Khrennikova 2016
Khrennikova 2016
Physica A
journal homepage: www.elsevier.com/locate/physa
Q1
Application of quantum master equation for long-term
prognosis of asset-prices
Q2 Polina Khrennikova
University of Leicester, University Road, LE1 7RH, United Kingdom
highlights
• The model is a contribution to the fields of econophysics and behavioral finance.
• Formalism of quantum physics is used to devise a dynamical model of asset price formation.
• The expectations of the agents on the finance market play a key role in this model.
• The notions of entanglement and superposition encode the non-classical uncertainty related to price formation.
• A numerical simulation of the model for two correlated assets is performed.
http://dx.doi.org/10.1016/j.physa.2015.12.135
0378-4371/© 2016 Elsevier B.V. All rights reserved.
2 P. Khrennikova / Physica A xx (xxxx) xxx–xxx
1 1. Introduction
Q3
2 This paper is a contribution to the field of behavioral finance, with an emphasis on mathematical modeling of
3 Q4 irrationality.1 The problem that we search to approach in this study cannot be formulated better than in the following
4 extended citation from Takahashi and Terano, [3]:
5 ‘‘Most conventional financial theories build the models by assuming representative rational investors based on the
6 hypothesis that ‘‘even if each investor has different future prospects, these errors are canceled out as a whole and do not
7 affect the prices’’. However, the participants of real markets only have limited information and they do not necessarily
8 process the obtained information appropriately. There is a good possibility that the market participants have a common
9 bias that deviates from rationality. If this is the case, the deviation from rationality will not be canceled out and will have a
10 large impact on the prices. To unravel the mechanism of price fluctuation in real markets, it is quite important to understand
11 the difference between the decision-making in the real world and the one based on the maximization of expected utility
12 and to clarify how the prices are affected by the deviation from rationality on decision-making’’.
13 We proceed with modeling the formation of such biases2 with the aid of the novel approach based on exploring the
14 formalism of quantum theory (and its methodology) outside of physics. Recently, this formalism was actively applied to
15 a wide range of social science problems, especially in cognitive psychology, behavioral finance and decision making, see
16 Refs. [9–23]. At heart of this formalism is the quantum probability theory, based on the calculus of complex probability
17 amplitudes (or more generally density matrices). This calculus leads to a nonclassical scheme of probability update, different
18 from the classical Bayesian scheme. Thus, the quantum approach is able to capture violations of Bayesian rationality. For
19 example, in models of behavioral economics it is important to explain and model contexts for which the Savage Sure Thing
20 Principle can be violated. Effectively, in a situation where one cannot presume that agents are ‘‘Bayesian rational’’, a deeper
21 type of uncertainty than the uncertainty of the rational agents, reasoning and acting in accord with some probabilistic
22 information, can be considered. In the mathematical formalism of quantum theory such nonclassical uncertainty is
23 represented in the form of superposition of alternatives. Please, see Remark 1 for a more extensive interpretation of the
24 distinction, made between the classical and the quantum approaches to uncertainty.
25 Along with the other interdisciplinary applications, in this work, quantum physics is used as an operational formalism,
26 describing statistical data generated by experiments. Its crucial difference from classical statistical theories is that it ignores
27 all the parameters of the model, besides those directly related to the measurements.3
28 With this in mind, the superposition of state vectors, representing some choice alternatives, is just a mathematical tool
29 to express uncertainty, which is deeper than the uncertainty described by a classical probability. The irrationality of decision
30 making of the market agents that is manifest in the quantum superposition of their beliefs and decisions, is one of the main
31 factors behind the deep quantum-like uncertainty in the formation of asset prices. Such a price formation of an asset is
32 expressed in the form of superposition of the different prices and can yield a price that deviates from the fundamental value
33 predicted by the classical finance theories. Another profoundly quantum information tool, which we use in our modeling is
34 entanglement, i.e., the states’ non-separability. The prices of a group of assets (e.g. in a portfolio) can be correlated with each
35 other and with a ‘‘financial bath’’, by correlations that are stronger than the possible classical probabilistic correlations.
36 We devise a model that dynamically captures the process of asset price formation and can be applied for preparing a long-
37 term financial prognosis. A decision maker can be e.g., an individual financial consultant, trader, financial forecast agency,
38 or a financial trading corporation. In order to develop some future trading strategy the decision maker (we refer to her as a
39 financial expert) would need to come with a prognosis for a group of asset prices. This model can be expanded for forecasting
40 Q5 the movements of different types of financial instruments, e.g., currency exchange rates. We model the process of decision
41 making related to the asset price formation. The resolution of uncertainty in future prices of assets is described with the
42 aid of the theory of open quantum systems [26]: the most powerful and general apparatus to describe adaptive dynamics
43 of a system (in our case assets) interacting with some bath (in this representation, the bath is of a purely informational
44 nature, consisting of the expectations of the agents acting on the finance market, the mass-media news and other signals of
45 financial, social, and political origins that can be regarded as ‘‘events’’ to contribute to price change).
1 The concept of irrationality can be coupled to the notion ‘‘bounded rationality’’ coined by Herbert Simon [1,2] of finance market players. The deviations
from the rational mode of processing the financial information and the implications it has for the formation of asset prices, will be discussed in Section 5.
2 See also Shiller [4] who emphasized that financial markets are sometimes irrational as investors often make their decisions based on personal
preferences and e.g. over-confidence bias. Irrationality of investors is actively explored in the studies related to the emergence of financial bubbles,
e.g., Bailey [5], Kindleberger [6,7]. It is beyond the scope of this work to concentrate on the specific interrelation, irrationality-bubbles. We proceed under
the assumption that the investors may behave irrationally even in the absence of bubble-like events, see agent based model simulation by Ref. [8]. Of course,
we understand well the role of financial bubbles and the subsequent downturns in increasing the interest to behavioral finance. Bubbles call for a behavioral
explanation. The recent financial crisis of 2008 clearly demonstrated that strict opponents of the Efficient Market Hypothesis, claiming that bubbles are
impossible in well-organized markets, where all financial information is publicly available, have neglected some factors behind the real price formation of
financial assets. Behavioral factors seem to be determining for the price evolution even on informationally-efficient markets, see, e.g., Bailey [5].
3 In quantum physics, one even employs the so called no-go theorems for such parameters, for example, see Ref. [24]. However, for our purposes, these
theorems are not of such a value as in physics. Consider, for example, one of the most famous no-go theorems, on Bell’s inequality [25,24]. It states that if
the ‘‘hidden parameters’’ were existing, they would be nonlocal. However, it seems that, e.g., for the financial markets, the issue of nonlocality is not critical.
