ECONOMETRIC MODELING OF ECONOMIC AND FINANCIAL PROCESSES IN
INFORMATION SOCIETY
Bogomolov A.I, Ph.D., Associate Professor Nevezhin V.P, PhD., Professor Financial University under the Government of the Russian Federation, Moscow
ЭКОНОМЕТРИЧЕСКОЕ МОДЕЛИРОВАНИЕ ЭКОНОМИЧЕСКИХ И ФИНАНСОВЫХ ПРОЦЕССОВ В ИНФОРМАЦИОННОМ ОБЩЕСТВЕ
Богомолов А. И., к. т. н., доцент, Невежин В. П., к. т. н., профессор, Финансовый университет при Правительстве Российской Федерации, г. Москва
Abstract. The article proposes to overcome the shortcomings of classical econometrics, noted economists scientists and practitioners using the capabilities of modern computer and information technology, network approach to the description of models and methodology agent - based modeling. The authors propose to introduce the concept of economic network and the network of natural agent, the latter is a random event whose probability of occurrence can be determined on the basis of belief networks Bayes.
Key words: network economics, agent-based model, the web of trust networks Bayes.
Аннотация. В статье предлагается преодолеть недостатки классической эконометрики, отмеченные учёными экономистами и практиками, используя возможности современных компьютерных и информационных технологий, сетевой подход к описанию моделей и методологию агент - ориентированного моделирования. Авторы предлагают ввести понятие сетевого экономического и сетевого природного агента, причём последний представляет собой случайное событие, вероятность появления которого может быть определена на основании сети доверия Байеса.
Ключевые слова: сетевая экономика, агент-ориентированная модель, сеть доверия Байеса
Economics — the science that studies quantitative and qualitative economic relationships using mathematical and statistical methods and models [1]. The modern definition of the subject of econometrics was developed in the Charter of the Econometric society, which is the main purpose called the use of statistics and mathematics for the development of economic theory. Econometrics is usually considered part of economic theory, along with macro and microeconomics.
However, in our opinion, it has a much greater epistemological value than is given in this definition. The fact is that in econometrics we study the processes and phenomena are essentially probabilistic and random in nature. And the chance and probability lie at the Foundation of our physical world, our existence, of human behavior and economic phenomena and processes. So, in the microcosm the behavior of elementary particles is probabilistic in nature and describes, respectively, quantum or wave mechanics. Events in our world are also largely dependent on "the will of the case". The behavior of the person who received from God the freedom of the will, to a large extent, is unpredictable, spontaneous character. Economic events and processes, even if they can be predicted, always characterized by a measure of uncertainty. And the mission of econometrics as a science, to overcome this fundamental uncertainty to predict the future and to link a chain of events, not only in economy but also in the entire socio-economic sphere, puts it in a special position among all life Sciences companies.
Despite the obvious successes of econometrics to explain and predict the behavior of economic systems and processes, many scholars and practitioners are not satisfied with the results and even consider it a pseudoscience or, at
«Хроноэкономика» № 1 (1) май 2016
least, useless. The most famous and far-sighted economists have long criticized econometrics. For example, the criticism of econometrics the great American economist Keynes due to his refusal "to treat economy as accurate science." In his presentation "Economic environment is volatile and unpredictable, and most economic variables connected by many complex non-linear dependencies. This is followed by the instability of correlation coefficients and the inability to solve predictive tasks. Therefore economic science cannot claim accurate quantitative measurements. It should be based on realistic assumptions and provide the tools that help to understand and explain the environment" [2].
In the early 1970-ies, the famous British economist Warwick sharply criticized economists-mathematicians for the lack of communication with specific facts" [3]. At the same time F. brown [4] claimed that "the construction of time series regressions is only good for cheating". V. Leontiev characterized econometrics as "an attempt to compensate for the conspicuous lack of widely available data by the use of increasingly sophisticated statistical techniques". Strongly opposed to econometrics belonged and representatives of the Austrian school of Economics [5, 6].
In his book "the Black Swan" N.N. Taleb [7] writes "Then I looked at all the scientific work and the dissertation which has managed to unearth. None of them no convincing evidence that economists (as a community) are able to make predictions; and if sometimes capable, then their predictions are only slightly better than random — make them on the basis of serious decisions it is impossible."
