for the supply elasticity. It follows

7
 The algebra of the affine demand case proceeds from assuming that

¡1
( )= , (9)

where 0 and 0 are constants. This assumption implies that

2
( )= (10)

so that

=1 , (11)

= 2, (12)

and
¤ 2 ¤
( )+ = + =0. (13)

Solving this (latter) equation for using the quadratic formula gives us
2 4 ¤
¡1 ¤
( )= (14)
2
and then
2 4 ¤
¡1 ¤
( )= ¤
, (15)
2
which is not a function—its inverse is, however, albeit one that is non­monotonic with a unique
¡1
global maximum at =2 . In the related special case of perfectly elastic demand (or supply),
¡1
= 0 so that ¡1
( )= , + ¤
= 0, and ( ¤
)= ¡1
.

The “augmented power” (supply or demand) case assumes that

¡1 ¡1
( )= + , (16)

that is positive when the supply curve is negatively inclined and negative when it is positively inclined.”

Therefore, the statement of his result in terms of the present notation is .

8
where , = 0, and are constants, which implies that

+1
( )= + (17)

and then
+1
+
= +1
, (18)

+1
( + 1)
= +1
, (19)

and
1
¤ +1
¡1 ¤ ¤ ¡1
( )=( ) . (20)
¡1
In the related special case of isoelastic demand, = 0, 0, and 0 so that ( )=

and
1
¡1 ¤ ¤ ¡
( )=[ ( ) ] +1 , (21)

which is an increasing (isoelastic inverse supply) function over all ¤

0 if and only if 1.

In the related special case of linear supply, = 0, 0, and = 1 so that ¡1

( )= and
1
¡1 ¤ ¡2
( ¤
)=( ) , which is a non­affine (isoelastic inverse demand) function.

The “depressed quadratic” supply case assumes that

¡1 2
( )= + , (22)

where and 0 are constants, which implies that

3
( )= + (23)

so that
2
+
= 2
, (24)
2

9
 2
+3
= 2
, (25)
2
and
3 ¤
+ =0. (26)

Solving this (“depressed cubic”) equation for using del Ferro’s (circa 1515) formula7 gives us

¤ 2 3 ¤ ¤ 2 3 ¤
¡1 3 3
¤
( )= + + + , (27)
2 3 2 2 3 2
which is an increasing, strictly concave, non­negative, real­valued (inverse demand) function over

all ¤
such that
3 ¤ 2 3 2
27 ( ) +4 0 (28)

or, equivalently,
¡ 3
¤ 2 0 if 0
27 . (29)
0 if 0

The “generalized depressed cubic” supply case assumes that

¡1 2 +1
( )= + , (30)

where , 0, and are constants,8 which implies that

+1 2( +1)
( )= + (31)

so that
+1
+
= +1
, (32)
+ (2 + 1)

7
See Contreras (2015, pp. 25–26).

8
Note that together with and would render this case to be that of affine demand above.

10
 +1
( + ) 2 ( + 1)
= , (33)
+ (2 + 1) +1
and
2( +1) +1 ¤
+ =0. (34)

Solving this equation for := +1

using the quadratic formula gives us
2 +4 ¤
¡1 ¤ +1
( )= , (35)
2
which is an increasing, strictly concave, non­negative, real­valued function (ignoring paired

non­positive real values when is even and paired non­real values when is odd) over all ¤
0.
¡1
In the related special case of affine supply, = 0 so that ( )= + and
+ 2 +4 ¤
¡1 ¤
( )= ¤
, (36)
2
which is a decreasing, strictly convex (and hence non­affine), non­negative, real­valued (inverse

demand) function over all ¤

0.

The “augmented logarithmic” (supply or demand) case assumes that

¡1
( )= + ln , (37)

where and 0 are constants, which implies that

( )= + ln (38)

so that
+ ln
= , (39)
1

ln 1
= , (40)
1

11
and
¤
[exp(ln )] ln = , (41)

which implies in turn that

¤¡
ln = (42)

by the definition of the Lambert W function.9 Consequently,

¤¡
exp
¡1 ¤
( )= ¤
, (43)

which is not necessarily a function and not necessarily everywhere decreasing over all ¤
0.