Definitely, financial correlations are nonlocal, but they are not super-luminary. Another no-go theorem is the Kochen–Specker theorem [24]. It says that if
‘‘hidden parameters’’ were existing, they would be contextual. Indeed, human behavior is intrinsically contextual and contextuality is a key feature of the
quantum formalism, which is resonating with its notion in behavioral studies, cf. with Refs. [16,17].
P. Khrennikova / Physica A xx (xxxx) xxx–xxx 3
The original dynamical equation from the theory of open quantum systems, the quantum master equation, is too complex, 1
and typically physicists use its (quantum) Markov approximation [26]. We also work with this quantum Markovian dynamics. 2
In physics the process of the resolution of uncertainty of the superposition–entanglement type leading to the equilibrium 3
state is called decoherence. We emphasize that the process of decoherence of the asset price states into their equilibrium 4
price state is essential in the presented model. If a devised model would not involve decoherence, then a superposition of 5
various asset prices would fluctuate and the decision maker would not be able to make the concrete investment decision. 6
We can draw a comparison with quantum physics to elucidate the importance of decoherence in this setting. Decoherence 7
is one of natural scenarios of the determination of the concrete value of a quantum observable, in the so called decoherence 8
model of measurement, Ref. [27]. Another approach to quantum measurement is based on the notion of state reduction 9
(collapse). In cognitive modeling the latter approach, i.e., instantaneous decision making, does not correspond well to the 10
real nature of cognitive processes. The main feature of the cognitive processes, including decision making, is that they have 11
In the setting of this work the reader will find a financial application of a general theory of decision as decoherence 13
developed in a series of papers of Asano et al., e.g., Refs. [9,10], who modeled, e.g., irrationality of behavior of players in 14
games of the Prisoners Dilemma type. This approach, based on theory of open quantum systems, is a part of the generalized 15
During the recent years, quantum probability and decision making were successfully applied to describe a variety of 17
problems, paradoxes, and probability judgment fallacies, e.g., conjunction and disjunction effects, Allais paradox (violations 18
of von Neumann–Morgenstern expected utility axioms), Ellsberg paradox (violation of Aumann–Savage subjective utility 19
In the standard Capital Asset Pricing Model (CAPM) the equilibrium asset price is derived on the assumptions of rational 21
investors [28,29] and the mean variance efficiency of the market. However, this model as well the traditional financial 22
theories (among them various extensions of CAPM, with additional explanatory loading factors, Refs. [30,31]) have been 23
criticized in terms of their explanatory power and the validity of their assumptions [3]: Even when CAPM was suggested, 24
Simon already pointed that the rationality of human being is bounded [1]. However, an asset price in financial markets is 25
considered to be priced based on the fundamental value for the reasons that ‘‘irrational investors behave random and their 26
behaviors are canceled out’’ and ‘‘arbitrage transactions adjust the market price to the fundamental value’’. 27
At the same time, Ref. [3] advocates that the assumptions of traditional financial theories have been questioned by the 28
behavioral finance school, by pointing that ‘‘there are good reasons to believe that the behavior of irrational investors is biased’’ 29
and ‘‘the capability of arbitrage is limited’’. Our approach based on the quantum adaptive dynamics, representing adaptivity 30
of prices to beliefs and expectations of investors can be considered as an extension of CAPM giving a possibility to include 31
irrationality and biases into the time dynamics of the asset prices, approaching the state of an equilibrium.4 The ideology 32
of our model is similar to the notion of ‘‘noise traders’’, applied as a point of reference in the so called ‘‘Behavioral CAPM’’, 33
Ref. [32]. We emphasize that we search to depict the dynamics of the asset prices, not as reflections of some objective 34
factors (e.g. as P/B ratio or firm size), but as the result of the beliefs of investors about the value of the asset that can deviate 35
from its fundamental value. The traders can be categorized as bounded rational, as a result of their non-Bayesian mode of 36
processing the financial information, as well as other factors, such as reluctance to perform such an analysis and instead the 37
usage of heuristics, by coping the behavior of other traders (aka herding bias). Further, the investors are not perceiving the 38
systematic risk of the market in a rational way and thus act over-confident and over-optimistic, or vice versa. The complexity 39
of information and a lack of transparency cannot be neglected either. It has been recognized that the financial markets can 40
periodically (or constantly, depending on the type of market) deviate from all forms of Market Efficiency, in particular the 41
semi-strong and the strong forms, see a comprehensive account in Ref. [33]. All the aforementioned factors contribute to the 42
formation of a special system of beliefs of the investors that shape their investment patterns. In this work we go beyond the 43
classical Markovian models of investors probabilistic beliefs’ evolution and appeal to the quantum formalism, to capture the 44
non-classical processing of information. The beliefs of the investors are in a sense creating the price of the asset that often 45
deviates from the mean–variance efficiency standards of classical financial models. We illustrate our model of preparation 46
of a prognosis (by taking into account the ‘‘financial bath’’) by a numerical simulation. This is an illustrative example for 47
the prognosis of the dynamics of prices of two correlated assets that is adaptive, with respect to the financial bath. Here we 48
explore a mathematical construction borrowed from quantum physics: the Markov dynamics of systems of ions, interacting 49
In general, applying the mathematical apparatus of quantum theory outside of physics brings one to a very advantageous 51
position, in terms of being able to utilize the advanced tools and models developed in statistical and quantum physics over 52
the last hundred of years, by endowing them with novel interpretations; in our case the financial one. This is a general idea Q6 53
of econophysics, Ref. [34], where, for example, the methods of classical statistical physics were successfully explored in 54
economics and finance over the last decades. This paper can be considered as a contribution to the rapidly expanding area 55
of econophysics, quantum finance and quantum economics see, e.g., Refs. [35–47]. 56
In line with these contributions to finance, we search to unify the methods of (quantum) econophysics with dynamical 57
4 This final price realization state is not changed by further inputs from the financial environment.
4 P. Khrennikova / Physica A xx (xxxx) xxx–xxx
1 Finally, we remark that the proposed formalism is based on a financial analogue of a Fock space, where assets’ prices
2 play the role of numbers of quantum particles with specific states. Such a space is infinite dimensional, i.e., in the theoretical
3 model prices can grow up to infinity. The application of the infinite dimensional space is merely a computational tool thus
4 Q7 allowing for a mathematical convenience. It is always possible to consider finite dimensional subspaces of the financial Fock
5 space as, e.g., we do in our numerical simulation, in Section 4. The appropriate state space can be derived empirically, by
6 analyzing the historical price statistics of the particular assets.
