Chapter 39ACCOUNTING SCANDALS AND IMPLICATIONS FOR DIRECTORS: LESSONS FROM ENRON PEARL TAN, Nanyang Technological University, Singapore GILLIAN YEO, Nanyang Technological University, Sin
Trang 1Chapter 39
ACCOUNTING SCANDALS AND
IMPLICATIONS FOR DIRECTORS:
LESSONS FROM ENRON
PEARL TAN, Nanyang Technological University, Singapore GILLIAN YEO, Nanyang Technological University, Singapore
Abstract
We analyze the Enron case to identify the risk
fac-tors that potentially led to its collapse and specific
issues relating to its aggressive accounting and
high-light the lessons for independent directors In Enron,
the interactions between external stimuli, strategies,
corporate culture, and risk exposures possibly
cre-ated an explosive situation that eventually led to its
demise Much of the post-Enron reforms have been
directed towards regulating the roles and
responsi-bilities of executive directors and auditors However,
the role of independent directors has received
rela-tively lesser attention Independent directors should
analyze the risks of their companies and understand
the pressures that arise from market conditions and
firm-specific policies and incentive structures They
also need to close the information gap between
ex-ecutive directors and themselves A post-Enron era
also requires independent directors to change their
focus Traditionally, independent directors have to
strike a difficult balance between maximizing
re-turns and minimizing risks Independent directors
may now have to focus on the management of risks,
the design and functioning of an effective corporate
governance infrastructure, and the moderation of the
power bases of dominant executives Practically, they
may also have to reduce the number of independent
director appointments to enable them to focus more effectively on a fewer companies
Keywords: corporate governance; independent dir-ectors; risks; incentives; accounting scandals; spe-cial purpose entity; hedging; volatility; Sarbanes– Oxley Act; audit committee
39.1 Introduction
The recent spate of accounting scandals raises ser-ious concerns about the opportunistic use of accounting procedures and policies to camouflage fundamental problems in companies The series of corporate collapses also highlight the failure of corporate governance mechanisms to prevent and detect accounting irregularities The convergence
of several factors, including competitive pressures, conflicts of interest, lack of market discipline, and inherent limitations of accounting standards resulted in an explosive situation whereby man-agers use aggressive accounting practices to pre-sent financial statements that do not reflect economic reality In this essay, we analyze the Enron case with the objective of determining the risk factors that potentially led to its collapse and specific issues relating to its aggressive accounting and highlight the lessons to be learnt for corporate
Trang 2governance from the perspective of an independent
director
39.2 The Competitive Environment and Incentives
for Aggressive Accounting
Enron was formed as a result of merger of two
companies in 1985 The merger was funded by
debt and pressure had existed from the start for
the new company to reduce its debt burden At
about the same time, deregulation of the natural
energy industry exposed Enron to substantial
op-erating and price risks arising from the increase in
gas supply and volatility in spot prices However,
deregulation also increased opportunities for more
flexible and innovative contracts to be drawn up
between the producer and buyers To survive,
Enron had to capitalize on these opportunities
and became a primary market player through its
development of the idea of a Gas Bank Under this
scheme, Enron facilitated the market for energy
contracts by buying gas from suppliers and selling
to buyers In acting as an intermediary, Enron
guaranteed both the supply and the price, and
assumed the related risks in return for transaction
fees Innovations were subsequently extended to
markets for basic metals, pulp and paper, and
broadband products Its diversification strategy
also included investments in other countries in
South America, Europe, and Asia The business
and geographical diversification created new risks
for Enron Its heavy investment in projects such as
broadband network assets would pay off only in
the long term However, an immediate debt burden
from these acquisitions placed pressure on Enron’s
balance sheet that was already weighed down by
existing debt (Powers et al., 2002).1
Although Enron began as an operator of
en-ergy-related assets, by the end of the 1990s, the
firm had divested a significant portion of its
phys-ical assets in what is known as an ‘‘asset light
strategy’’ (Permanent Subcommittee on
Investiga-tions of the Committee of Governmental Affairs,
2002)2 and was primarily focused on its trading