However, "if stars light, means, someone needs it". That is, despite the criticism and the obvious failures of
www.hronoeconomics.ru
econometrics, for example, when trying to predict economic crises, its research and application in the world becomes more and more widespread. Econometrics develops, added all new economic theories, methods and tools from systems theory, chaos theory, swarm intelligence, fuzzy logic, hybrid computing, etc. Especially significant changes in performance occur econometric methods and expected applications of new information technologies and their ability to process and analyze huge volumes of data.
Increasing demands for the results of econometric studies and a significant expansion of the range of economic tasks that require econometric methods solution occurs in the conditions of information society, when the "price" of knowledge increases dramatically. In the information society being "accelerated" and solving econometric problems often are required to obtain in real-time.
This fundamental possibility and provide modern information technology.
A major failure of the use of econometrics to explain and predict the behavior of economic systems and management decision-making is "unexpected" events, which affect the characteristics (variables) and parameters of economic objects and make the results of econometric modelling are unsatisfactory.
To overcome the above mentioned disadvantages, in contrast to the classical econometric models, new models should include further factors (variables), which represent random events that affect the factors (variables) of the model and alter their values. The combination of all these variables (including random events) must be represented in the network model, i.e. nodes and links between them.
In the information society the dominant economic and social functions and processes are increasingly organized on the principle of networking. It networks constitute the new social morphology of our societies, and the proliferation of "network" logic significantly affects the course and the results of the processes associated with the production, daily life, culture and power.
One of the network models together of economic actors in the information society is a description of them as economic agents. Economic agents in classical Economics — economic actors involved in the production, distribution, exchange and consumption of economic benefits [8-11].
For example, in a four-sector (open) economy we believe there are four macroeconomic agent:
1. Households — households (individuals and families).
2. Firms are institutions in the form of factories, farms, mines, shops that perform several functions in the production and distribution of goods and services.
3. The state agent, consisting of government agencies, whose task is to regulate the economy.
4. Foreign sector — rest of the state.
Agent-oriented modeling is known in foreign literature
as Agent-Based Modeling (abbr. ABM) [12, 13], the development of which directly from the increasing computational capacities of modern computers, allows one to imagine (to simulate) a system of any complexity from a large number of interacting objects, without resorting to aggregation. Agent-oriented models are used for numerous
«Хроноэкономика» № 1 (1) май 2016
commercial and technological problems. The examples may serve the following tasks:
1) optimization of supply chain and logistics;
2) modeling of consumer behavior (including social networking);
3) distributed computing;
4) the management of labor resources;
5) transportation;
6) management of investment portfolios.
In the General theory of agent-based models agents are understood under the enough abstract entities, granted by certain properties. We offer the concept of economic agents to consider as part of a more General concept of network agents. In the framework of the network econometric model has been proposed to refer to the nodes of the network agents, which may be economic agents (economic actors), and there may be economic variables (explanatory and explained).
In fact, in the two-factor model Cobb-Douglas Y=AKaLP, (1)
exogenous variables, K - capital, L - labor , may well be considered sufficiently independent economic agents, explaining the behavior of endogenous variable (dependent of the undertaking).
In agent-oriented modeling or computer simulation important method of specification for the network of agents. Although this is a separate interesting question, however, for the further development of our ideas about network econometric models offer the color specification for the network of agents. In the graphical display of patterns of network economy, network agents, depending on their nature can be assigned a different color. For example, financial performance can be indicated by the network agents in yellow and industrial -blue, etc. The usefulness of color classification is proved, for example, in theoretical physics, where the quarks have "color". Since color corresponds to a certain wave length, then the number can be an identifier of classification of a specific network agent.
The need for classification of network agent is evident in connection with the use of computer simulation of their interaction. Network agents are located in a common information space, for example, in Internet.
The network agents can influence some events, not necessarily related to economic activity, for example, drought, revolution or technological disaster, etc. For the community will consider them as natural events or natural agents.
The need for classification of network agent is evident in connection with the use of computer simulation of their interaction. Network agents are located in a common information space, for example, in Internet.
The network agents can influence some events, not necessarily related to economic activity, for example, drought, revolution or technological disaster, etc. For the community will consider them as natural events or natural agents. We assume that the occurrence of these events is subject to cause and effect relationships and is characterized by a certain probability. These nodes in our scheme can be denoted, for example, squares are also different colors depending on their essence. To distinguish
www.hronoeconomics.ru
network and natural agents to their indices (classifiers) are added respectively 0 or 1.
