Bounded Rationality and the SD Approach for energy modelling
ISAAC DYNER and CARLOS JAIME FRANCO
Universidad Nacional de Colombia. Sede Medellin
idyner@ perseus.unalmed.edu.co; cjfranco@ isa.com.co
ABSTRACT
The liberalisation of energy markets provides a challenge for what is known as rational energy
use, and therefore for the understanding of end-users behaviour under both government
policies and traders strategies. In this environment, it is important to assess the decision
making processes of energy end-users for policy and strategy intents.
The methodologies reported in the literature generally make strong assumptions with respect
to the decision making processes of end-users, including: complete information, full rationality
and lack of risk perception. In this paper, Simon’s bounded rationality (BR) is picked up,
seeking for alternative grounds to the traditional methodologies.
System Dynamics (SD) modelling, with the help of other approaches, seems to provide a most
appropriate environment for a wide number of BR applications in the energy field, as it has
been claimed in the literature that this approach is well grounded on bounded rationality.
However, both this claim, as well as to how SD handles uncertainty, which basically accounts
for BR, needs to be clearly justified.
In this paper we undertake BR under the SD framework in two areas: modelling
methodologies and applications. First by indicating, explicitly and implicitly how to incorporate
uncertainty in SD and second by providing guidelines for establishing how variables are
included and discarded from models.
1 FROM CENTRAL PLANNING TO OPEN ENERGY MARKETS
During the late 80s, there was a world trend of disbelief with respect to the effectiveness of
state owned monopolies in charged of public services, which gave way to important reforms
initially in Chile, the United Kingdom and Norway, and few years later in other European and
American countries as well as in Australia.
Changes intended to:
¢ Introduce competition in the electric sector.
* Allow private investment and the privatisation of state owned companies.
* Eliminate vertical integration, unbundling the transmission, distribution and generation
businesses.
* Focus governments actions in the area of industry regulation and in few cases also
performing indicative planning for expansion.
One of the main changes introduced by these reforms was the promotion of competition at the
final user's level, allowing consumers to freely procure electricity.
Against this background, some thought that these reforms were contradictory with Demand
Side Management -DSM- which encompasses planning, implementation and follow up
activities, with the purposes of modifying levels and patterns of electricity consumption. By
means of these programmes, important benefits were achieved by both end users and power
companies, especially with respect to: energy efficiency initiatives, direct load control, load
interruptible and hourly pricing (EIA, 1996).
However, it was not clear that the conditions for rational energy use under a centrally planned
scheme prevail when markets are liberalised.
2 EFFICIENCY AND RATIONAL ENERGY USE UNDER OPEN MARKETS
Under competition, companies can not be forced to undertake programmes of rational energy
use. Government may implement indicative planning or create incentives but cannot force
utilities or final users to undertake such programmes. Utilities would only offer such
programmes if these are profitable, but which companies in the electric sector maybe interested
in undertaking such programmes? To provide answers to this question we would try to
understand the motivations of the different agents that intervene in the electric market.
Transporters would be interested in efficiency programmes, directed to reduce losses at
economic levels. As regulated revenues depend on the transported energy, the business logic
seems simple: transport as much as possible. Thus, they would not be interested in undertaking
other programmes.
Generators may be interested in efficiency and production “optimisation”. Therefore, the
regulator may incentivate highly competitive markets, forcing generators to produce at optimal
levels to survive in the market place.
Traders may have incentives towards undertaking rational energy use programmes.
For established traders:
¢ Implementing rational use programmes to end users by means of binding contracts.
* Promoting rational use to reduce demand when prices in the wholesale market are higher
than prices which have been agreed upon with end-users.
* Promoting rational use to those sectors where electricity is subsidised by law.
For new traders or those who intend to increase their market share:
¢ Promoting rational energy use as part of a strategy to attract clients.
¢ Promoting rational energy use to offer lower prices, using a similar strategy to the ESCOS
~Energy Service Companies.
As has been discussed, open markets do not seem to discourage programmes of rational
energy use and efficient production. It is not clear though as to when and which agents would
undertake demand side management and efficiency programmes under open markets, because
of possible market barriers.
Sioshansi (1995) considers that under open markets DSM programmes should operate better
than under centrally planned schemes. To understand this issue better, let us revise the
evolution of DSM based on Sioshansi’s perspective. He identifies three periods for DSM:
1. 1973-1989 Command and control without incentives.
2, 1989-1994 Command and control with incentives.
3. 1994 - Consumer driven to achieve better efficiency.
The third wave of the DSM is marked by two events that can completely change the concept
of the electricity business: competition in the trading business and free access to nets.
Sioshansi argues that under market schemes DSM can be more effective than ever before,
however the literature has already reported some difficulties. Several researchers have noticed
that substantial differences exist among the current levels of energy efficiency and those that
should exist if they had undertaken the appropriate decisions under a rational approach. This
problem is referred to in the literature as the energy efficiency gap (Jaffe y Stavins, 1994),
which will be discussed next.
