Parabola Volume 57, Issue 2 (2021)

The Black-Scholes model in the context of econophysics

Noah Bergam1 and Teodora Kolarov2

1 Abstract
The Nobel Prize-winning Black-Scholes model (BSM) for financial derivatives pricing
is inextricably linked to the study of econophysics, where concepts from statistical
physics are applied to economic systems. The BSM’s connections to this field exist
on various levels, some more coincidental than others. By design, the BSM treats the
dynamics of stock options as Brownian motion, a phenomenon first discovered in 1827
in the context of random particle movement. However, the fact that the Black-Scholes
partial differential equation can be solved using a transformation into the heat equation
is more of a mathematical accident. Likewise, the traditional Black-Scholes model’s
practical inability to predict the behaviour fat-tailed (as opposed to Gaussian) distri-
butions of stock prices may reflect the limitations of the econophysical modelling ap-
proach. This paper reviews these three econophysical connections and explores their
implications as to the merits of the statistical physics perspective in economics writ
large.

2 Introduction
Econophysics is concerned with adapting the mathematical abstractions of physics
to economic models and mindsets. The term itself was coined in 1995 by H. Eugene
Stanley [6], but its relevance extends well before then, arguably back to Adam Smith’s
1776 Wealth of Nations [13] and his discussion of an equilibrium-chasing “Invisible
Hand” [11]. The subject of econophysics has been approached from many different
angles since then, with varying degrees of success. Jan Tinbergen, winner of the Nobel
Memorial Prize in Economic Sciences in 1969, developed the landmark gravitational
model of international trade. Entropy and information theory have been used to anal-
yse day trading [21]. In 2002, Smith and Foley [14] attempted to put utility maximiza-
tion theory in terms of thermodynamic formalisms but their methods were in 2003
refuted by McCauley [11]. The list goes on.
There have been many reputed benefits to the physical perspective in economics,
including an overall paradigm shift from purely deductive, analytical models to com-
putational, agent-based ones [2].
1
Noah Bergam is a freshman at Columbia University in the City of New York.
2
Teodora Kolarov is a freshman at the Georgia Institute of Technology in Atlanta, Georgia.

1
 This paper reviews one of the most influential products of econophysics and math-
ematical economics: the Black-Scholes model for European put-option pricing. There
are three major points of discussion:

• The underlying, physics-related assumptions which help derive the Black-Scholes

partial differential equation (PDE).
• The solution to said PDE which involves a transformation into the classical heat
equation [10].
• The Black-Scholes model’s inability to account for “fat-tailed” distributions which
are often observed in natural group dynamics such as human interaction and
thus illustrate a certain limitation to econophysics [2].

This paper focuses on the statistics and calculus as well as the general econophysi-
cal philosophy which is core to all three observations.

3 Deriving the Black-Scholes model from econophysical

principles

3.1 Background and assumptions

Black-Scholes is a pricing model for financial instruments known as options, which are
essentially contracts to exchange any underlying security (often a stock) at a predeter-
mined price. Call options give an individual the right to buy a given security, while
put options give an individual the right to sell. European call options, with which the
Black-Scholes model was originally concerned, can only be exercised at an expiration
date that is set by the contract (as opposed to American options, which the buyer can
choose when to exercise) [18]. In creating the model in 1973, economists Fischer Black
and Myron Scholes wanted to investigate if investors are truly protected, or have the
ability to be protected, against huge risks from changes in the asset price before expiry.
Their theory of option pricing can be summed up as follows: Buying a certain amount of
the stock is crucial for the long position in the stock to offset the loss in the short call [3].
This is the key proposition in modern hedge funding, which uses this strategy to
limit risks in financial assets. The Black-Scholes PDE demonstrates that investors can
create a riskless portfolio by continuously adjusting the proportions of the short and
long call. In an efficient or frictionless market, any portfolio with a zero market risk
must have an expected rate of return equal to the riskless interest rate [3]. This deriva-
tion establishes an equilibrium condition between the expected return on the option
and the expected return on the stock, with the final condition relying on the premise of
the riskless interest rate [18].
For the Black-Scholes model, we derive a partial differential equation for the price
of an European put option. Here are central assumptions of the derivation (see [16]):

2
 (a) The market has a constant riskless interest rate and constant volatility.