8 In this representation, we proceed with a discrete price model. It is assumed that there is chosen some monetary unit,
9 say dollar, and all prices are expressed in terms of this unit, i.e., they are given by natural numbers, n = 1, 2, . . . . If the
10 price of a financial asset is less than one dollar, then it is identified with the discrete level n = 0, i.e., in this model the price
11 level n = 0 does not correspond to complete collapse of a financial asset in terms of its value. We emphasize that the model
12 allows for a long term prognosis (at some future point in time) of the dynamics of N financial assets Aj : j = 1, 2, . . . , N.
14 In the real finance market, at each moment of time, each single asset Aj has the definite price pj . This quantity can be
15 treated as classical (as any result of measurement in quantum theory). However, in the process of a prognosis of asset’s price
16 this assumption, about the fixed (‘‘objective’’) price of an asset, is too constrained. Even when one considers classical finance
17 theories, an asset’s price is modeled probabilistically, e.g., with the aid of theory of stochastic processes [48]. The quantum
18 approach gives a possibility to model a deeper type of uncertainty, associated with the asset price determination. Such type
19 of an intrinsic uncertainty of the actual price is represented by superposition of the fixed-price states.5
20 The state space of a single financial asset Aj is infinite dimensional Hilbert space Hj with the basis |nj ⟩, nj = 0, 1, 2, . . . .
21 As was pointed out, besides the fixed price states given by the basis vectors |nj ⟩, this space contains superpositions of these
22 states of the form:
23 |ψ⟩ = cnj |nj ⟩, (1)
nj
|cnj |2 = 1. Here |cnj |2 give the probabilities Pnj that the jth asset has the price nj .
24 where cnj are complex numbers, nj
iθ
25 In such a quantum structure the complex numbers have not only amplitudes, but also phases, cnj = |cnj |e nj . In the
26 quantum formalism the phases, more precisely the relative phase θnj − θnj , play an important role. The presence of
1 2
27 relative phases contribute nontrivially to the state dynamics, that can be either Schrödinger dynamics or dynamics based
28 on the quantum master equation. The latter is taking into account interaction with a ‘‘market expectations environment’’
29 representing the expectations of agents acting at the market, see Sections 3.2 and 3.3 for further consideration. In this sense
30 such a quantum dynamics differs crucially from a classical Markovian state dynamics, which takes into account only the
31 probabilities Pnj .
33 As in standard QM representation, the complete price configuration of N assets is depicted as the tensor product of the
34 state spaces Hj :
35 H = ⊗Nj=1 Hj .
36 Its vectors have the form:
37 |Ψ ⟩ = Cn |n⟩, (2)
n
39 This state space represents another purely quantum information effect (which is not reduced to states superposition),
40 namely, entanglement: entanglement of the asset-prices for different assets.
41 The states of the space H can be separable and non-separable (entangled). Separable states that can be split into isolated
42 states, each in its respective state space, can be represented as:
43 |ψ⟩ = ⊗Nj=1 |ψj ⟩ = |ψ1 . . . ψN ⟩, (3)
44 where |ψj ⟩ ∈ Hj . (For a moment, we consider only pure states.) The states which cannot be represented in this way are
45 called non-separable, entangled.
5 It important to remark that the classical probabilistic uncertainty of the price values can be in a similar mode modeled with the aid of the quantum
formalism—by using so-called mixed states, statistical mixtures of fixed price states. However, for a moment we prefer to proceed with pure states (although
the model that is applied poses a need to introduce mixed states as well, since the decision making dynamics driven by the quantum master equation
transfers a pure state into a mixed state).
P. Khrennikova / Physica A xx (xxxx) xxx–xxx 5
The degree of entanglement can be quantified with the aid of various measures. Their precise mathematical form is not 1
important for our considerations, see, e.g., Ref. [24] for details. We just point to the existence of a possibility to order these 2
states with the aid of some fixed measure of entanglement. In particular, we can speak about maximally entangled states. For 3
example, consider two assets A1 and A2 and two prices k1 and k2 for A1 and m1 and m2 for A2 . Then so-called Bell’s states 4
√ √
|Φ ⟩ = (|k1 m1 ⟩ + |k2 m2 ⟩)/ 2; |Φ − ⟩ = (|k1 m1 ⟩ − |k2 m2 ⟩)/ 2;
+
(4) 5
√ √
|Ψ + ⟩ = (|k1 m2 ⟩ + |k2 m1 ⟩)/ 2; |Ψ − ⟩ = (|k1 m1 ⟩ + |k2 m2 ⟩)/ 2 (5) 6
are entangled. In the case of the infinite dimensional asset-price state space the price states of assets are not maximally 7
entangled. However, if there were only two possible price configurations for each asset, i.e., k1 , k2 and m1 , m2 , respectively, 8
then such states can be maximally entangled, giving rise to Bell states. 9
From the interpretational viewpoint, the notion of entanglement is one of the most complicated and debated themes in 10
quantum physics. One of the features of entanglement (in the framework of our model) is that generally, in the process of 11
preparing a forecast of asset-prices, it is impossible to treat the asset price dynamics isolation. Moreover, it is even impossible 12
to treat the prices and returns of assets as classically correlated, as obtained through the covariance–variance relation. The 13
correlations, which are encoded in an entangled state, are stronger than correlations described by the classical probabilistic 14
Finally, we remark that even separable asset price state carries an essential degree of quantumness (the associated 16
Remark 1 (Classical and Quantum Representation of Uncertainty). Classical uncertainty is mathematically modeled with the 18
aid of probability measures. Quantum uncertainty is mathematically modeled with aid of state superposition. What is the 19
main interpretational difference between these two approaches, especially in applications of quantum-like models to finances? 20
Classical probability measure accounts chances of realization of actual states of the financial market. For example, consider 21
two assets A1 and A2 with possible prices n1 = 0, 1, 2, . . . and n2 = 0, 1, 2 . . . . We can consider some classical prob- 22
ability measure p defined on the configurations space of possible actual states of the market, Ω = {ω = (n1 n2 )}. Here 23
p(ω) quantifies the chance of realization of the actual price configuration ω. Consider now some quantum superposition, 24
e.g., c1 |k1 m1 ⟩ + c2 |k2 m2 ⟩. This state encodes expert’s beliefs in chances of realization of the actual states of the financial 25
market. In her belief state space both price configurations (k1 , m1 ) and (k2 , m2 ) can coexist peacefully. A the same the clas- 26
sical probability measure p encodes the ‘‘either-or’’ situation. This difference between classical and quantum uncertainties 27