The effectiveness of market mechanisms to reach socially desirable levels of energy efficiency
had generated a debate among specialist and policy makers. The neoclassical assumptions of
perfectly informed markets and rational individuals that consider all costs, when deciding what
appliances to buy, seem insufficient for understanding DSM.
When observing electricity consumption, agents seem to implicitly use discount rates above
89%, which is much higher than those rates applied to other common investments (Howart and
Andersson, 1993; Hausman, 1979; and Gately, 1980).
Some authors claim that consumers use such high discount rates due to uncertainties related to
the operation and penetration of new technologies. Other authors argue that these high rates
maybe explained by the consumers’ low revenues and their difficulty for investment in new
technologies. This phenomenon reveals that consumers expect to recover investments in very
short periods. These two problems may be solved by: a) providing guarantees or b) subsidising
low revenues customers. To this respect, a critical issue is establishing the size of the energy
gap and the definition of the optimal level of energy efficiency (five independent notions of
optimality have been identified; Jaffe and Stavins, 1994).
This section has been focusing on issues related to failures in electricity markets. Apparently
some of these failures are due to market uncertainties and other aspects related to the decision
making processes of final users: cultural barriers, aesthetic parameters, habits, asymmetric
information or knowledge and high risks.
The above discussion suggests that the problem of demand side management and efficiency
should incorporate the perspective of rational energy use under some sort of bounded
conditions as suggested by Simon (1978) - this could be referred as rationally bounded energy
use.
3 DECISIONS MAKING
A simple way to explain an action, is to consider it as depending on two successive filtering
operations (Elster, 1990). The initial step begins enquiring about all possible actions that the
individual can undertake. The first filter comprises all the physical, economic, legal and
psychological restrictions that the decision maker faces. We are left after the first filtering with
an opportunity group. The second filtering operation determines which actions are included in
the opportunity group, figure 1A.
We can observe that due to the first filtering operation, the election may not be carried out to
the second filtering operation. That is to say, the opportunity group decreases in the first step.
Opportunities are more basic than desires in one aspect: they are easier to observe. The
military strategist states that his plan should be based on the enemy's capacity (Verifiable)
rather than on his intentions (unverifiable) (Sun Tzu, 500 BC). This means, getting ready for
the worst case, in which the enemy will harm us as much as he can possibly do.
It is important to realise that it is usually easier to change people's opportunities than their way
of thinking. Desires and opportunities can be influenced each other and they can both be
influenced to produce an action, as illustrated in figures 1 B,C.
Desires
——— i
Gatien eo A Action (A)
Sa Opportunities
Desires
. ~— > Action —@)
oulenunmes””
Desires
— Action (C)
to) — “nt
Figure 1. Desires and Opportunities
When facing a decision, people usually do what they believe is best on his behave. This
sentence, deceivingly simple, summarises the theory of the rational election (Elster, 1990a).
For an action to be rational should be the result of three optimal decisions (Elster, 1990a).
First, it should be the best means to carry out one person’s desire according to his beliefs.
Second, these beliefs should be optimal given the evidence that the person has. Finally, the
person should gather a good amount of evidences, neither too much nor too little. The correct
amount depends on his desires, the importance assigned to the decision and his beliefs about
costs and benefits. Figure 2 (Elster, 1990b) shows the process.
Desires A
Beliefs
aan fr
Evidences
Figure 2. Optimisation process in rational decision making.
When making a decision, we confront risk for acting too soon with little information or for
delaying the process of information gathering, until it is too late.
Decisions making is necessary limited because of the processing capacities of human beings
and the failure of rationality as a consequence of physiologic or sociological problems.
The logic is to establish a set of rules to progress from an unknown condition to one which is
known (Le6n, 1994). The problem is to define what is true in a series of situations and to
established their limits. For example, the expression “the water is hot” becomes true under
certain circumstances in a specific environment. Also, this logic does not establishes what is
true or false, but rather provides a great number of combinations between both, such as it is
done in fuzzy logic.
Human intelligence does not work as this logic prescribes. According to Simon (1978), human
beings are fundamentally adaptive to the environment. An environment of high complexity
would paralyse a totally logical system, forcing the decision maker to discard a large amount of
information. In highly complex environments, calculations should be made by procedures that
allow us to reach the objectives on time, although not always providing the best or optimal
solutions.
Against this background, two important characteristics of the problem solving process are: a)
that mental representations are supported in work memory with finite capacity only, and b) that
human decisions differ from computationally logical systems, since they are affected by the
environment. While formal system disregard context, humans seem to integrated it better in its
decision making process. Is it correct then to attribute full rationality to man and to model it
under these assumptions?