and financial activities relating to physical energy
commodities Effectively, the company was trans-formed from a natural gas supplier into an energy trader and intermediary It offered specialist ser-vices in price risk management strategies and mar-ket-making activities Its dominance in the market for energy contracts gave Enron a first-mover ad-vantage in exploiting information economies of scale However, the lucrative profits it enjoyed attracted other entrants to the industry and Enron’s profit margins began to erode by the end
of 2000 Further, as a trader, Enron was compelled
to maintain an investment grade rating in order to lower its counter-party risk
Against this backdrop of competitive pressures, Enron’s senior management developed incentive schemes that turned the firm environment into a highly competitive internal market place An in-ternal ranking system administered by the com-pany’s Performance Review Committee became a means of allocating bonus points and determining dismissals The entire process was described as a
‘‘blood sport’’ (Chaffin and Fidler, 2002) and for-mer employees believed that the basis for reward was largely determined by whether a deal could be reported as revenue or earnings rather than com-mitment to the company’s core values of Respect, Integrity, Communication, and Excellence Enron’s annual incentive awards and the long-term incentive grants are closely tied to company performance measures and stock prices The annual incentive bonus was pegged to a percentage of recurring after-tax profit, while its long-term incentive grants provided for accelerated vesting provided Enron achieved performance targets linked to com-pounded growth in earnings-per-share and cumula-tive shareholder returns.3 A Senate report on the Enron collapse concluded that Enron’s Board of Directors approved lavish and excessive executive compensation and failed to stem the ‘‘cumulative cash drain’’ arising from its incentive schemes.4 Hence, Enron appeared to react to risk by creat-ing an environment that generated new risk expos-ures through its business strategies and reward system that focused on short-term results Figure 39.1 summarizes the competitive pressures at Enron
Trang 339.3 Aggressive Accounting Practices
Enron’s accounting practices resulted in removing
the liabilities of its balance sheet, improving
prof-itability, and reducing profit volatility These
de-sired accounting effects were achieved through
structuring numerous complex and ‘‘innovative’’
transactions Many of these transactions involved
dealing with special purpose entities that Enron set
up in partnership with related parties The
investi-gating Senate Committee described these practices
as ‘‘high-risk accounting.’’ The manner in which
certain transactions were reported was deemed to
be at variance with their true economic substance
The main question that underlies these practices
relates to the issue of whether Enron had retained
the risks that were purportedly transferred to the
special purpose entities
39.3.1 Effectiveness of ‘‘Hedging’’ Transactions
An example included the entering into transactions
that were purported to hedge the volatility of its
‘‘marked to market’’ investments The hedging
transactions were entered into between Enron and a special purpose entity (SPE).5 A hedge is effective only if a loss suffered by a hedged party
is transferred out to an outside party In its first hedging transaction, Enron transferred its own stock to the SPE in exchange for a note The intention of the hedge was to transfer losses to the SPE, through the exercise of an option, should the stock price of a profitable ‘‘merchant’’ invest-ment decline The SPE purported to take on the risk of price volatility of the investment and to compensate Enron for the loss on its investments However, cash was available to the SPE only if the latter sold the Enron stock Since the SPE was financed by Enron’s stock, the transaction was effectively a self-hedging arrangement as the creditworthiness of the SPE was tied to Enron’s fortunes When Enron’s stock fell in value in late
2000 and early 2001, the SPE faced a liquidity crisis and could not honor its obligations under the op-tion Hence, the ‘‘hedge’’ was ineffective because the counter-party’s risk was inextricably inter-twined with Enron’s risk and the hedge did not constitute a true economic hedge
39.3.2 Control and Risks Relating to
Unconsolidated Entities
There are two broad approaches in accounting for
an SPE If an SPE is controlled by an investing company, the assets and liabilities of the SPE are consolidated entirely on to the investing com-pany’s balance sheet Alternatively, if it is not under the investing company’s control, it is treated
