=Paper= {{Paper |id=Vol-2416/paper7 |storemode=property |title=Analysis of the credit turnover in the "Retailer-Bank-Insurer" system with variations in market factors |pdfUrl=https://ceur-ws.org/Vol-2416/paper7.pdf |volume=Vol-2416 |authors=Mikhail Geraskin,Olga Kuznetsova }} ==Analysis of the credit turnover in the "Retailer-Bank-Insurer" system with variations in market factors == https://ceur-ws.org/Vol-2416/paper7.pdf

Analysis of the credit turnover in the "Retailer-Bank-
Insurer" system with variations in market factors

M I Geraskin1, O A Kuznetsova1

1
Samara National Research University, Moskovskoye shosse, 34A, Samara, Russia, 443086

e-mail: olga_5@list.ru

Abstract. The article considers the possible states of the "Retailer-Bank-Insurer" system. The
Pareto-efficient agents’ strategies, which maximize the total profit of the system, are described.
The agents’ models parameters influence on the change in the state of the system is analyzed.
Additionally, the influence of the complementarity effect on the change in the retailer’s optimal
sales volume is investigated. The values of the agents’ parameters, which cause the transition
of the system from one state to another are determined.

1. Introduction
The integrated organizational systems in the economy are formed in the event that the buyer's demand
for one product is conditional upon the fact of purchasing another product. A typical example of such
integration is the “retailer-bank-insurer” system, which is formed in the framework of the retailer’s
credit turnover. Hereinafter, the retailer, the bank, and the insurer are called the agents. The demand
for retailer products encourages customers to apply for loans from banks and insurance services to
insurers, and the possibility of obtaining credit resources, in turn, expands the demand for retailer
products [1]. Therefore, the complementary demand leads to the integrated systems emergence [2].
The integrated system is tightly coupled [5], if the utility function (profit) of agent depends on its
action and on the actions of the environment, i.e., other agents [3, 4]. In particular, the agents’ costs of
the “retailer-bank-insurer” system are interdependent (non-separable), since the system stability is
ensured through mutual payments (commissions, discounts, etc.). The system agents revenues can be
interdependent, if the system has a mechanism for the aggregate utility distribution [6, 7]. In this case,
the agents’ utilities can be considered as transferable.
As a result of the agents heterogeneity in economic activity, the problem of coordinating the
agents’ interests in the integration is considered. The agents, whose goods initiate demand for the
goods of other agents, have prevailing economic activity. One of these agents is called the meta-
agent. The meta-agent status is realized in the awareness about other agents’ utility functions and the
utility values. The meta-agent chooses the distribution mechanism of the aggregated system utility. In
the “retailer-bank-insurer” system the retailer is the meta-agent. The Pareto efficient [8] algorithm for
the transferable utility distribution for the tightly coupled system was developed in [9].
In this paper, we use this algorithm to investigate the influence of the market environment
parameters on the states of the “retailer-bank-insurer” system. Two types of states can be implemented
in the system [10]. First, the meta-agent dominates, when the meta-agent’s optimal sales volume is

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019)
Data Science
M I Geraskin, O A Kuznetsova

less than other agents’ optimums; as a result, the environment agents reduce their volumes of sales to
the meta-agent’s volume. Second, the environment dominates, when the meta-agent’s optimum
exceeds other agents’ optimums. Hence, there are two options of coordinating: either the meta-agent
reduces sales to the volume of the environmental optimum level, or the environment increases sales to
the meta-agent’s optimum. The choice of these options is carried out in the system consistently [11,
12], according to the maximum of the total profit criterion. Consequently, the state of the system has a
significant effect on the equilibrium after the agents’ total profits distribution [13].
The subject of the paper is to study the influence of the agents' market demand functions
parameters on the resulting system states and to determine the boundary values of these parameters,
which cause transitions from one state to another.