motivates for the use of the terminology ‘‘deeper uncertainty’’ for quantum uncertainty, as an uncertainty about uncertainty. 28
An expert, decision maker, who searches to come with a long term prognosis of prices for multi-asset configurations, 30
i.e., with probabilities that the price configuration of assets will be characterized by the vector of prices |n⟩ = |n1 . . . nN ⟩. As 31
such her aim is to resolve the uncertainty encoded in a superposition (2). In financial terminology, the expert can analyze 32
the output of the model, which provides the expert with the following information: the time of asset price stabilization and 33
the associated probabilities of the possible future prices. Naturally, this data enables the expert to establish the future price 34
configuration of an asset or a series of assets at a given time point. Technically, the time of asset price fluctuations in the 35
proposed model is going to infinity. The fluctuations of the asset prices are damped under the pressure of ‘‘expectations 36
bath’’, but the small price fluctuations still persist. This property of the model demands the expert to fix the exact timing 37
of the equilibrium of state prices fluctuations in an artificial manner, by setting some fluctuation threshold. We discuss the 38
In the quantum formalism the dynamics of the state of an isolated quantum system is described by the Schrödinger’s 40
equation and the dynamics of the state of a quantum system interacting with an environment is described by the quantum 41
master equation [26]. In general the latter equation is too complicated mathematically to treat it straightforwardly 42
i.e., without using some approximations. Therefore, various approximations are used. The most popular is the quantum 43
Markovian approximation leading to the Gorini–Kossakowski–Sudarshan–Lindblad (GKSL) equation [26]. We will discuss its 44
By applying the mathematical formalism of quantum physics to the problem of the asset-price prognosis one has to 47
specify the notions of ‘‘isolation’’ and ‘‘environment’’. We start with the notion of environment. We point out that our 48
approach to asset pricing is concerned with the information dynamics, the information about the possible asset-prices in 49
a future point in time. For this specific purpose, it is natural to identify the environment, with the ‘‘bath’’ of expectations of 50
the agents on the finance market. The notion of ‘‘isolated dynamics’’ corresponds to the ignorance of the impact of such a bath 51
and the assets’ price interaction. Effectively, in physics the notion of an isolated quantum system is merely an abstraction. 52
In reality no completely isolated systems would exist. If we consider quantum field theory, even the vacuum has to be 53
treated as an environment generating an influence on the photons. However, in some contexts one can ignore the impact of 54
6 P. Khrennikova / Physica A xx (xxxx) xxx–xxx
1 the environment with some degree of approximation. By considering the sources of a system’s dynamics, we can split the
2 generator of dynamics into two parts; the first one corresponding to the ignorance of the environment (this would generate
3 a Schrödinger’s dynamics) and the second generator is encoding the impact of the ‘‘market expectations environment’’.
4 In the model under construction, a decision maker working on the long term prognosis of the asset price dynamics of a
5 group of financial assets, analyzes the expectations of the finance market participants. Since the expectations of the agents
6 often create market prices that do not correspond to the real values of the assets (and even if these values coincide at a
7 given point in time, the real asset prices, as understood by the classical finance theory, are not sustainable over time) this
8 model can be considered as a part of the field of behavioral finance and economics. The analysis of investors’ expectations
9 about the market prices, rather than the analysis of the company’s fundamentals, plays a key role in the generation of the
10 probabilistic prognosis of the future prices in this setting.
11 3.2. Prognosis of the asset price dynamics without the expectations of investors
12 In the process of asset pricing, the decision maker surely analyzes the price dynamics of each single financial asset as well
13 as their mutual interactions. However, typically such an analysis is not sufficient to come to the concrete financial prognosis,
14 since the internal characteristics of the assets do not encode all the factors impacting upon its market price. This situation
15 can be mathematically expressed as fluctuations of the solutions of Schrödinger’s equation:
γ dΨ
16 (t ) = H Ψ (t ), (6)
i dt
17 where H is a Hermitian operator acting in H and generating the state dynamics, without accounting for the impact of an
18 environment and γ > 0 is a factor determining the time scale of the dynamics (if the H is chosen as a dimensionless
19 quantity).
20 Remark 1 (On the Meaning of the Scaling Factor γ ). In quantum mechanics the Hamiltonian, H has the dimension of energy
21 and γ = h̄ is the reduced Planck constant h̄ = h/2π . It has the dimension of action = energy × time. One may attempt to
22 proceed in the same way by inventing the notion of the ‘‘energy of market expectations’’ which would be a quite natural
23 approach. However, the main problem of this approach is the development of the measurement methodology for such a kind
24 of the ‘‘mental energy’’. This is a complicated problem related to the interdisciplinary application of physical models. We
25 postpone a more broad discussion to future publications. Nevertheless, heuristically it is convenient to refer to the ‘‘energy of
26 agents’ expectations’’ to couple the abstract quantum formalism with the process of investment decision making in finance.
27 In the absence of ‘‘interactions’’ between assets prices the Schrödinger’s dynamics is described by the Hamiltonian of the
28 form
30 where H0j : Hj → Hj , j = 1, 2, are Hamiltonians generating the price dynamics of different assets. These free Hamiltonians
31 are typically defined with the aid of ‘‘number operators’’
34 H0j = Ej Nj , (9)
35 where the parameter Ej plays the role of the discussed above ‘‘energy of expectations’’ of the agents on the financial market
36 for the price pj = nj of the financial asset Aj . In the state space of jth asset, such free Hamiltonian generates the dynamics
37 of the form
where Ψ0 = |c0nj |2 = 1 is the initial state. From this representation we can see that the solutions of
39
nj c0nj |nj ⟩, nj
40 Schrödinger’s equation different from the stationary ones will fluctuate forever, i.e. no stabilization to the definite state
41 will take place. This feature of Schrödinger’s dynamics does not allow for its application in the process of preparation of
42 a financial prognosis. Moreover, such a fluctuating dynamics cannot resolve the deep quantum (‘‘intrinsic’’) uncertainty
43 encoded in superposition and entanglement of the price states.