as an investment in a separate entity, with off-balance sheet treatment of the SPE’s assets and liabilities Under applicable accounting rules in the United States, an SPE could receive off-bal-ance-sheet treatment only if independent third-party investors contributed at least 3 percent of the SPE’s capital Some of Enron’s dealings raised serious questions about whether this rule was effectively met
For example, from 1997 to 2001, Enron did not consolidate an SPE called Chewco In 1997, Enron
Strategy Change in focus from production to trading
Diversification
Consequences Price risks and volatility Operating risks from new businesses Financial risks from increased debt burden pressure to maintain investment grade credit rating
Corporate culture Competitive Profit-driven Short-term focus
Stimuli
Deregulation
of the natural
energy
industry
Debt burden
incurred
during the
merger that
resulted
in Enron
Figure 39.1 Competitive pressures at Enron
ACCOUNTING SCANDALS AND IMPLICATIONS FOR DIRECTORS: LESSONS FROM ENRON 645
Trang 4and the California Public Employees’ Retirement
System (CalPERS) were joint venture partners in
an off-balance sheet investment vehicle called Joint
Energy Development Limited Partnership (JEDI)
To enable CalPERS to cash out its investment in
JEDI in order to invest in a larger Enron venture,
Andrew Fastow, the then Chief Financial Officer
at Enron, and others at Enron formed an SPE
called Chewco to buy CalPERS’ interest in JEDI
Thus, Enron was able to continue accounting for
JEDI as an off-balance-sheet entity on the basis
that the holdings by Enron staff members and
related parties constitute outside capital at risk
According to SEC investigations,6Fastow, secretly
controlled Chewco Hence, a serious question
arose as to whether Enron, through a related
party, had effective control over major operating
and financial policies of Chewco Further, Enron
and its related SPEs provided guarantees and cash
collateral on bank funding to Chewco, indicating
that equity at risk was effectively borne by Enron
rather than independent third parties In
Novem-ber 2001, both Enron and its auditors, Andersen,
concluded that Chewco was an SPE without
suffi-cient outside equity and should have been
consoli-dated The retroactive consolidation of Chewco
from 1997 through 2001 had an astounding effect
on the financial statements Profits decreased by a
total of $405 million over the period of restatement
and additional debt of $711 million was recognized
on the balance sheet in 1997.7
39.4 The Role of Corporate Governance
Theoretically, Enron had in place an impressive
array of corporate governance mechanisms
Out-side directors were well respected and highly
quali-fied individuals in the fields of accounting, finance,
and law The Board of Directors had several
com-mittees to review various aspects of the company’s
policies and operations There was separation of
the offices of the Chairman and Chief Executive
Officer The external auditors were a Big Five
accounting firm However, following the
com-pany’s massive financial collapse, serious doubts
arose as to the effectiveness of these institutional arrangements The Senate Investigating Commit-tee found that the Enron’s Board failed to safe-guard Enron shareholders and contributed to the collapse of the company by allowing Enron to engage in high-risk accounting, inappropriate con-flict of interest transactions, extensive undisclosed off-balance-sheet activities, and excessive executive compensation.8Further, the Board was also found
to have failed to ensure the independence of the company’s external auditor, Andersen who pro-vided internal audit and consulting services as well.9
Many valuable lessons can be learnt from the Enron case to prevent the derailing of the effective functioning of governance mechanisms We focus our discussion on the role of independent directors Much of the post-Enron reforms have been direc-ted towards regulating the roles and responsibil-ities of executive directors and auditors However, the role of independent directors has received rela-tively less attention than that of other corporate governance agents We discuss below some impli-cations of the Enron collapse on the role of inde-pendent directors
(i) What is the primary role of independent direct-ors? The multiple roles that independent directors have to undertake require them to strike a difficult balance between maximizing returns and minimizing risks Their purview is wide, ranging from activities that have a
‘‘profit’’ focus to others that have a ‘‘defensive’’ focus Independent directors potentially find themselves in an identity crisis For example,
if an independent director has to operate within
an Enron-type environment, the director is confronted with an aggressive risk-taking in-ternal environment The question arises as to whether the independent director should act as
a thorn in the managers’ flesh or go with the flow of an aggressive managerial style for the sake of profit maximization?