(b) There are no transaction costs for the option.

(c) It is possible to buy or sell any number of stocks, including fractional quantities.

(d) The option is European, meaning it can only be exercised at expiration.

An investor would not be able to exercise the option prior to this date.

(e) The price of the underlying stock follows a stochastic process.

(f) Stocks are non-dividend paying.

(g) There exists approximately instantaneous trading of the underlying asset.

Note the connections to physics. Some are more obvious than others. Assumption
(e) notably leads us to consider financial dynamics as Brownian motion, an idea which
was originally applied to particle movement in fluids. By making stock prices con-
tinuous, assumption (c) offers a useful analogy between price and physical space (the
relevance of this analogy can be witnessed in Section 3.2). Assumption (b) lends itself
well to the idea of frictionlessness in the market.
Due to assumption (e), the modelling of stock option valuation is based on stochas-
tic calculus, or Itô calculus, meaning the motion of the asset price is partially ran-
dom [20]. Many stochastic processes are based on functions which are continuous,
but nowhere differentiable. In other words, stochastic processes make for very jagged-
looking functions. As we will see in equation (2), stochastic functions generally consist
of a non-stochastic component which account for the long-term behaviour of the func-
tion, and a stochastic component which accounts for the sharp motions in the short-
term [20].

3.2 Derivation
We can write the function of option price as V (S, t; σ, µ, E, T, r). The two independent
variables, S and t, make up the asset’s primary dimensions, where S is the asset price
and t is time. In terms of parameters, σ is the annual volatility (assumed to be constant),
µ is the average rate of growth, E is the strike price, T is the maturity time associated
with the option, and r as the interest rate (constant) [16].
Essentially, the function V (S, t) represents the long position, which is positive be-
cause one sells the option, minus the short position, some constant value multiplied by
the value of the asset from shorting ∆ units of the underlying [3]. If ∆ < 0, then we
are in fact buying ∆ amount of the underlying asset [3]. The idea is to offset the risk
of buying the option by buying a certain amount of stock. The Black-Scholes idea is
first to find the proportion ∆ such that the portfolio becomes non-stochastic in practice.
We assume this asset follows the model of a stochastic process, meaning it follows a
log-normal random walk. The log-normal distribution is bound by 0 on the lower side,
making it perfect for modelling asset prices which cannot take negative values [8]. This

3
 Figure 1: A simulation of Brownian motion with slight upward drift [20].

model describes the movements of stock prices as independent and primarily random
events.
Now, we are going to build a portfolio Π of one long option position and a short po-
sition in some quantity ∆ of the underlying asset S. The quantity ∆ will be considered
to be a constant. The value of this portfolio is represented as follows (see [8]):

Π = V (S, t) − ∆S . (1)

We can write the valuation of the asset price in a dt time interval by the following
differential representation (see [19]):

dS = µSdt + σSdX(t)dt , (2)

where µSdt is the non-stochastic component, and the change is going to be propor-
tional on the asset price and time interval. Here, σSdX is the stochastic component,
composed of unknown volatility σ. The stochastic contribution is given by dX, a func-
tion of t.
To explain the function of dX, we consult the Wiener process [16]. A Wiener process
is defined the following way: The change in X on time interval dt can be expressed as
the normal distribution of mean 0 and standard deviation 1 multiplied by the square
root of dt. Symbolically, this is expressed as follows:
√
dX = N (0, 1) dt , dX 2 = dt . (3)

The average of the square of N (0, 1) is 1. The average of the square of dX is pro-
portional to dt. These distinctions will be necessary as we analyze the change in the
portfolio and keep the linear contributions on dt, despite going up to a second order

4
derivative on the change of the asset [16]. Ultimately, we must analyse the change of
the portfolio with respect to the change of the asset on a specific time interval. The
change on the value of the portfolio, by linearity, from t to t + dt is

dΠ = (dV − ∆dS) . (4)