Fig. 1. Color (wavelength) as element identification the classification of network agent The need for classification of network agent is evident in connection with the use of computer simulation of their interaction. Network agents are located in a common information space, for example, in Internet.
The network agents can influence some events, not necessarily related to economic activity, for example, drought, revolution or technological disaster, etc. For the community will consider them as natural events or natural agents. We assume that the occurrence of these events is subject to cause and effect relationships and is characterized by a certain probability. These nodes in our scheme can be denoted, for example, squares are also different colors depending on their essence. To distinguish network and natural agents to their indices (classifiers) are added respectively 0 or 1.
Among these nodes can be "centers of power", which manifest themselves giving signals to change the state economic or natural agents. Such "centers of power" at the macro - level can be a major capitalists or their families (the Rockefellers, Rothschilds, etc.), at the state level, clicks, clientele at the level of individual producers and consumers, the clout and L3.
All these agents (network and natural) are signaling, that is, in the information interaction. As a result of receipt of a signal from one agent to another changes the state of the agent, that is, for the network agent, this leads to changes in the values of its characteristics (economic variable), and the natural agent (events) changes the probability of its occurrence. In General, the model of the network economy can be represented as a system of heterogeneous nodes, each of which is painted in its color, and the connecting links (Fig.2).
«Хроноэкономмка» № 1 (1) Mafi 2016
Fig. 2. The network economy model of heterogeneous economic and natural agents (events) Thus, the network model of the economy, at both macro-and micro-level, the network contains heterogeneous nodes (network and natural agents) which exchange signals, enter into economic relations (dependencies) and change their state over time, i.e. take in information interaction. Information interaction is more fundamental level to describe the functioning of socioeconomic systems compared with market and hierarchical.
Consider how to explain the behavior of an economic agent (an endogenous variable) depending on explaining behavior of economic agents (exogenous variables) and the classical econometric network model. Consider the same model with Cobbo-Douglas.
Classical econometrics solves the problem by means of transforming a nonlinear model into a linear model of multiple regression (LMR) to explain the behavior of endogenous variable, depending on the behavior of the explanatory (exogenous) variables, representing the relationship between the economic variables in the form of the regression relationship.
Classical regression dependence can be defined as follows. LetXi,X2, ...,Xn is a random variable with a given joint probability distribution. If for each set of values Xi=xi,X2=x2, ...,Xn=xn is defined conditional expectation (x1,x2, ..., xn) = E(Y/ X1=x1, X2=x2, ..., Xn=xn) (2) the function y(x1,x2, ..., xn) is called regression values X1, X2, ..., Xn, and its graph is the regression line forX1, X2, ..., Xn, or a regression equation values for the variables X1, X2, ..., Xn. This model is a linear model of multiple regression (LMR).
The dependence of X1, X2, ..., Xn is manifested in the change of mean values From when you changeX1, X2, ..., Xn. Although for each fixed set of values X1=x1, X2=x2, ..., Xn=xn the value of U remains a random variable with a certain distribution. Graphic LMR as follows (Fig. 3):
Fig.3. Classic LMR from one explain and two explanatory variables
Collected sample of n observations in the form of table
1.
Table 1. The sample source data
n K I L '|Y "
J__K11 L21 Y1_
2 K12 L22 Y2
_n_ Kin I L2n Yn
Using OLS, estimate the parameters of the model. It is assumed that the simultaneous change of 2 exogenous variables Xi instantly leads to a change U, i.e. the principle www.hronoeconomics.ru
of "long-range, long-range interaction Principle States that if body A, located at point a acts on another body b, then body, located at the point b is experiencing this impact at the same time. Newton considered it necessary to have a transmitter of this action, "agent", however, allowing him to be intangible nature [15]. Thus, in the classical LMR the principle of "long-range", Classic LMR to determine how accurately the regression analysis estimates the change when you change Xi, X2, ..., Xn, is used the average value of dispersion in different sets of values Xi, X2, ..., Xn (in fact, it is a measure of the dispersion of the dependent variable around the regression line).
Thus, it is believed that classical LMR able to predict the average value of the endogenous variable with the accuracy of the average value of the variance based on "historical" data and the assumption that the probabilistic characteristics of the explanatory variables will not change. This model is quite far from reality and does not allow to make predictions and to make informed decisions in a rapidly changing socio-economic sphere. The reason for this lack LMR is that "picture" spoil "unexpected" occurrence of events influencing certain economic variables of the model, as well as random non-simultaneous change of the values of economic variables themselves. Classic LMR at different points in time, its structure does not change, and the econometric model as a result of information interaction, it is constantly changing.