The lesson from Enron is very clear that it does not pay to sacrifice the defensive role when risk
Trang 5factors are overwhelming and the long-run survival
of the company is at stake While post-Enron
legislation such as the Sarbanes–Oxley Act of
2002 is primarily directed towards establishing
mandates for insiders, audit committee board
members and external auditors, much less is said
about the responsibilities of independent directors
per se However, the implicit responsibilities of
independent directors are clearly reinforced by
laws that impose fiduciary duties on directors to
act in good faith, with reasonable care, and in the
best interest of the corporation and its
share-holders The Conference Board also reiterates
dir-ectors’ role to monitor management and to ensure
their ethical and legal compliance (The Conference
Board, 2003).10
Hence, independent directors owe a primary
duty of care to outside investors Their priority
should be towards establishing and ensuring a
cor-porate environment and infrastructure wherein
managerial stewardship is executed without
com-promising the long-run interests of the firm and its
stakeholders They, more than anyone else, are
best placed to limit the excesses of a dominant
Chief Executive
(ii) Independent directors have to bridge the
infor-mation gap between executive directors and
themselves The Conference Board emphasizes
that directors need to understand, among other
things, the business strategies they approve, the
risks and vulnerabilities arising from the
strat-egies, growth opportunities, debt levels, and
company’s capital allocation of the companies
under their purview.11 Following the Enron
experience, independent directors are well
ad-vised to understand the internal dynamics,
managerial incentives, and power bases within
the corporate environment and to adopt a
healthy skepticism of strategies that potentially
advance managerial interests over that of
exter-nal investors They should be keenly aware of
the threats posed by dominant Chief Executive
Officers and key personnel and the risks of
opportunistic managerial behaviour
(iii) Greater commitment in terms of time and effort are expected of independent directors
to meet the governance objective Independ-ent directors must take a proactive role in governance and not rely solely on external auditors, legal counsel, or key executives to provide them the necessary assurance For example, when the Enron Board was asked why they moved so quickly in their approval
of an unusual hedging transaction, the re-sponse was that the company had obtained a fairness opinion from an outside accounting firm.12On another proposal, the Board relied
on the company’s legal counsel to advise if anything was amiss on a particular memoran-dum Had the directors reviewed the memo-randum for themselves, they would have noted that key company executives were in-volved in the arrangement that gave rise to conflicts of interest.13 Interviewed Board members told the investigating Senate Sub-committee members that they assumed that the then Chief Executive Officer had actively reviewed and approved the fairness of the unusual business proposals and the compen-sation controls.14Enron’s directors were also found to have knowingly allowed Enron’s use
of ‘‘high-risk’’ accounting without enforcing restraint.15 Hence, the Senate Report under-scores the principle that evidence of a suspect transaction or activity that is known to a director must be questioned and examined diligently and thoroughly, regardless of the views of other experts
The implications for independent directors are enormous The days when an independent director held several of such appointments concurrently are likely to be over Independent directors may have
to be selective in choosing appointments so as not
to spread themselves too thinly They must also be prepared to commit resources and time and change the mindset that their appointment is a ‘‘part-time’’ one They may also have to assess the risks
of companies to determine if they are willing to ACCOUNTING SCANDALS AND IMPLICATIONS FOR DIRECTORS: LESSONS FROM ENRON 647
Trang 6undertake the fiduciary responsibility of
monitor-ing such a company
39.5 Conclusion
The Enron case has painful lessons for the business
community A seemingly successful company was
apparently derailed through the use of highly risky
transactions and aggressive accounting that
tem-porarily boosted profits and reduced debt The
question arises as to why the corporate guardians
of Enron did not prevent these transactions from
occurring Following Enron and other accounting
scandals, a re-examination needs to be carried
out of the role and responsibilities of independent
directors This paper suggests that significantly
greater challenges are posed to independent
direct-ors in a post-Enron world to understand more of
the risks, accounting practices, and managerial
op-portunism existing in the companies under their
purview and to take a more proactive role in
gov-ernance, which inevitably requires a substantial
commitment of their time and resources
NOTES
1 Hereinafter referred to as the ‘‘Powers Report.’’
2 Hereinafter referred to as the ‘‘Senate Report,’’ p 7.
3 Proxy Statement Pursuant to Section 14(a) of the
Securities Exchange Act of 1934, 2 March 2001,
EDGARPlus(R).
4 The Senate Report, p 3.
5 Details of the hedging transactions are found in The Powers’ Report, pp 13–15.
6 Securities and Exchange Commission, Litigation Release 17762, 2 October 2002.
7 The Powers’ Report, p 42.
8 The Senate Report, p 11.
9 The Senate Report, p 54.
10 Hereinafter referred to as The Conference Board Report.
11 The Conference Board Report, p 9.
12 The Senate Report, p 27.
13 The Senate Report, p 28.
14 The Senate Report, pp 30–31.
15 The Senate Report, pp 14–24.
REFERENCES Chaffin, J and Fidler, S (2002) ‘‘The Enron Collapse.’’ Financial Times, London, 9: 30.