2. Methods and materials
We consider the "retailer-bank-insurer" system. The utilities of agents are calculated by the following
formulas [3]:
, (1)
where ; is the agent’s profit function; pk(n) is the price of k-th
agent's goods; uk is k-th agent’s integration costs; ρk is k-th agent risk; ak(n), bk(n) are the price function
coefficients for the n-th system state (n = 1 is the integrated system with Q-l* > Ql*, n = 2 is the
integrated system with Q-li* > Ql* > Q-lj*, n = 3 is the integrated system with Q-l* < Ql*); K is the
agents’ set; the meta-agent is denoted by the symbol «l», the environment is denoted by the symbol «–
l», the symbol «*» denotes the optimum of the agent; Qk is k-th agent sales; ρk is direct costs per unit
of k-th agent.
The system states are classified as follows:
 1, если Ql*  Q*l ,

n   2, если Ql*  Q*l , (2)

 l l1 
3, если Q *  Q * , Q * ,
l 2 
* *
where the symbols Ql1, Ql 2 denote the environmental agents of the meta-agent.
The meta-agent is an agent, which profit exceeds the total profit of the environment for the optimal
strategy. The meta-agent is determined by the following criterion:
 
l Ql*   k Qk*  
kK \l
. (3)
The risk costs characterize the share of probable losses of the agent's average revenue from the
price. The risk cost of the retailer is the bank’s arrears on loans for goods sold. The bank’s risk
exposure is the overdue debt on loans issued, which is taken into account by the discount rate. The
insurer's risk costs are payments for insured events, which are taken into account by the probability of
their occurrence. In the case of uk > 0, the integration costs are interpreted as discounts or commissions
in price of agent with greater economic activity. In the case of uk <0 the integration costs represent the
income of the more active agent as a transfer in the form of price premiums or commissions from
other agents for participation in the integrated system.
The agents’ optimums are calculated by the following formulas [3]:

(4)
The complementary demand is taken into account as follows: the meta-agent sales volume through
its demand function coefficient depends on the environment sales volume:

(5)

where αlk is the complement ratio of k-th and l-th goods, αlk(1) > αlk(2) > 0 are constants. The
complementarity effect is expressed as follows: the growth in lending leads to a faster growth in

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019) 51
Data Science
M I Geraskin, O A Kuznetsova

retailer’s turnover under low interest rate (n = 1), and it leads to a slow increase under high interest
rate (n = 2).
The boundary values of the demand functions parameters are determined in the form of the
following conditions:
,
,
where are the values of the demand functions coefficients under which the
transition occurs between the states n = 1, 2 (denoted by the symbol “~”) and n = 2, 3 (denoted by the
symbol “≈”). Because the system is tightly coupled, that is, the profits redistribution between agents is
possible, the equilibrium is selected according to the agents' total profits criterion:
 
Q p  Qkp , k  K  arg max  Q,   k Qk*
Q 0
  
(6)
k K

is the Pareto-efficient (equilibrium) vector of the agents’ actions Qkp ;  is the total
p
where Q
system profit.
We consider the problem of searching for the Pareto-effective strategies (i.e., actions) in the
possible states of the "retailer-bank-insurer" system.

3. Results
Assertion 1. The Pareto-effective strategies of agents are as follows:
  Ql* , n  1,
 
Q p  Qkp , k  K   Q*l , Ql* , n  2,


 l1 
 Q* , Q* , n  3.
l 2 
Proof of assertion 1: for n = 1, by criterion (6) we get
 k 
Q  Q , k  K  arg max l Q  l1 Q
p p
Q0
  *
l     Q  ; because in this case, according to (2),
*
 l1 l 2
*
l 2

Ql*  Q*l , then, accounting (3), we can get Qkp  Ql* , k  K . When n = 2, by criterion (6) we get
Q*l  Qkp  Ql* , k  K . When n = 3, we get Q*l1  Qkp  Q*l 2 , k  K .
In order to simulate the Pareto-effective strategies, we consider the “retailer-bank-insurer” system,
in which the initiator of the coordination is the retailer. The parameters of the system are presented in
Table 1.

Table 1. System parameters.
Parameters Retailer Bank Insurer
ak 49000 0,785 0,35
uk -3230 0,031 0,018
ρk 0 0,001 0,077
0,09 0,19
pk 49000Q1 0,785Q2 0,35Q30,165
ck 1500 0,053 0,03
bk -0,09 -0,19 -0,165

As follows from table 1, the maximum price of the retailer's product and its costs exceed the prices
and costs of other agents, therefore, according to the maximum profit criterion, the retailer acts as a the
meta-agent. The retailer receives additional income in the integration, because the integration costs of
the retailer are negative.
In Figure 1, agent the profits dependence on sales volumes are plotted. It should be noted that due
to the large difference in the prices of agents, the realtor's profit is estimated on the right scale, and the
bank's profit and the insurer's profit are presented on the left scale.