Consider for simplicity the network model of a single endogenous (dependent) and two exogenous (independent) variables. In contrast to the classical LMR is under constant surveillance by agents xi, X2 and y, and recalculate the model parameters based on OLS. This procedure is carried out on the basis of the automated system for the collection and processing of data connected to company network, or a public research center, etc. In a network econometric model should take account of the possibility of a natural agent (a random event dependent on the occurrence of other events) lead to changing the status of a network agent (Fig.6).
Fig. 4. Network model of multiple regression of the 2
economic variables and the 3 events This model relates to the field of statistical information modelling is a graph of a probabilistic and causal relationship between the variables. In the above model we have included Bayesian belief network, suggesting that the change L may be affected by the event A, which is a consequence of the events b and C. for Example, such events can be inflation, trade unions and other public associations (the so-called "third sector" of the economy), economic policy of the state. In Bayesian networks of trust vertices represent random variables and arcs of the probabilistic dependencies that are defined through conditional probability tables. The table of conditional
probabilities of each vertex contains the state probabilities of this node, provided it States "parents".
In Bayesian networks can be combined organically empirical frequency of occurrence of different values of the variables, subjective assessment of "expectations" and theoretical ideas on the mathematical probabilities of certain consequences from a priori information. This is an important practical advantage and distinguishes Bayesian network from other methods of information modeling. In the traditional formulation Bayesian networks are not designed to operate with a continuous set of States (for example, a real number in a given interval). To represent real numbers you can split the range of possible values of the variable L into segments and consider the discrete set of centers in Fig. 3 K, L and Y are economic agents (exogenous and endogenous variables), A, B and C -natural agents (random events). Between all these agents is communication. Understand that there may be another scheme. For example, agent A may act directly on the Y.
Thus, the econometric network model in addition to the classic implies the use of methodologies and tools agent-based models, considering as agents and variables influencing them random events, which can also be modelled on the basis of networks of trust Bayes.
The procedure works with a network econometric model can be represented as a repeating sequence of four steps:
1. The inclusion of a substantial network and natural agents
2. Tracking values of the agents and storing them in the database
3. Estimation of parameters for model information Bayesian belief networks, developing (ready to use) program and the calculation of conditional probabilities and the corresponding values of the natural and network agents.
4. Training program by comparing the results of the actions and expectations and return to the first stage.
Thus, the inclusion of classic econometric models of representations of variables as network agents and the addition of natural agents model (random effects in the model variables) translates into a statistically-information interaction, where the most effective way they can Express themselves Bayesian belief networks.
The present level of development of information and communication technologies allows to track and store data about the state of the network agents and to acquire new knowledge and reliable predictions about the economic processes on the basis of a network of econometric models and automated expert systems.
References
1. V.A. Byshev. Econometrics. M., finances and statistics, 2008, 480 p.
2. I. Rozmainsky. Methodological foundations of the theory of Keynes and his "dispute about method" with Tinbergen (Rus.) M., problems of Economics. 2007, № 4.
3. The crisis of modern economic theory. - URL: http://nanobukva.ru/b/altern/blaug metodologija ehkono
«Хроноэкономмка» № 1 (1) Mafi 2016
www.hronoeconomics.ru
micheskoj nauki_ili_kak_ehkonomisty_ob'jasnjajut_(2-
e_izd)_44.html
4. A.I. Klimin, V.A. Urvalov. Ferdinand Braun — the Nobel prize in physics // "Elektrosvyaz', No. 8, 2000 (on the website "Virtual computer Museum"
5. D. Hendrie. Econometrics: alchemy or science (brown) M., Ekovest. 2003. No. 2. P. 172-196.
6. The economic crisis and the collapse of econometrics. - URL: ¡http://vened.org/economy/3720 -2010-07-09-08-42-18.html
7. N.N. Taleb .Black Swan. - URL: http://e-libra.ru/ read/255375-chernyj -lebed. -pod-znakom-nepredskazue-mosti.html
8. Alan Kirman. Interaction, Organisation and Aggregate Economic Activity, 1998, Working Paper, GREQAM, EHESS and UniversitH d'aix-Marseille III, Institut Universitaire de France.