44 In the price basis the operator H0 is diagonal. Interaction between asset-prices is described by non-diagonal operators
45 coupling the dynamics of asset-prices. In quantum theory one typically considers bipartite interactions (although in some
46 problems, especially in quantum field theory, more complicated interactions are also explored).
47 As was already pointed out, the quantum state dynamics in the presence of non-negligible influence of a bath cannot be
48 formulated in terms of pure states represented by normalized vectors in the complex Hilbert space. The characteristics of
P. Khrennikova / Physica A xx (xxxx) xxx–xxx 7
the finance market demand to introduce an environment based dynamics. One has to reformulate Schrödinger’s dynamics 1
dρ i
(t ) = − [H , ρ(t )]. (11) 3
dt γ
In the next section we will consider a modification of Eq. (11) in the presence of a financial environment. 4
One of the main distinguishing features of the solutions of the Markovian quantum master equation is that, for a wide 6
class of equations, a nonstationary solution ρ(t ) stabilizes to a stationary solution ρprog representing the long term prognosis 7
As was already emphasized, opposite to Schrödinger’s equation, the quantum master equation can transform pure states 9
into mixed states. Such a process is called decoherence. This is a dynamical equation in the space of density operators. 10
Therefore the limiting prognosis state ρprog can be a mixed state even if the initial price configuration was a pure state. 11
Such a steady state is (under natural conditions) diagonal in the price basis and hence, it represents the resolution of the 12
quantum-like uncertainties present in superposition and entanglement that are present in the initial state ρ0 . 13
In general ρprog determines only the probabilities for the price values. For example, consider only two assets A1 , A2 . If e.g., 14
then the probability that p1 = n1 , p2 = m1 equals to P and the probability that p1 = n2 , p2 = m2 equals to Q ; the prob- 16
abilities of other price configurations equal to zero. We again remark that the prognosis-states, e.g., (12), are in some sense 17
classical states. Superposition indeterminacy, present in the initial state, say ρ0 is one of Bell’s states (4), disappears in the 18
process of decoherence. This is a most typical scenario of the states’ evolution, driven by quantum Markov master equation. 19
Let us formulate the Markovian approximation of the quantum master equation, the Gorini–Kossakowski–Sudarshan– 20
dρ i
(t ) = − [H , ρ(t )] + L(ρ(t )), (13) 22
dt γ
where H is a Hermitian operator acting in H and L is a linear operator acting in the space of linear operators B(H ) in 23
H (such maps are often called super-operators). Typically the operator H represents the state dynamics in the absence 24
of environment, cf. with von Neumann equation (11). However, H can also contain a contribution of the impact of the 25
environment. The superoperator L has to map density operators into density operators, i.e., it has to preserve ‘‘Hermitianity’’, 26
positive definiteness, the trace. These conditions constrain essentially the class of possible generators L. By adding some 27
additional conditions, the so called complete positive definiteness, one can describe the class of generators precisely, see, e.g., 28
1
Lρ = αk [Ck ρ Ck∗ − (Ck∗ Ck ρ + ρ Ck∗ Ck )/2] = αk Ck ρ Ck∗ − {Ck∗ Ck , ρ} . (14) 30
k k
2
In the above devised mathematical model the equilibrium state, is determined as t approaches infinity. Of course, in 32
the setting of a financial market the investment decisions have to be taken during a finite time interval. It is natural to 33
assume that a decision maker terminates the process by setting the threshold for state-fluctuations, quantified by some 34
small parameter ϵ . Thus, when the magnitude of the price state fluctuations becomes less than the ϵ , the asset price state is 35
Consider two assets, A1 , A2 , with possible prices n1 = 1, 2 and n2 = 1, 2. Thus each asset has one qubit state space 38
with the basis |1⟩, |2⟩; the pair of assets has the four dimensional state space C4 with the basis e1 = |11⟩, e2 = |12⟩, e3 = 39
|21⟩, e4 = |22⟩. 40
For the numerical simulation we explore the mathematical model from quantum physics—we refer to it as a three level 41
system of ions, Ref. [49]: two ions interacting with each other, the strength of interaction is represented by the parameter V . 42
There is given a collection of such systems, prepared in the ground state e1 , the vacuum state (in physics, it is denoted 43
|00⟩, but we do not want to operate with assets of a zero price). These systems of ions are exposed to the light pump 44
of specially selected frequency to move ions from the ground state to one of the states e2 , e3 . The intensity of the light 45
pump is represented by the parameter A. The state e4 is the so called dark level, i.e., it is unapproachable from the ‘‘light 46
levels’’ e1 , e2 , e3 . This characteristic reduces the dimension of the model’s state space from four to three, as such, one can 47
operate with the three dimensional complex Hilbert space C3 . In the absence of an interaction with a bath, the population 48
8 P. Khrennikova / Physica A xx (xxxx) xxx–xxx
1 of these levels fluctuates. By playing with the frequency of the radiation, the ions’ energy levels and the parameters V and
2 A, responsible for the strength of interaction, it is possible to move all the ion systems from the ground state e1 to e2 or
3 e3 and vice versa. Again, by adjusting the parameters, it is possible to realize a transition from e1 → e2 much easier than
4 from e1 → e3 . However, these transitions are connected with each other i.e., it is impossible to approach high intensity of
5 e1 → e2 transition without a nontrivial, although weak intensity transition from e1 → e3 . By using so called pseudo-spin
6 formalism [49] it can be done with a following Hamiltonian6 :
V A 0
7 H= A 0 A . (15)
0 A V
8 We emphasize that the quantum physical process under consideration is extremely complex. However, the operational
9 description is reduced to the matrix (15).
10 Now we remark that in physics it is impossible to create an isolated system interacting with a pump of the fixed frequency
11 radiation. There is always present a sufficiently strong noise radiation. The presence of this background affects crucially the
12 behavior of these systems of ions. Their states fluctuations are suppressed and the elements of density matrix stabilize
13 to some fixed numbers. In the model under consideration, the off-diagonal elements representing an interference between
14 states ei and ej , i ̸= j, will vanish (in the limit) and the diagonal ones will approach some fixed probabilities. Again, physicists
15 would apply an operational strategy, not searching to describe explicitly the quantum stochastic process of the bath, but
16 instead to apply a phenomenological (super-)operator L in the GKSL-equation (13). The pseudo-spin formalism is again in
17 use7 [49], where the bath is operationally described with the aid of the following operator:
1 0 0
18 Jz = 0 0 0 . (16)
0 0 −1
19 In physics [24] this is the operator of the spin’s projection on the z-axis. Using Jz we introduce the following operator in the
20 space of density operators in H = C3 :
22 The simplest phenomenological quantum (super-)operator representing interaction with the bath can be selected in the
23 form:
25 (It can be shown that it can be also represented in the canonical form (14).) Finally, there is the coupling constant g > 0
26 describing the strength of interaction of the systems of ions with the bath. The operator L was constructed in such a way
27 that the solution of the corresponding GKSL-equation
dρ i
28 (t ) = − [H , ρ(t )] − gL(ρ(t )) (19)
dt γ
29 exhibits stronger decoherence for larger interaction constants g.