Permanent Subcommittee on Investigations of the Committee of Governmental Affairs.(2002) United States Senate, The Role of the Board of Directors in Enron’s Collapse Report, 107–170.
Powers, Jr W.C , Troubh, R.S., and Winokur, Jr H.R (2002) ‘‘Report of investigation by the special inves-tigative committee of the board of directors of enron Corp.’’
Securities and Exchange Commission (2002) ‘‘Litiga-tion Release No 17762.’’
The Conference Board (2003) ‘‘Commission on public trust and private enterprise.’’
Trang 7Chapter 40
AGENT-BASED MODELS OF FINANCIAL MARKETS NICHOLAS S P TAY, University of San Francisco, USA
Abstract
This paper introduces the agent-based modeling
methodology and points out the strengths of this
method over traditional analytical methods of
neo-classical economics In addition, the various design
issues that will be encountered in the design of an
agent-based financial market are discussed
Keywords: agent-based models; computer
simula-tion; bounded rationality; heterogeneous agents;
learning; co-evolution; complex adaptive system;
artificial intelligence; neural networks; classifiers;
genetic algorithms; genetic programming
40.1 Introduction
The sort of phenomena that are interesting in
fi-nance and yet difficult to investigate
analytic-ally involve the complex interactions among many
self-interested heterogeneous boundedly rational
agents acting within the constraints imposed by
either formal or informal institutions or
author-ities To outrival their opponents, each and every
agent must continually evolve to adapt to changes
that may arise either from exogenous
perturba-tions to the environment or endogenous transiperturba-tions
caused by agents changing their strategies or
modi-fying their behaviors as they learn more about the
behaviors of the other agents and the environment
they reside in A good example of such complex
adaptive systems is the stock market
A natural way to study a complex adaptive sys-tem like the stock market is to use an agent-based model which entails simulating the stock market on
a computer from the bottom up with a large number
of interacting heterogeneous boundedly rational artificial agents that are created to mimic the traders in the stock market Once the environment
of the stock market and the behaviors of the agents are specified and the initial state of the model is set, the dynamics of the model from the initial state forward will be driven entirely by agent–agent inter-actions, and not by some exogenously determined systems of equations Hence, if any macroscopic regularity emerges from the model, it must be a product of the endogenous repeated local inter-actions of the autonomous agents and the overall institutional constraints This is the spirit of the agent-based modeling approach
What makes the agent-based modeling method-ology particularly appealing? To begin with, ana-lytical tractability is not an issue since this approach relies on computer simulations to under-stand the complex model Quite the reverse, it is inconceivable how one could obtain closed form solutions of a model as complex as the stock market without first diluting drastically the au-thenticity of the model Although analytically tractable heterogeneous agent rational expect-ations models have been around, the complexity and realism that are captured in agent-based models are beyond the reach of those analytical models
Trang 8For instance, consider the problem that a
deci-sion maker faces when the outcome is contingent
on the decisions to be made by all the participating
heterogeneous decision makers, each with their
own unique preferences and quirks and private
information that are not directly observable by
the other decision makers This decision problem
is inherently ill defined and cannot be solved
through mathematical deduction or analytical
modeling In real life, when confronted with such
an ill-defined situation, decision makers often rely
on the rules of thumb that they have distilled
from years and years of experience to guide them
in their decision-making This decision making
process is formally known as inductive reasoning
and it can be captured naturally with the
agent-based approach by running computer simulations
of a large number of interacting artificial agents
who make decisions using rules of thumb that they
distill from their repeated interactions with each
other
The ability to build more realistic models with
the agent-based method often allows agent-based
models to reveal a much richer set of behaviors
that are embedded in a system which may
other-wise be overlooked by traditional equation-based
models For instance, Parunak et al (1998) in
comparing the differences between
equation-based modeling and agent-equation-based modeling of a
supply network have found that equation-based
model fails to produce many of the rich effects,
such as memory effect of backlogged orders,
tran-sition effects, or the amplification of order
vari-ation, which are observed in an agent-based model
of the same supply network In addition, various
agent-based models (Farmer and Joshi, 2000;
Johnson et al., 2001; LeBaron et al., 1999; Tay
and Linn, 2001) have been successful in accounting
for real financial markets phenomena such as
market crashes, mean reversion, relatively high
level of trading, technical trading, excess volatility,
and volatility clustering These are phenomena