The preceding five cases would seem to constitute all those that are both analytic and have at

least subdomains of positive prices (exchange rates) for which the associated supply and demand

relations are, respectively, non­decreasing and non­increasing functions. The key to constructing
¡1
them was to first specify a functional form for ( ) that has an analytic inverse ( ). In doing

so, it also has to be true that some parameterization of the form for ( ) has an associated form

for ( ) that is non­increasing or non­decreasing on some part of its domain and that the same
¡1
parameterization of the form for ( ) has an associated form for ( ) that is non­decreasing or

non­increasing on some part of its domain.

Illustrations of the five cases are provided by the eight figures herein. Each figure includes

(i) the graph of a specific parameterization of one of the five functional forms of ( ) labelled

“S” or “D” as appropriate with the vertical axis measured in units of foreign currency ( ) per

unit of domestic currency ($) and the horizontal axis measured in ­illions10 of units of domestic

9
is defined as the solution to the equation and thereby “answers the question ‘What

power of Euler’s number, multiplied by itself, produced the number ?’” (Lehtonen, 2016, p. 1111)

10
That is, ones, tens, hundreds, thousands, tens of thousands, hundreds of thousands, millions, etc. corresponding to

12
currency, (ii) the graphs of the associated ( ) and its mirror image in relation to the 45­degree
¡1
line ( ) with the horizontal axis measured in ­illions of units of domestic or foreign currency

and the vertical axis measured in ­illions of units of foreign or domestic currency, (iii) the graph

of the associated ( ) labelled “D*” or “S*” with the vertical axis measured in units of domestic

currency per unit of foreign currency and the horizontal axis measured in ­illions of units of

foreign currency, and (iv) a pair of dashed straight­line segments highlighting the one­for­one
¡1
crossing points of ( ) and ( ) and of ( ) and ( ). In addition, Figure 1 includes a sequence

of dotted­and­dashed straight­line segments that show how to read the constituent graphs from

(a) = 2 units of foreign currency per unit of domestic currency yielding (3 9 2) 0 9 = 2 ¹1

­illions of units of domestic currency demanded by foreigners to (b) 2 2 ¹1 = 4 ¹2 ­illions of

units of foreign currency sold by foreigners in exchange for 2 ¹1 ­illions of units of domestic

currency to (c) 2 ¹1 ­illions of units of domestic currency bought by foreigners in exchange for

4 ¹2 ­illions of units of foreign currency to (d) 4 ¹2 ­illions of units of foreign currency supplied

by foreigners at 2 ¹1 4 ¹2 = 0 5 = ¡1
units of domestic currency per unit of foreign currency.

Figure 5b is read in the same way; Figures 2a through 5a are read similarly, but in relation to the

domestic supply of domestic currency S first and the domestic demand for foreign currency D*

last (or in the reverse order starting with a value for ¡1

).

3. Conclusion

The “factors generally thought to influence exchange rates . . . include a country’s inflation

equal to zero, one, two, three, four, five, six, etc. Note that the etymology of the word “zillion” is “Z (perh[aps] =

unknown quantity) + MILLION” according to the Oxford English Dictionary (and other expert sources).

13
rate, real economic growth rate, interest rates relative to the rest of the world, and private

speculation. Theories of the exchange rate differ because of the assumptions they make about

the importance of these factors” (Pearce, 1983, p. 21). As criticisms have been levelled against

all of these theories mostly on the basis of what each leaves out of consideration of necessity in

aid of maintaining tractability, it would seem that direct estimation of foreign­exchange supply

and demand is the best approach to modelling exchange­rate determination empirically. To do

so would require solving the associated identification problem, of course, which can be done

in principle using supply­ or demand­side instruments as explained by MacKay and Miller

(2019). The product demand schedule specified therein (by equation 1) takes a semi­linear

form analogous to ( ) = ( ( )), where the function ( ) is such that the composition

is affine in and hence amenable to linear regression analysis. The specification of

the foreign­demand­for­domestic­currency function ( ) would be informed by the results in

Section 2 above and the (translation) part of that is not dependent on would be a linear

function of the remaining relevant independent variables. A similar procedure could be used

to construct a corresponding estimable supply equation. Further details on the construction and

estimation of such a system are beyond the scope of this paper and await future research.