Let V = V (S(t), t), where S satisfies equation (2). Itô’s Lemma [9] tells us that

1 2 2 ∂ 2V
 
∂V ∂V ∂V
dV = µS + + σ S dt + σ dX(t)dt . (5)
∂S ∂t 2 ∂t2 ∂S

The fundamental difference between stochastic calculus and ordinary calculus is that
stochastic calculus allows the derivative to have a random component determined by
a Brownian motion [20]. The derivative of a random variable has both a deterministic
component and a random component, the latter of which is normally distributed. Itô’s
Lemma is essentially the chain rule for stochastic calculus [20]. Changes in a variable
such as stock price involve a deterministic component which is a function of time and
a stochastic component which depends upon a random variable.
From here, we substitute dV into our portfolio equation (4), rearrange, and find that

1 2 2 ∂ 2V
 
∂V ∂V ∂V

dΠ = µS + + σ S dt + σ dX(t) − ∆ µSdt + σSdX(t)dt . (6)
∂S ∂t 2 ∂t2 ∂S

After rearranging variables, we find that

1 2 2 ∂ 2V
     
∂V ∂V ∂V
dΠ = µS −∆ + + σ S dt + σS − ∆ dX . (7)
∂S ∂t 2 ∂t2 ∂S

In simplifying the expression, we would like to eliminate the random term dX. Since
∆ is an arbitrary constant which is meant to fit our needs for equilibrium, we can set
∆ = ∂V∂S
and simplify to
∂V 1 ∂ 2V
dΠ = + σ2S 2 2 . (8)
∂t 2 ∂S
By making this important distinction, we reduce the randomness, or the risk, to 0, a
concept more familiarly known as delta hedging [18]. By definition, setting ∆ = ∂V ∂S
delineated the amount of assets that we need to hold to get a riskless hedge [3]. For
example, at maturity T , V (S, T ) = max(S − E, 0), a boundary condition that will be
elaborated on later. Hence, ∆ = 1 if S > E and ∆ = 0 if S < E. That means we need
to hold 1 stock if S > E. Meanwhile, if S < E, there is no need to hold any stock since
there is already a guaranteed profit [3]. The value of ∆ is therefore of fundamental
importance in both theory and practice.
Our portfolio must also be non-stochastic. Its value has to be the same as it if it were
on a bank account with interest rate r. This is known as the fundamental economic
assumption of no arbitrage [8]. Investing in the portfolio should be no different than
the risk-free alternative. If the change is completely riskless, then it might be the same

5
as the growth we would expect if we put the equivalent amount of cash in a risk-free
interest-bearing account [19].
Thus,
0 = rΠdt − dΠ . (9)
The difference in return should be 0, so our conclusion is supported. Now, we substi-
tute in equation (7) and (1), rearrange, and find:

∂V 1 ∂ 2V ∂V
+ σ 2 S 2 2 = rV − rS . (10)
∂t 2 ∂S ∂S
The left side represents the change in the price of the option V due to time t in-
creasing and the convexity (second derivative of the price of the stocks with respect to
interest rates) of the option’s value relative to the price of the stock [18]. The right side
represents the risk-free return from a long position option and short position with ∂V ∂S
shares of the stock [18].
The differential equation is often represented in terms of Greek letters as follows:

Θ + 1/2σ 2 S 2 Γ = rV − rS∆ . (11)

The key observation of Black and Scholes was that the risk-free return of the portfolio
of long and short position options (right hand side) over any time interval was ex-
pressed as a weighted sum of Θ and Γ. This is known as the “risk neutral argument”.
Meanwhile, the value of ∂V ∂t
is most likely negative because the value of an option de-
creases as it gets closer to its expiration date, and the convexity of the option’s value is
usually positive from the assumption that the overall portfolio will receive gains from
holding the option [19]. Ultimately, the negative from theta and positive from gamma
cancel out, resulting in risk-free returns.
Assuming the options contract gives the investor the right to buy the underlying
at strike price, E at any time up to and including time T (maturity), then we have the
following boundary conditions [19]:

V (0, t) = 0 for all t .