30 We explore how this model works in finance context, by mapping assets’ states onto ions’ states. Suppose that an expert
31 preparing a financial prognosis decides to start with a pair of some low-priced shares A1 , A2 . We remind that in this section
32 the minimal price n = 1; so the shares with prices n1 = 1, n2 = 1 are selected. In the model, the initial state is given
33 by the vector |ψ0 ⟩ = |11⟩, or in the density operator of a form by ρ0 = |ψ0 ⟩⟨ψ0 | (the financial analogue of the ground
34 state in physics). We assume that the assets under consideration are correlated and the parameter V gives the intensity of
35 correlations. We also assume that at the market there is present an ‘‘information pump’’ towards these assets leading to
36 high intensity transition e1 → e2 and essentially lower intensity transition e1 → e3 (heating of these pair of assets). We
37 also assume that the financial context under consideration is such that these transitions are coupled, i.e., it is impossible to
38 approach high intensity of e1 → e2 transition without nontrivial, although weak intensity transition e1 → e3 . The intensity
39 of the information pump towards the pair of assets A1 , A2 is given by the parameter A. We ask the reader for understanding:
40 for a moment, our aim is mainly to illustrate our general model by relatively simple numerical examples. We are not yet able
41 to establish the explicit correspondence with financial engineering, e.g., to specify the real financial technique of ‘‘financial
42 pumping’’ into the pair of the correlated assets.
6 By proceeding operationally we are not interested in the details of the physical considerations of this formalism. We just use the knowledge, see
Ref. [49], that this operator generates the aforementioned transitions.
7 It was found that the spin operators can operationally represent the basic features of the radiation bath. This is a purely formal mathematical modeling
approach; in fact the stochastic features of the bath do not correspond to the real spins. Therefore, the terminology ‘‘pseudo-spin’’ is applied. Moreover,
the real bath is a physical system of large complexity and with an infinite number of degrees of freedom. The below simple description is a consequence
of a well developed approximation technique, leading to the quantum Markov dynamics (13), see Ref. [26] for details.
P. Khrennikova / Physica A xx (xxxx) xxx–xxx 9
Fig. 1. The dynamics of the matrix element ρ11 (t ) = ⟨ρ(t )e1 |e1 ⟩: Solid line—strong decoherence (g = A/30); dash line weak decoherence (g = A/90);
dotted line—no decoherence (g = 0).
Suppose now that an expert E works on planning of the investment of S dollars into A1 , A2 ; he plans to buy xj entities 1
of the Aj -assent, j = 1, 2. And E uses (maybe unconsciously) the aforementioned model to describe the possible dynamics 2
of a pair of correlated assets. In the absence of the environment, based on the expectations of the agents of the financial 3
market, the financial pumping can periodically lift the state e1 = |11⟩ to the state e2 = |12⟩. By estimating the period of 4
oscillations, E would be able to make a higher profit by investing money solely in A2 and selling it at the instant of time of 5
approaching the latter state. However, the model is complex enough and it is not easy to find explicitly this exact period in 6
time. Of course, the real financial market has to be modeled by taking into account the interaction with the bath. It seems 7
impossible to solve Eq. (19) analytically. We performed simulation for the parameters V = 10 and A = V /15 for a few 8
values of the coupling constant g, see Fig. 1. Initially ρ11 (0) = 1. We can see that ρ11 (t ) → 0.5, t → ∞; we also found that 9
ρ22 (t ) → 0.4, ρ33 (t ) → 0.1, ρij (t ) → 0, i ̸= j. Thus, the expectations of E about the possible state of the financial market 10
(reduced to a pair of assets in our numerical simulation) are partially transferred from e1 to (mainly) e2 and (essentially less) 11
to e3 .8 12
0.5 0 0
ρprog = 0 0.4 0 . (20) 14
0 0 0.1
We remind that in this equilibrium state the quantum uncertainty has vanished completely. The final stabilized state can 15
be treated as representing the classical probability distribution associated with assets’ prices.9 16
We introduce the portfolio operator (here x1 and x2 are amounts of money which are initially put into A1 and A2 , 17
respectively): 18
A(x1 , x2 ) = (ix1 + jx2 )|ij⟩⟨ij|. (21) 19
ij
8 The three-level system of ions that we apply in our model has a unique steady state, i.e., by starting with any initial state ρ , its dynamics leads to
0
the same limiting steady state. In particular, if we start with a pure state, which is a superposition of basis vectors with some phases, the final result of
observation will not depend on these phases. At the same time, we should note that the dynamics of the system at the initial stage of its evolution depends
nontrivially on the phases. In the process of the model application, the expert is only interested in the final state of asset prices that does not depend on
the initial state and thus does not depend on the phases associated with the initial state.
9 As was pointed out at the very end of the Section 3, an expert in the real financial setting cannot wait infinitely long for the price states to approach a
complete stabilization to the equilibrium state. This situation is illustrated in the process of the presented numerical simulation. Here the time-steps are
discrete, t1 , t2 , . . . , tn , . . . To stop the calculations, we estimate the difference between the consecutive values of the state-matrix, until some threshold
is reached. The matrix (20) presents the probabilities for possible price configurations on the market. As such, the output of this matrix is providing the
expert with the probabilistic prognosis of the future price configuration on the finance market.
10 P. Khrennikova / Physica A xx (xxxx) xxx–xxx
1 or
2 ⟨A(x, S − x)⟩ρprog ≡ Trρprog A(x1 , x2 ) = 0.5S + 0.4(2S − x) + 0.1(x + S ) = 1.4S − 0.3x. (25)
5 Hence, by investing solely to A2 one maximizes the profit approaching the value 0.4S.
6 We remark that, although initially the state of a pair of assets A1 , A2 is a basis state e1 = |11⟩, which is characterized
7 by the fixed prices and the output state ρprog is the classical mixture of the basis states—the states with the fixed prices, the
8 adaptive dynamics generates the states ρ(t ), 0 < t < ∞, with non-zero interference coefficients, the off-diagonal terms.