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019) 52
Data Science
M I Geraskin, O A Kuznetsova

Figure 1. Dependence of agent profits on sale volumes.

In Figure 1, the following notation is used: π1(b1 = -0,1) is the retailer's profit, when its optimum is
less than that of other agents; π1(b1 = -0,09) is the retailer's profit, when its optimum greater than that of
other agents; π2 is the bank’s profit, π3 is the insurer's profit.
The criterion for solving the problem is determined as the total profit of the system. In Figure 2,
the total profit coincides with the retailer's profit due to its significant superiority over other agents.
Thus, the total profit of the system is approximately equal to the retailer's profit. Consequently, the
optimal sales volume of the system is determined by the retailer’s model.

Figure 2. Dependencies of agent profits and total system profits.

In Figure 2, the following notation is used: π1(b1 = -0,1) is the retailer's profit, when its optimum less
than that of other agents; π1(b1 = -0,09) is the retailer's profit, when its greater than that of other agents; π2
is the bank’s profit, π3 is the insurer's profit, πΣ1(b1 = -0,0915) is the total profit of non-integrated agents,
when the optimal retailer’s volume is greater than the optimal volumes of other agents, πΣ1(b1 = -0,096) s
the total profit of non-integrated agents, when the optimal retailer’s volume is less than the optimal
volume of other agents.
The changing in the market parameters leads to a variety of system states, which is characterized
by the position of Ql* relative to other agents. In Figure 2, the set of the possible states reflects its
boundary states, when Ql(1)*= Q2* and Ql(1)*= Q3*. Accordingly, it is possible to distinguish three
intervals: Ql*< Q2* < Q3*, Q2* < Ql*< Q3*, Q2*< Q3* < Ql*.
Provided that the agent optimal volume depends on b1, it is possible to determine the boundary
values b1. The situation Ql*= Q2* < Q3* is realized, when b1 = -0,098. The situation Q2*< Q3* = Ql* is
reached, when b1 = -0,098. In the situation (b1 = -0,1) the retailer’s optimal volume Ql(1)* is less than
the optimal volumes of other agents Q2*, Q3*. In the situation (b1 = -0,09) the retailer’s optimal
volume Ql(1)* is greater than the optimal volumes of other agents Q2*, Q3*. Because the retailer's
profit significantly exceeds other agents’ profit, the total profit of agents in a non-integrated system

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019) 53
Data Science
M I Geraskin, O A Kuznetsova

almost coincides with the retailer's profit. Accordingly, it is possible to determine the optimum of the
system as the optimum of the retailer.
We carry out the factor analysis of the market parameters influence a1, b1, a2, b2 on Q1*.

Figure 3. Dependencies of agents’ optimal volumes on market factors a1, b1.

We consider the following initial market environment parameters : a1= 49000, b1= -0,019. When
changing the market parameters a1, b1, the complementarity effect does not arise, because the sales
volumes of other agents do not change. Accordingly, the parameter αlk in formula (5) is not taken into
account. At the same time, it is obvious that an increase in a1 is accompanied by a less sharp change in
the retailer’s optimal volume, than with a change in b1. Thus, an increase in а1 from 0 до 4% provides
that it enters the zone Q2*Q3*. At the same time, as a1 decreases, the retailer’s optimal volume
decreases, and the situation Q1*Q3*. A decreasing in b1 leads to the retailer’s optimum reducing, and a situation Q1*
Q3*. While decreasing in a2 more than 2%, the retailer’s optimal volume is reduced, and a situation
Q1* < Q2* arises.

Table 2. Characteristic intervals.
n=1 n=2 n=3
Q1*< Q2* Q2*< Q1*< Q3* Q1* > Q3*
∆a1(-∞,0%), ∆a1(0%,-4%), ∆a1(-4%,+∞),
∆b1(-∞,3%) ∆b1(3%,4%) ∆b1(4%,+∞)

∆a2(-∞,-2%), ∆a2(-2%,-1%), ∆a2(-1%,+∞),
∆b2(-∞,-1%) ∆b2(-1%,1,6%) ∆b2(1,6%,+∞)

A changing in b2 from -1% to 1,6% provides that the retailer’s optimum is the zone Q2* < Q1* <
Q3*. An increasing in b2 more than 1,6% provides that the retailer’s optimum switches to the zone
Q1*> Q3*. When b2 decreases more than 1%, the retailer’s optimal sales volume decreases and the
situation Q1* < Q2* arises.