If S is ever zero, then dS, the change in asset price, is also zero, and therefore S can
never change [3]. Since if S = 0 at expiration, then the payoff will be zero, making the
option worthless regardless of time T to maturity.

V (S, t) → S as S → ∞.

As S → ∞, the likelihood of the option to be exercised increases, so the strike price

becomes less and less important [3]. Thus, the value of the option approaches that of
the asset:
V (S, T ) = max(S − E, 0) .
The final condition outlines the payoff at T . This stems directly from the definition of
a call option, which will only be exercised if S > E, or asset price is greater than the
strike price, thereby procuring the gain S − E [8].

6
4 Connections to the heat equation
4.1 Introduction to the heat equation
The one-dimensional partial differential heat equation is as follows:

∂u ∂ 2u
=k 2. (12)
∂t ∂x
where x is position, t is time, and u(x, t) describes the temperature at a point [7]. It is
best to imagine this PDE as acting on a thin, thermally conductive rod with varying
temperature. The PDE describes a fairly simple principle: that the rate of change of
temperature at a given point on the rod is directly proportional to the concavity of
temperature as a function of position. In the context of a continuous, wavelike function
(see Figure 2), this means that peaks and troughs will change temperature most rapidly.

Figure 2: A simulation of the one-dimensional heat equation with initial conditions

u(x, 0) = sin(2πx) − sin(5πx) and u(0, t) = 0. Think of the cross section along the x axis
as denoting the temperature along a rod. Note how temperature diffuses over time,
and how extrema cool most rapidly [15].

Note that the heat equation can be easily adapted to describe thermodynamic be-
haviour in higher dimensions:
∂u
= k∇2 u , (13)
∂t
where u is the heat function, k is a constant, t is time, and the Laplacian operator (∇2 )
essentially denotes the second spatial derivative [12]. The same intuition from before
applies.
The analytical solution to the Black-Scholes equation involves a transformation into
the solvable one-dimensional heat equation. This paper will not provide a solution of

7
the heat equation; we do note, however, that it involves decomposing the function at
hand into an infinite sum of sine waves using the Fourier transformation [7].

4.2 Transforming Black-Scholes into the heat equation

Below is the Black-Scholes PDE in its full glory:

∂V 1 ∂ 2V ∂V
+ σ 2 S 2 2 + rS − rV = 0 . (14)
∂t 2 ∂S ∂S
If we think of V as the analogue to temperature, we can already see structural sim-
ilarities with the heat equation. There is a first derivative of V with respect to time,
and a second derivative of V with respect to stock price (analogous to a spatial deriva-
tive) [10]. However, there are some notable differences, such as the presence of the
∂V
∂s
and rV . We apply transformations in order to get rid of these terms. See the three
subsequent transformations, and their respective purposes, below:

τ = T − t. (15)

This reverses the progression of time, such that we look forward for our V rather than
backwards (since we know the price at maturity, or time T ).

1
x = ln(S) + (r − σ 2 )τ . (16)
2
∂2V ∂V
This serves to remove the S coefficient in front of the ∂S 2
term, and get rid of the ∂S
term altogether.
F = V ert . (17)
This brings us to the final heat equation form. The strategy with each of these
transformations is to look at every variable on the right hand side that shows up in
PDE, compute each necessary partial derivative, substitute, and simplify.
After applying these transformations, we are left with the following equation:

∂F 1 ∂ 2F
= σ2 2 . (18)
∂x 2 ∂x
Behold, the heat equation. This can now be solved using Fourier analysis.

5 “Fat Tails”: A practical limitation

The past two examples of econophysical connections were based on the Black-Scholes
model in theory. This final connection is concerned with a well-established practical
limitation of the model: its inability to handle fat-tailed, or leptokurtic, distributions of
securities pricing.