that analytical representative agent models of
fi-nancial markets have tolled to explain without
much success
Another serious shortcoming of analytical rep-resentative agent models of financial markets is that by design these models do not specify the dynamic process that will need to happen in order
to arrive at the equilibrium or equilibria that are characterized in these models Consequently, for models that produce multiple equilibria, it is un-clear which equilibrium among the multiple equi-libria agents would converge on In contrast, the events that unfold in a computer simulation of an agent-based model are completely transparent, and can be recorded hence providing the modeler a means to go back in the time line of evolution to understand how certain equilibrium or other global regularities came into existence
The agent-based methodology therefore offers important advantages over the traditional analyt-ical tools of neoclassanalyt-ical economics as it allows a researcher to obtain more germane results Need-less to say, the use of computer simulations as a tool for studying complex models has only became feasible in recent years because of the availability
of fast and cheap computing power Although the agent-based modeling methodology is still in its infancy, there is already a considerable lit-erature on agent-based models Leigh Tesfatsion
at the Iowa State University maintains a website
at http:==www.econ.iastate.edu=tesfatsi=ace.htm to facilitate access to the extensive resources related
to the agent-based modeling methodology, and to keep researchers in this field abreast of the latest developments
In the introductory remarks on her website, Tesfatsion observes that agent-based research may generally be organized according to one of the following four research objectives: (1) empirical understanding, (2) normative understanding, (3) qualitative insight and theory generation, and (4) methodological advancement The first objective focuses on seeking answers that are established
on the repeated interactions of agents to explain the emergence of global regularities in agent-based models Some examples of global regularities in financial markets are mean reversion and volatility clustering Researchers in this group are interested
Trang 9in understanding if certain types of observed global
regularities can be attributed to certain types of
agent-based worlds The second objective concerns
using agent-based models as laboratories to aid in
the discovery and design of good economic policies
or good institutional structures Researchers with
this objective in mind are interested in using
agent-based models to evaluate whether certain
eco-nomic policies or institutional designs and
pro-cesses will promote socially desirable outcomes
over time among agents that are driven solely by
their self interests Tesfatsion phrased the third
objective as ‘‘How can the full potentiality of
eco-nomic systems be better understood through a
better understanding of their complete phase
portraits (equilibria plus basins of attraction)?’’
Unlike analytical models, the causal mechanisms
in agent-based models are not direct and are very
difficult to discern because of the complex nature
of the interactions among the agents and between
the agents and the environment The goal here is to
use the phase portraits as a means to enrich our
understanding of the causal mechanism in these
systems The fourth objective addresses issues
re-lated to improving the methods and tools used by
agent-based researchers
For someone who is just starting out in this line
of research, it is worthwhile to begin by reading
‘‘A Guide for Newcomers to Agent-based Modeling
in the Social Sciences’’ by Axelrod and Tesfatsion
which is available on the homepage of Tesfatsion’s
website In addition, it is beneficial to read the
survey articles written by Hommes (2004), Duffy
(2004), LeBaron et al (1999), LeBaron (2000,
2004a), and Tesfatsion (2002) and a book by Batten
(2000) that provides an overview of agent-based
models and offers some historical perspectives of
this methodology
The next section discusses the design issues that
will be encountered in the design of an agent-based
model This discussion benefited greatly from the
insights that LeBaron has provided in his excellent
overviews of the various design issues (LeBaron,
2000, 2001c, 2004a)
40.2 Design Considerations
A typical agent-based model is made up of a set of autonomous agents that encapsulate the behaviors
of the various individuals in a system we are inter-ested in studying and the investigation involves simulating on a computer the interactions of these agents over time Accordingly, there are two important design considerations in the develop-ment of an agent-based model – the design of the agents and the design of the environment
How naive or sophisticated the agents should be modeled really depends on the objective of the research For instance, if the research objective