In addition to providing necessary structure for the possible estimation of the demand for

and the supply of foreign exchange, the mathematical analysis herein constitutes a pedagogy for

introducing and elaborating upon the peculiar geometry of such markets. An understanding of the

basis for the two equivalent (economic) perspectives on a given market for foreign exchange as

well as the relationship between the shape of supply or demand in one and the shape of demand or

supply in the other would seem to be essential learning outcomes for any course on open­economy

14
macroeconomics or international finance. Pace Machlup (1939) quoted in Section 1, the depth of

understanding in relation to these learning outcomes would depend on the level of the course and

its prerequisites in terms of prior mathematics training.

15
 ¥/$5 S* 45° line
(b)
4
D

2 (a) (c)

(d)
0
0 1 2 3 4 5
z­illions of $s

F IGURE 1. The affine demand case with = 3 9 and = 0 9.

16
 ¥/$
45° line

S
1
D*

0
0 1 2 3
z­illions of $s

F IGURE 2 A . The “augmented power” (linear supply) case with

= 0 , = 0 4 , and = 1 .

17
 ¥/$ S
45° line

D*

0
0 1 2 3
z­illions of $s

F IGURE 2 B . The “augmented power” (nonlinear supply) case with

= 3 , = 0 6 , and = 2 5 .

18
 ¥/$5 45° line
S
4

1
D*
0
0 1 2 3 4 5
z­illions of $s

F IGURE 3. The “depressed quadratic” supply case with = 08

and = 0 2 .

19
 ¥/$ S
45° line

D*

0
0 1 2 3
z­illions of $s

F IGURE 4 A . The “generalized depressed cubic” supply case with

= 1 3 , = 0 42 , and = 1 .

20
 ¥/$
45° line

2
S

1
D*

0
0 1 2 3
z­illions of $s

F IGURE 4 B . The “generalized depressed cubic” (affine) supply

case with = 0 25 , = 0 55 , and = 0 .

21
 ¥/$5 45° line

2 S

1
D*

0
0 1 2 3 4 5
z­illions of $s

F IGURE 5 A . The “augmented logarithmic” (supply) case with

= 1 5 and = 1 35 .

22
 ¥/$5 45° line

2
S*
1
D

0
0 1 2 3 4 5
z­illions of $s

F IGURE 5 B . The “augmented logarithmic” (demand) case with

= 2 5 and = 0 15 .

23
References

Contreras, José N., “An Episode of the Story of the Cubic Equation: The del Ferro­Tartaglia­

Cardano’s Formulas,” Journal of Mathematical Sciences & Mathematics Education, Vol. 10,

No. 2 (September 2015), pp. 24–37.

Gandolfo, Giancarlo, International Finance and Open­Economy Macroeconomics, Berlin:

Springer­Verlag, 2001.

Haberler, Gottfried von, The Theory of International Trade with its Applications to Commercial

Policy, trans. Alfred Stonier and Frederic Benham, William Hodge & Co., 1936.

Haberler, Gottfried, “The Market for Foreign Exchange and the Stability of the Balance of

Payments: A Theoretical Analysis,” Kyklos, Vol. 3, No. 3 (August 1949), pp. 193–218.

Lehtonen, Jussi, “The Lambert W Function in Ecological and Evolutionary Models,” Methods in

Ecology and Evolution, Vol. 7 (2016), pp. 1110–1118.

Levi, Maurice D., International Finance, Fifth Edition, Routledge, 2009.

Machlup, Fritz, “The Theory of Foreign Exchanges,” Economica, New Series, Vol. 6, No. 24

(November 1939), pp. 375–397; reprinted in International Monetary Economics: Collected

Essays by Fritz Machlup, London: George Allen & Unwin Ltd., 1966, Ch. I (pp. 7–50).

Machlup, Fritz, “Elasticity Pessimism in International Trade,” Economia Internazionale, Vol. 3

(1950), pp. 118–137; reprinted in International Monetary Economics: Collected Essays by Fritz

Machlup, London: George Allen & Unwin Ltd., 1966, Ch. II (pp. 51–68).

24
MacKay, Alexander, and Nathan H. Miller, “Estimating Models of Supply and Demand:

Instruments and Covariance Restrictions,” Harvard Business School, Working Paper 19­051

(October 2019).

OpenStax College, Principles of Economics, OpenStax CNX, 24 October 2018,

http://cnx.org/contents/69619d2b­68f0­44b0­b074­[email protected].

Pearce, Douglas K., “Alternative Views of Exchange­Rate Determination,” Economic Review,

Federal Reserve Bank of Kansas City, Vol. 68, No. 1 (February 1983), pp. 16–30.

25