8
Figure 3: Solution surface of the Black-Scholes equation. Note the similarities to the
heat equation surface [18].

One of the fundamental assumptions behind the Black-Scholes model is the idea
that securities are normally distributed according to price. However, it has been ob-
served that Black-Scholes overprices call options which are “at the money”, such that
S ≈ K. It also underprices call options at either end, whether “in the money”, S ≫ K,
or “out of the money”, S ≪ K. This indicates that securities pricing is not normally
distributed, but rather follows a more fat-tailed distribution, where extreme events are
more likely [5].

Figure 4: A normal distribution (red) and a more fat-tailed Cauchy distribution (blue).

One of the major contributions of econophysics has been the observation of fat-
tailed distributions in financial markets [2]. Note that this does not necessarily demon-
strate that fat tails are a common sight in physics and the natural sciences; if anything,
fat tails are a more common sight when analysing group dynamics such as human be-
haviour. Nassim Nicholas Taleb’s 2007 book The Black Swan [17] discusses this idea
in more depth, arguing that leptokurtic distributions are much better models for ex-
amining unpredictable human phenomena than the traditional bell curve [4]. From
the Black Swan perspective, the traditional Gaussian Black-Scholes model reflects the
inherent limitations of the physics perspective in economics and the social sciences in

9
general.
Note that the Black-Scholes model can be and has been adapted to look at fat-tailed
distributions, such as Levy Processes and power law distributions [1]. Nonetheless, the
question remains: to what extent can a single distribution, no matter how leptokurtic,
capture the chaos of collective human behaviour?

6 Conclusions
Econophysics is a powerful but ostensibly limited enterprise. Its pros and cons shine
perhaps most clearly through the Black-Scholes partial differential equation. In one
sense, it is remarkable how physics-based assumptions of stock motion can yield an
analytic (and coincidentally thermodynamics-related) solution to the future value of
a call option. The trade-off, at least with the vanilla model of Black-Scholes, is that
we are wed to the normal distribution when it appears stock behaviour may follow
something more fat-tailed in practice. The philosophical implication, at least at a first
glance, is that econophysics has met its match, but judging from the history, this field
has a way of adapting.
Adam Smith seemed to believe that the laws that govern economics and the finan-
cial market parallel those which regulate the motion of simple harmonic oscillators and
falling objects [3]. Neither Smith nor his contemporaries anticipated the sheer volatil-
ity of the modern market. Nonetheless, his fledgling concept of econophysics grew
from simple mechanics analogies to those of intricate thermodynamics.
Random walk literature arose in the 20th century, asserting that stock price changes
are random and cannot be predicted. Cue stochastic, or Itô, calculus, and mathemati-
cians had officially devoted an entire study to differentiation and integration of ran-
dom variables [20].
In the most recent applications of econophysics, mathematicians have treated fric-
tionless markets as heat baths. The assumption of adequate liquidity, or “market
depth” is analogous to the assumption of a heat bath in thermodynamics: Removing
a small number of shares from the market does not effect the market price of the asset
traded, just as reversibly removing a small amount of energy from the heat bath does
not affect overall temperature [11]. With these thermodynamic analogies, it is very
possible to draw mathematical relationships with regard to market entropy, which in
turn offers insight into a market’s sense of disorder [11].
On and on, the field continues to grow in unforeseen and decentralized ways – and
yet, on a surface level, there seems to be a certain fundamental absurdity to econo-
physics. Of course, humans and human transactions don’t quite behave as obediently
or predictably as particles. But when we zoom out, when we consider the macroeco-
nomic lens, we observe emergent, mathematical properties. The statistics we use to
analyse the emergence called an economy can sometimes parallel the statistics used
to look at large collections of well-behaved particles. Therein lies the insights, and
potentially the folly, of econophysics, in drawing these parallels and chasing them.

10
Acknowledgements
We would like to thank Dr. Alexandra Lasevich for helping coordinate this research
project, and Brad Poprik and Davidson Barr for their corrections and suggestions.

References
[1] S. Boyarchenko and S.Z. Levendorskii, Non-Gaussian Merton-Black-Scholes The-
ory, World Scientific Publishing, 2002.