9 These states represent deep uncertainty of the nonclassical type about the possible prices. Moreover, one doest need to start
10 Q8 (by mimicking straightforwardly the formalism of physics) from the ‘‘ground state’’ e1 . It is possible to perform a simulation,
11 beginning with some superposition of the fixed price states, ψ0 = ij cij |ij⟩. We aim to proceed with such a development in
12 one of the future works. In such a model, a financial expert E is not planning for an instantaneous investment, but in some
13 near future. Hence, for E , the initial state would be uncertain as well. Such a model advancement would potentially reflect
14 to a better degree the real process of a financial prognosis for the future investments.
15 5. Concluding remarks
16 We remind that classical probability theory (including the Bayesian probability updating scheme and the corresponding
17 modern economic theories of rationality) is fundamentally based on classical Boolean logic. This contribution is concerned
18 with modeling the behavior of decision makers and their subsequent impact on the asset prices. The investors in many
19 contexts elaborate their investment strategies by using a reasoning based on quantum logic, Ref. [50]. The latter essentially
20 relaxes the constraints on the rules of reasoning posed by Boolean logic.10
21 We apply the notion ‘‘irrational’’ in a positive constructive sense, as providing for creative novel solutions, which are
22 impossible in the framework of Boolean logic. In a sense, we can speak of quantum rationality. By using quantum logic,
23 decision makers can proceed with complementary (mutually exclusive) financial contexts and make decisions based on
24 the analysis of such contexts. The Hilbert space formalism provides for the unique possibility to create a unifying mental
25 representation of complementary contexts.11 One of the main features of the theory of open quantum systems [26], which
26 is used to model decision making and more broadly a resolution from uncertainty, about the future asset prices is precisely
27 the possibility to describe context-adaptive dynamics, in the very general mathematical setting.
28 In spite of the aforementioned original features of quantum rationality, it cannot be considered as superior with respect
29 to the classical Boolean (Bayesian) rationality. For example, by getting a novel possibility to unify complementary contexts, a
30 decision maker violates the rules of Boolean reasoning and may generate biases, that are considered as a proof of irrationality
31 from the viewpoint of ‘‘a more careful Boolean analysis’’. The notion of quantum rationality is very close to the notion of
32 bounded rationality, Ref. [1]. A decision maker proceeds using the rules of quantum logic and quantum operational update
33 of events’ probabilities, because such information processing simplifies the reasoning. It allows the decision maker to take
34 short cuts in her decision making process instead of constructing a complete Boolean model. Such type of reasoning emerged
35 in the process of human evolution; in many cases it is more important to proceed faster and use less computational resources
36 than to create a (Boolean) rational model of some situation, cf. Refs. [2,51].
37 As was pointed out in Ref. [5], Sir Isaac Newton said that he can calculate the motion of heavenly bodies, but not the
38 behavior of investors. This statement reflects perfectly the main idea behind our study: behavior of investors is intrinsically
39 indeterministic and it does not match with the laws of classical physics, even statistical physics, based on the theory of
40 classical stochastic processes. On the other hand, it might match the methods of quantum physics and might be properly
41 described on the basis of quantum probability, reflecting a deeper type of uncertainty represented by the superposition and
42 entanglement of investors belief-states that lead to deeper uncertainty associated with asset prices.
43 We hope that this paper and the presented mathematical model of asset price formation, may stimulate the interest to
44 further applications of the mathematical apparatus of quantum theory to the field of behavioral finance and even financial
45 engineering.
46 Uncited references
Q9
47 [52].
10 As was recently emphasized in the talk of K. Gustafson [18], Boolean logic corresponds to the information processing by computers and artificial
intelligence systems, whereas human mind deviates from Boolean reasoning. This claim is based on the results of an extended study of human behavior,
Ref. [11].
11 The analysis of various no-go theorems of quantum mechanics, for instance the theorems elaborating against a possibility to reproduce a quantum
statistical data on the basis of deterministic models, demonstrates that counter-facts play an important role in quantum reasoning.
P. Khrennikova / Physica A xx (xxxx) xxx–xxx 11
References 1
[1] H. Simon, A behavioral model of rational choice, Quart. J. Econ. 69 (1955) 99–118. 2
[2] H. Simon, Bounded rationality and organizational learning, Organ. Sci. 2 (1) (1991) 125–134. 3
[3] H. Takahashi, T. Terano, Agent-based approach to investors’ behavior and asset price fluctuation in financial markets, J. Artif. Soc. Soc. Simul. 6 4
(2003) 3.
[4] R. Shiller, Irrational Exuberance, Princeton University Press, Princeton, 2000. 5
[5] R.E. Bailey, The Economics of Financial Markets, Cambridge University Press, Cambridge, 2005. 6
[6] Ch.P. Kindleberger, Bubbles, in: The New Palgrave Dictionary of Money and Finance, Vol. 1, Macmillan, London, 1992, p. 243. 7
[7] Ch.P. Kindleberger, Manias, Panics, and Crashes, third ed., John Wiley & Sons, Inc., 1996. 8
[8] H. Takahashi, H. Takahashi, T. Terano, Exploring risks of financial markets through agent-based modeling, in: SICE-ICASE, 2006. International Joint 9
Conference, 2006, pp. 939–942. 10
[9] M. Asano, M. Ohya, Y. Tanaka, A. Khrennikov, I. Basieva, On application of Gorini–Kossakowski–Sudarshan–Lindblad equation in cognitive psychology, 11
Open Syst. Inf. Dyn. 18 (2011) 55–69.
[10] M. Asano, I. Basieva, A. Khrennikov, M. Ohya, Y. Tanaka, Quantum-like generalization of the Bayesian updating scheme for objective and subjective 12
mental uncertainties, J. Math. Psych. 56 (3) (2012) 166–175.
[11] J. Bernasconi, K. Gustafson, Contextual quick learning and generalization by human and machine, Netw. Comput. Neural Syst. 9 (1994) 85–106. 13
[12] J.R. Busemeyer, P.D. Bruza, Quantum Models of Cognition and Decision, Cambridge Press, 2012. 14
[13] J.R. Busemeyer, Z. Wang, J.T. Townsend, Quantum dynamics of human decision making, J. Math. Psych. 50 (2006) 220–241. 15
[14] J.R. Busemeyer, Z. Wang, A. Lambert-Mogiliansky, Empirical comparison of Markov and quantum models of decision making, J. Math. Psych. 53 (5) 16
(2009) 423–433.