In organisations, although man does not behave totally rational, he intends to - this is
denominated rationally bounded behaviour. The importance of bounded rationality resides in
better understanding the decision making process for better policy modelling.
4 MODELLING
Modelling is centred in the problem of understanding decisions for most effective policy
making, specifically for understanding the decisions of the agents that intervene in the energy
market. Problems have been detected at three levels:
Producers: These agents produce electricity for sale in the wholesale market, by way of
contracts or directly in the pool. Their clients are traders and other generators. They do not
have direct contact with final users.
Traders: These are agents that, by means of purchases and sales, provide electricity to final
users. These agents also carry out covering operations, selling electricity to generators and
other traders. Sales to final users are carried out by way of contracts, in the case of large
consumers, and through tariffs determined by the regulator, in the case of regulated users.
Final Users: There are two types of final users. Those who can freely negotiate prices with a
trader, and those that have fixed tariffs. The first ones are denominated not regulated users and
the second ones regulated users.
In this research we will focus on the limitations of agents from point of view of both final users
and traders, investigating what effects these produce on other agents in the electricity market.
These two types of agents were chosen since they are the most important ones from the point
of view of demand. Policies evaluated in this paper are geared towards overcoming the barriers
of final users for the adoption of certain type of energy technologies. This research may also
support traders strategy intents.
The methodologies for bounded rationality should be able to overcome some of the
deficiencies exhibited in the classical approaches, specifically with respect to efficiency, rational
energy use and DSM, as these usually assume full rationality. It has been argued that the
classic approaches are ineffective (Bunn and Dyner,1996), as they are:
NORMATIVE
DETERMINISTIC
LINEAL
NON-SY STEMIC
MECHANISTIC
STATIONARY
These feature become modelling requirements for a new methodologies. To improve our
understanding about demand evolution and the interactions between traders and final users,
alternative approaches also need to support simulation capabilities for policy assessment. With
this research we expect to improve our vision on technology diffusion, rational use of energy,
efficiency, substitution, losses, cultural aspects and externalities such as environmental
contamination, all these under a deregulated environment and strong competition between
traders.
The precedent section identified that desires and opportunities are relevant aspects in the
decisions making process. It was also observed that man's limitations in decision making maybe
explained by bounded rationality arguments. A framework is now proposed for the
understanding of decision making for energy investment from the point of view of end users.
This framework uses the inherent sociological and psychological limitations of mankind.
In general terms the decisions making framework is represented in the figure 3. This system
evolves according to the actions of participants. These actions consider desires and
opportunities, and policies induce changes in desires and the opportunities.
System
Government
policies
Desires Finalusers _ Oportunities
Trader
Estrategies
Figure 3. General scheme of the system.
In general, for the problem of technology choice, decisions making operates as shown in figure
4, where there exists filters for desires and the opportunities.
Technologies
Availability of:
Communication,
Filtering opportunities
Selection
Knowledge Filtering desires
Risk aversion..ete
Population
Figure 4. Double filtering of desires and opportunities.
To develop a methodology that allows incorporating the concepts of bounded rationality is
necessary to find a tool capable of supporting the implicit limitations of the decisions making in
energy systems. We discuss next our modelling approach using systems dynamics. This
modelling approach focuses on the representation of end user’s decision making.
When evaluating several tools for the representation of energy systems, the literature provides
foundations to carry it out using systems dynamics. In the first place Bunn and Dyner (1996)
state that systems dynamics may be a sensible option, especially if one works with changing
systems or with recently restructured markets. On the other hand Morecroft (1983) argues that
systems dynamics implicitly uses bounded rationality in its formulations.
The Camegie School states that the behaviour of complex organisations can only be
understood taking into account the knowledge of individuals and their psychological
limitations (Morecroft, 1983). This point of view focuses on the flows of information within
complex systems, the quantity and quality of the information that is manageable in the decision
making processes and the decision rules used.
It is at the level of information flows and decision making where most of the fundamental ideas
of the Camegie School coincide with systems dynamics, but it is also at this level where
Camegie School radically brakes apart from the point of view of economics and the
operational research - specifically with respect to desires and opportunities in the selection of
efficient technologies and with respect to electricity planning.
In relation to problem solving by way of simulation, systems dynamics and the Carnegie School
share the belief that simulation is a powerful way of gaining understanding of the decision
making process.
There are three important ideas of the Carnegie School that could positively influence systems
dynamics (Morecroft, 1983):
Improve communication with other disciplines: The Carnegie School offers an altemative
methodology in which feedback structures can be explained in terms of flows of information
and the processing of this information in the decisions making process.
Conceptualisation focusing in the organisation: Models conceptualisation can be improved
if those who builds the models clearly recognise that they are building models of human
organisations and that they are governed by the principle of bounded rationality. This does not
mean that we would construct models for better decisions, but rather models that are more
adjusted to reality.