[2] M. Buchanan, What has econophysics ever done for us?, Nature Physics 6 (2013),
317.

[3] R. Chan, Y. Guo, S. Lee and X. Li, Financial Mathematics, Derivatives and Struc-
tured Products, Springer Singapore, 2019.

[4] S. Dunbar, Stochastic Processes and Advanced Mathematical Finance - Solution of

the Black-Scholes Equation, 2016, http://mmquant.net/wp-content/uploads/
2016/08/solutionBSE.pdf, last accessed on 2021-08-19.

[5] S. Dunbar, Stochastic Processes and Advanced Mathematical Finance - Limita-

tions of the Black-Scholes Equation, 2016, http://mmquant.net/wp-content/
uploads/2016/08/solutionBSE.pdf, last accessed on 2021-08-19.

[6] K. Gangopadhyay, Interview with Eugene H. Stanley, IIM Kozhikode Society &
Management Review 2 (2013), 73–78.

[7] M. Hancock, The 1-D Heat Equation, 18.303 Linear Partial Differential Equations,
MIT OpenCourseWare, 2006. https://ocw.mit.edu/courses/mathematics/
18-303-linear-partial-differential-equations-fall-2006/
lecture-notes/heateqni.pdf, last accessed on 2021-08-19.

[8] J.C. Hull, Options, Futures, and Other Derivatives, 9th ed., Prentice Hall, Engle-
wood Cliffs, NJ, 2017.

[9] K. Itô, Stochastic Integral, Proc. Imperial Acad. Tokyo 20 (1944), 519–524.

[10] C. Kwan, Solving the Black-Scholes Partial Differential Equation via the Solution
Method for a One-Dimensional Heat Equation: A Pedagogic Approach with a
Spreadsheet-Based Illustration, Spreadsheets in Education 12, 2019.

[11] J. McCauley, Thermodynamic analogies in economics and finance: instability of

markets, Physica A: Statistical Mechanics and its Applications 329 (2003), 199–212.

[12] M. Macauley, Lecture 7.3: The Heat and Wave Equations in Higher Dimensions,
http://www.math.clemson.edu/˜macaule/classes/m18_math4340/slides/
math4340_lecture-7-03_h.pdf, last accessed on 2021-08-19.

11
[13] A. Smith, An Inquiry into the Nature And Causes of the Wealth of Nations,
W. Strahan and T. Cadell, London, 1776.

[14] E. Smith and D. Foley, Is Utility Theory so Different from Thermodynamics?,

Santa Fe Institute Working Papers (2002), Paper #02-04-016.

[15] N. Stanton, Graph of Solution of the Heat Equation,

https://www3.nd.edu/˜nancy/Math30650/Matlab/Demos/demo10-5/
demo10_5.html, last accessed on 2021-08-19.

[16] B. Stehlı́ková, Black-Scholes Model: Derivation and Solution, 2015,

http://www.iam.fmph.uniba.sk/institute/stehlikova/fd14en/
lectures/05 black scholes 1.pdf, last accessed on 2021-08-19.

[17] N.N. Taleb, The Black Swan: The Impact of the Highly Improbable, Random
House, 2007.

[18] J. Veisdal, The Black Scholes formula, explained, Cantor’s Paradise, 2019,
https://www.cantorsparadise.com/the-black-scholes-formula-
explained-9e05b7865d8a, last accessed on 2021-08-19.

[19] R. Walker, An Introduction to the Black Scholes PDE, 2009,

http://www.ms.uky.edu/˜rwalker/research/black-scholes.pdf,
last accessed on 2021-08-19.

[20] W. Whitt, A Quick Introduction to Stochastic Calculus, 2007,

http://www.columbia.edu/˜ww2040/4701Sum07/lec0711.pdf,
last accessed on 2021-08-19.

[21] Wikipedia, Econophysics, https://en.wikipedia.org/wiki/Econophysics,

last accessed on 2021-08-19.

[22] quantpie, Transformation of Black Scholes PDE to Heat Equation, 2019,

https://youtu.be/IgMoOcO095U, last accessed on 2021-08-19.

12