[15] J.R. Busemeyer, E.M. Pothos, R. Franco, J. Trueblood, A quantum theoretical explanation for probability judgment errors, Psychol. Rev. 118 (2011) 17
193–218.
[16] E.N. Dzhafarov, J.V. Kujala, Quantum entanglement and the issue of selective influences in psychology: An overview, Lect. Notes Comput. Sci. 7620 Q10 18
(2012) 184–195.
[17] E.N. Dzhafarov, J.V. Kujala, All-possible-couplings approach to measuring probabilistic context, PLoS One 8 (5) (2013) e61712. 19
[18] K. Gustafson, The importance of imagination (of lack thereof) in artificial, human, quantum cognition and decision making, in: Conference paper: 20
Conference on Quantum Probability and the Mathematical Modeling of decision-making, March 2015, Toronto, Fields Institute, 2015. 21
[19] E. Haven, A. Khrennikov, Quantum mechanics and violation of the sure-thing principle: the use of probability interference and other concepts, J. Math. 22
Psych. 53 (2009) 378–388.
[20] E. Haven, A. Khrennikov, Quantum Social Science, Cambridge Press, 2012. 23
[21] P. Khrennikova, Evolution of quantum-like modeling in decision making processes, AIP Conf. Proc. 1508 (2012) 108. 24
[22] P. Khrennikova, A quantum framework for ‘Sour grapes’ in cognitive dissonance, in: Proceedings of Quantum Interaction-13, in: Lecture Notes in 25
Computer Science, University of Leicester Press, 2013.
[23] E. Pothos, J.R. Busemeyer, A quantum probability explanation for violation of rational decision theory, Proc. R. Soc. B 276 (2009) 2171–2178. 26
[24] G. Jaeger, Quantum Information: An Overview, Springer, Heidelberg, Berlin, New York, 2007. 27
[25] J.S. Bell, Speakable and Unspeakable in Quantum Mechanics, Cambridge Univ. Press, Cambridge, 1987. 28
[26] M. Ohya, I. Volovich, Mathematical Foundations of Quantum Information and Computation and its Applications to Nano–and Bio-Systems, Springer, 29
Heidelberg, Berlin, New York, 2011.
[27] W.H. Zurek, Decoherence, einselection, and the quantum origins of the classical, Rev. Modern Phys. 75 (7) (2003) 715–775. 30
[28] E.F. Fama, Efficient capital markets, a review of theory and empirical work, J. Finance 25 (1970) 383–417. 31
[29] J.E. Ingersoll, Theory of Financial Decision Making, Rowman & Littlefield, 1987. 32
[30] E.F. Fama, K.R. French, Common risk factors in the returns on stocks and bonds, J. Financ. Econ. 33 (3) (1993). Q11 33
[31] E.F. Fama, K.R. French, A five-factor asset pricing model, September 2014, Fama-Miller Working Paper, 2014. Available at SSRN: 34
http://ssrn.com/abstract=2287202. 35
[32] H. Shefrin, M. Statman, Behavioural capital asset pricing theory, J. Finan. Quant. Anal. 29 (3) (1994) 249–323. Q12 36
[33] H. Baker, N.B. Nofsinger, R. Kolb, Behavioural Finance: Investors, Corporation and Markets, John Wiley & Sons, 2010. 37
[34] R.N. Mantegna, H.E. Stanley, Introduction to Econophysics. Correlations and Complexity in Finance, Cambridge University Press, Cambridge, 2007. 38
[35] B.E. Baaquie, Quantum Finance: Path Integrals and Hamiltonians for Options and Interest Rates, Cambridge University Press, Cambridge, 2004. 39
[36] B.E. Baaquie, Interest Rates and Coupon Bonds in Quantum Finance, Cambridge University Press, Cambridge, UK, 2009. 40
[37] O. Choustova, Quantum Bohmian model for financial market, Physica A 374 (2007) 304–314. 41
[38] E. Haven, A discussion on embedding the Black Scholes option pricing model in a quantum physics setting, Physica A 304 (2002) 507–524. 42
[39] E. Haven, Private information and the ‘‘Information Function’’: A survey of possible uses, Theory and Decision 64 (2008) 193–228. 43
[40] E.W. Piotrowski, M. Schroeder, A. Zambrzycka, Quantum extension of European option pricing based on the Ornstein Uhlenbeck process, Physica A 44
368 (2006) 176–186.
[41] E.W. Piotrowski, J. Sladkowski, Quantum solution to the Newcomb’s paradox, Int. J. Quantum Inf. 1 (2003) 395–402. 45
[42] R.J. Hawkins, B.R. Frieden, Fisher information and equilibrium distributions in econophysics, Phys. Lett. A 322 (2004) 126–130. 46
[43] R.J. Hawkins, M. Aoki, B.R. Frieden, Asymmetric information and macroeconomic dynamics, Physica A 389 (2010) 3565–3571. 47
[44] H. Ishio, E. Haven, Information in asset pricing: A wave function approach, Ann. Phys. (Berlin) 18 (1) (2009) 33–44. 48
[45] P. LaMura, Projective expected utility, J. Math. Psych. 53 (5) (2009) 408–414. 49
[46] M. Schaden, Quantum finance: A quantum approach to stock price fluctuations, Physica A 316 (2002) 511. 50
[47] M. Shubik, Quantum economics, uncertainty and the optimal grid size, Econom. Lett. 64 (3) (1999) 277–278. 51
[48] A.N. Shiryaev, Essentials of Stochastic Finance: Facts, Models, Theory, World Scientific Publ., Singapore, 1999. 52
[49] R.H. Dicke, Coherence in spontaneous radiation processes, Phys. Rev. 93 (1954) 99. 53
[50] J. Birkhoff, J. von Neumann, The logic of quantum mechanics, Ann. of Math. 37 (4) (1936) 823–843. 54
[51] G. Gigerenzer, R. Selten, Bounded Rationality: The Adaptive Toolbox, MIT Press, ISBN: 0-262-57164-1, 2002. 55
[52] C. Schinckus, Economic uncertainty and econophysics, Physica A 388 (2009) 4415–4423. 56