is to understand how certain market structures affect the allocative efficiency of a market inde-pendent of the intelligence of the agents as in Gode and Sunder (1993), then one can simply model the agents as naive ‘‘zero intelligence’’ agents Zero intelligence agents are agents that are not capable of formulating strategies or learn-ing from their experience; hence their behaviors will be completely random Gode and Sunder populated their double auction market with zero intelligence agents that are designed to submit their bids and asks at random over a predefined range and remarkably they discover that zero intelligence agents when subjected to a budget constraint are able to allocate the assets in the market at over 97 percent efficiency The lesson to be learned here is that not all macroscopic regularities that emerge from agent-based models are necessarily conse-quences of the actions taken by the agents as they evolve and learn from their interactions In this case, the high level of allocative efficiency that is attained in a double auction market is due to the unique structure of the market itself
However, in many agent-based models, the ob-jective is to investigate the outcome of the inter-actions among many heterogeneous agents that are designed to mimic their counterparts in the real world In these models, the key design issues related to the design of the agents are the agents’ preferences and their decision-making behaviors
Trang 10Agents could have either myopic or intertemporal
preferences The latter is more realistic but will
make the model much more complex As we have
alluded to earlier, the decision problem that the
agents face is usually ill defined, and thus cannot
be solved by deductive reasoning A reasonable
solution is to assume that the agents rely on
in-ductive reasoning to arrive at a decision (see
Arthur, 1994, 1999; Rescher, 1980) Inductive
reasoning or induction is a means for finding the
best available answers to questions that transcend
the information at hand In real life, we often have
to draw conclusions based upon incomplete
infor-mation In these instances, logical deduction fails
because the information we have in hand leaves
gaps in our reasoning In order to complete our
reasoning, we fill those gaps in the least risky,
minimally problematic way, as determined by
plausible best-fit considerations Consequently,
the conclusions we draw using induction are
sug-gested by the data at hand rather than logically
deduced from them
Inductive reasoning follows a two-step process:
possibility elaboration and possibility reduction
The first step involves creating a spectrum of
plausible alternatives based on our experience and
the information available In the second step, these
alternatives are tested to see how well they answer
the question at hand or how well they connect the
existing incomplete premises to explain the data
observed The alternative offering the ‘‘best fit’’ is
then accepted as a viable explanation
Subse-quently, when new information becomes available
or when the underlying premises change, the fit of
the current alternative may degrade When this
happens a better alternative will take over
How can inductive reasoning be implemented in
an agent-based financial market model? Arthur
(1994, 1999) envisions inductive reasoning in a
financial market, taking place as follows Initially,
each agent in the market creates a multitude of
decision-making rules (this corresponds to the
pos-sibility elaboration step discussed above) Next,
the decision-making rules are simultaneously
tested for their effectiveness based on some
cri-teria Finally, effective decision-making rules are retained and acted upon in buying and selling de-cisions Conversely, unreliable rules are dropped (this corresponds to the possibility reduction step) The rules that are dropped are then replaced with new ones in the first step and the process is carried out repeatedly to model how individuals learn inductively in a constantly evolving financial market
Some examples of criteria that have been used for appraising the effectiveness of the decision rules includes utility maximization, wealth maxi-mization, and forecast errors minimization Once a decision has been made on a criterion for evaluat-ing the decision-makevaluat-ing rules, the next task is to decide the length of historical data to be used in computing the criterion Although many agent-based models tend to allow the agents to adopt identical history length, this is not necessary It is
in fact more realistic to permit agents in the same model to adopt different history length as in LeBaron (2001a,b)
To take the modeling to the next step, decision will have to be made concerning what the decision making rules look like and how they are to be generated in the models? One possibility is to model the decision-making rules after actual trad-ing strategies used in real financial markets The benefit of this approach is that the results are likely
to be tractable and precise and it will also shed light on the interaction among these actual trading strategies However, this approach does not allow the agents any flexibility in modifying the strat-egies or developing new stratstrat-egies This could im-pose ad hoc restrictions on the model’s dynamics Some common tools that have been employed to allow the agents more degrees of freedom in struc-turing and manipulating the decision making rules
as they learn are artificial neural networks (LeBaron, 2001a), genetic programming (Chen and Yeh, 2001), and classifiers that are evolved with genetic algorithms (LeBaron et al., 1999) Even with these artificial intelligence tools, the modeler will need to predefine a set of information variables and functional forms to be used in the