9. Economic agents and their interests. - URL: http://www.be5.biz/ekonomika/e013/05.htm
10. Economic agents. - URL: http://www.bgsha.com/ ru/education/library/fulltext/econom/r1 -7. htm
11. Wikipedia. - URL: http://commons.wikimedia.org/ wiki/File:Macroeconomics.PNG?uselang=ru
12. A.R. Bakhtizin. Agent-oriented modeling. - URL: http://gdzmail.ucoz.ru/news/agent_orientirovannye_mode li_ehkonomiki_bakhtizin_albert/2012-10-19-352
13. Agent-oriented modeling and simulation: prospects in the field of information technology. - URL: http://www.artsoc.ru/digest/agent-oriented-models/ index.php?ID= 181
14. I.V. Kotenko, A.V. Ulanov. Agent-oriented modeling the behavior of complex systems in the Internet environment. - URL: http://comsec.spb.ru/ru/papers/60/ getfile.
15. Naidysh V. M. textbook.: the concept of modern natural history. Textbook.-Ed. 2nd Rev. and EXT. M., Agfa. - M., 2004, 622 p.
Использованные литературные источники
1. В.А.Бывшев. Эконометрика. - М., Финансы и статистика, 2008, 480 с.
2. И. Розмаинский. Методологические основы теории Кейнса и его "спор о методе" с Тинбергеном (рус.). - М., Вопросы экономики. 2007, № 4.
3. Кризис современной экономической теории. -URL : http://nanobukva. ru/b/altern/blaug_metodolo
gija_ehkonomicheskoj_nauki__ili_kak_ehkonomisty_ob'j
asnjajut_(2-e_izd) 44.html
4. А. И. Климин, В. А. Урвалов. Фердинанд Браун — лауреат Нобелевской премии в области физики // «Электросвязь» № 8, 2000 г. (на сайте «Виртуальный компьютерный музей»)
5. Д. Хендри. Эконометрика: алхимия или наука (рус). - М., Эковест. 2003. № 2. С. 172-196.
6. Экономический кризис и крах эконометрики. URL: http://vened.org/economy/3720- 2010-07-09-08-42-18.html
7. Н.Н. Талеб. Чёрный лебедь. - URL: http://e-libra.ru/read/255375-chernyj-lebed.-pod-znakomnepred-skazuemosti.html
8. Alan Kirman. Interaction, Economic Organisation and Aggregate Activity, 1998, Working Paper, GREQAM, EHESS and Universiffi d'Aix-Marseille III, Institut Universitaire de France.
9. Экономические агенты и их интересы. -URL:http://www.be5.biz/ekonomika/e013/05.htm
10. Экономические агенты. - URL: http://www.bgsha.com/ru/education/library/fulltext/econo m/r1-7.htm
11. Википедия. - URL: http://commons.wikimedia/ .org/wiki/File: Macroeconomics. PNG?uselang=ru
12. А.Р. Бахтизин. Агент-ориентированное моделирование. - URL: http://gdzmail.ucoz.ru /news/ agent_orientirovannye_modeli_ehkonomiki_bakhtizin_al bert/2012-10-19-352
13. Агент-ориентированное моделирование и имитационное моделирование: перспективы в области информационных технологий. - URL: //www.artsoc.ru/ digest/agent-oriented-models/ index.php?ID= 181
14. И.В. Котенко, А.В. Уланов. Агентно-ориентированное моделирование поведения сложных систем в среде Интернет. - URL: http ://comsec.spb.ru/ru/papers/60/getfile.
15. Найдыщ В.М. Концепция современного естествознания: Учебник. - Изд. 2-е перераб. и доп. -М., Агфа-М., 2004, 622 с.
- V -
УДК 330.43:658:14
МОДЕЛИ ФОРМИРОВАНИЯ ФИНАНСОВОЙ СТРАТЕГИИ ПРЕДПРИЯТИЯ В УСЛОВИЯХ НЕСТАЦИОНАРНОЙ ВНЕШНЕЙ СРЕДЫ
Гурьянова Л.С., д.э.н., профессор Трунова Т.Н., к.э.н., преподаватель
Харьковский национальный экономический университет им. С. Кузнеца, г. Харьков
Аннотация. Рассматривается комплекс моделей, который на основе методов многомерного анализа, векторной авторегрессии, моделей коррекции ошибки, методов анализа панельных данных, имитационного
«Хроноэкономика» № 1 (1) май 2016
www.hronoeconomics.ru