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019) 54
Data Science
M I Geraskin, O A Kuznetsova

The situation n = 1 denotes that the retailer’s sales volume is in the interval Q1* < Q2. In this
situation, the retailer’s sales volume is the least in the system, hence, the retailer is interested in
reducing the volume of other agents to Q1.
The situation n = 2 denotes that the retailer’s sales volume is in the interval Q2* < Q1* < Q3*. In
this situation, the retailer’s optimal volume is greater than the bank’s optimum, but less than the
insurer’s optimum. The retailer encourages the bank to increase its sales, and it encourages the insurer
to reduce its sales.
Situation n = 3 denotes that the retailer’s sales volume is in the interval Q1* > Q3*. In this
situation, the optimal sales volume of the retailer is greater than other agents’ optimums, therefore, it
is interested to encourage them to increase the sales.
The emergence of each situations is a result of a change in the parameters of the retailer: the
maximum price of its product and the ratio of sales volume change to the price. Additionally, due to
the complementarity effect, the change in bank’s parameters through the coefficient of the demand
function affects the change in the retailer’s volume. Hence, the retailer’s parameters are determined:
the maximum price and the change in sales volume, when the transition from one situation to another
occurs.

4. Conclusion
The paper considers the “retailer-bank-insurer” system. The system takes into account the integration
costs and the joint influence of agents’ sales volumes on each other. Several frontier positions have
been defined, under which the Pareto-effective system strategy changes.
The zones Q1* < Q2*, Q2* < Q1* < Q3*, Q1* > Q3* are selected. In the first zone, the retailer, in
order to increase its profit, is interested in encouraging the bank and the insurer to reduce their sales
volumes to Q1*. In the second zone, to achieve the same result, the bank must increase its sales
volume, and the insurer must decrease. In the third zone, both the bank and the insurer must increase
their sales to achieve Q1*.
The influence of market factors ak, bk on the optimal sales volume of agents is analyzed. An
increase in these parameters leads to an increase in the optimal sales volumes of agents of the relevant
market. At the same time, the change in the bank’s parameters has an impact on the change in
retailer’s sales volume. The change in the retailer's optimum is more sensitive to changes in the bank’s
parameters.
5. References
[1] Kulikovskikh I M 2017 Anomaly detection in an ecological feature space to improve the accuracy
of human activity identification in buildings Computer Optics 41(1) 126-133 DOI:
10.18287/2412-6179-2017-41-1-126-133
[2] Geraskin M I and Manakhov V V 2015 Optimization of interactions in a multi-agent, tightly
linked "retailer-bank-insurer" system Problemy upravleniya 4 9-18
[3] Continuity and stimulating compatibility in cardinal voting mechanisms URL: https://papyrus.
bib.umontreal.ca (15.07.2017)
[4] Novikov D A 2012 Theory of management of organizational systems
[5] Shao R and Zhou L 2016 Voting and optimal provision of public goods Journal of Social
Economics 134 3-41
[6] Burkov V N, Gorgidze I I and Novikov D A 1997 Models and mechanisms for the distribution of
costs and revenues in a market economy (Moscow: ICS RAS)
[7] Moulin H 2017 Designing a single mechanism Theoretical economics 12(2) 587-619
[8] Ventura A, Kafiero K and Montibeller M 2016 Pareto Efficiency, Coase's Theorem and External
Effects: A Critical Look Journal of Economic Problems 50(3) 872-895
[9] Geraskin M I, Kuznetsova O 2017 Agents’ Interaction algorithm in a strongly coupled system
with a transferable utility CEUR Workshop Proceedings 1904
[10] Geraskin M I 2017 Optimal mechanism for the distribution of the effect in an integrated strongly
coupled system of anonymous agents with a transferable utility Problemy upravleniya 2 27-41
[11] Pattanyk P K, Salles M 2016 Social choice and welfare (Elsevier science)

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019) 55
Data Science
M I Geraskin, O A Kuznetsova

[12] Bondarik V N, Kordin N A 2012 Resource allocation mechanisms based on non-maximum,
symmetric, anonymous voting procedures with delegation Management problems 5 26-32
[13] Korgin N A, Korepanov V О 2016 Effective solution to the problem of resource allocation with
the Groves-Lediar mechanism for portable utility Automation and remote control 77(5) 914-942

V International Conference on "Information Technology and Nanotechnology" (ITNT-2019) 56