Analysis of behaviour, making use of intentional rationality: Systems dynamics models are
of two types: partial and total models. The partial ones represent a portion of a process and
will be possible to observe intentional rationality, this means that parts of the organisation
could behave slightly more rationally if they were allowed to behave freely. When several
groups interacts, policies based on intentional rationality can fail. This is probably why policies
that succeed in small environments, fail when facing more complex environments.
Systems dynamics can add more elements to the process of model formulation and a new
dimensions to the analysis of behavioural models than the Camegie School Approach
(Morecroft, 1983).
The construction of any systems dynamics model may begin by establishing the main variables
that have an impact on the system under consideration and the elaboration of chains of
causation among variables. In a later phase, a stocks and flows diagram is built. Finally
mathematical relationships among model variables are constructed and the values of the
parameters and the initial conditions of the level variables are determined.
When building a system dynamics model that represents decision making, it is also necessary to
consider: desires, opportunities and the decisions making mechanism itself, this is represented
next.
Determination of Opportunities
Modelling begins by establishing the set of all opportunities, that is determining all possible
decisions that the individual can make. This group may be reduced because of the technologies
that are available in the local market. In turn this group of technologies is limited by the
technologies that the decision maker considers.
The set of real opportunities is finally reduced by the economic limitations of end users. The
general diagram of the opportunities can be appreciated in figure 5.
With this vision, modelling opportunities can be approached by reducing values gradually,
before reaching a decision.
Determination of Desires
Modelling desires consists on determining the portion of population that will not be shifting
away from former energy technologies because they have no knowledge about alternatives, or
because they are reluctant to change (and they not even consider the possibility of change), or
even because they are risk averse. All these inhibit the entrance of technology alternatives.
There are some problems related to modelling. Perhaps the main one is that related to
modelling habits, which seems to be an important factor against technological changes. The
general diagram that describes the filtering processes of desires and opportunities can be
appreciated in figure 5.
Availabilty of
Technologies
Sveunuiy
7” mee
Soper Opportunity
set known by
the user
Availability of
energy ot
distribution
channels
Real
: opportunity :
: set .
: Availability of
: technological .
. change Economic i
. opportunity Filter of :
. set } Opportunities ‘
SELECTION
Filter of Desires :
High discount People with
rates for the desire and Population :
changing capability of that have the :
selection: Economical :
opportunity
6, set %
Change em Income, Total
“Aversion: Hats Knowledge...ete Population
Figure 5. Double filter in technology choice.
Establishing limits for decisions making
When individuals confront decision making with no limitations, a reasonable assumption is that
they intend optimality. This would be a rational decision since limitations have already been
filtered. These decisions could be modelled by means of the logit model using an infinite
gamma (see ahead), or simply assuming that the decision maker take the best option, under the
information available to him.
For the evaluation of investment decisions, it is necessary to restate the concepts of annual
equivalent costs, net present value, monthly equivalent cost, operational costs, and investment
costs, among others.
A model that considers limited rationality has the advantage that it allows to focus on policies
directed to both desires and opportunities.
Policies intending to stimulate opportunities:
Policies intending to stimulate opportunities are easy to detect and implement as stated by
Elster (1990): “Opportunities are external to individuals, while desires depend on him". It is
easier to change people's opportunities than transforming their way of thinking. Policies intend
to increase opportunities, attacking the causes that restrict them. The different policies that can
increase the elements included in the opportunity sets are now listed:
* Set of reduced opportunities: Research and development of new technologies.
* Set of opportunities known to user: Popularisation or marketing.
* Set of real opportunity: Improvement of distribution channels of technologies.
¢ Set of economic opportunity: financing, soft credits.
Policies intending to stimulate desires:
There is no use for an extensive opportunity group if people have limited desires. To focus on
policies to stimulate desires it is necessary to consider people’s habits, then trying to modify
them. This could be accomplished by means of education or by demonstration (pilots plans).
Changing habits may be achieved by children education.
5 SD MODELLING AND OTHER MODELLING APPROACHES
To show how methodologies differ we will show alternative representations for decision
making under both bounded rationality and full rationality, using: The Logit Model and
Financial schemes. The problem under consideration is market split between incandescent and
fluorescent compact bulbs.
A systems dynamics model was constructed (figure 6) and different decision making
mechanisms were investigated for comparison purposes. Next we describe in brief the
mechanisms under examination.
i)
‘Aucilary_8
Awdlary_7
Decision making
Mechanisms
Figure 6. Model diagram.
Optimisation: If optimisation is used compact fluorescent bulbs would immediately take the
whole market, as they have lower annual equivalent costs. This result is obviously impossible
because of production, technological and cultural barriers.
Logit Model: The Logit model is used for decisions making analysis when the annual
equivalent cost of an alternatives is evaluated as a percentage of the annual equivalent cost of
other technologies. The model diagram is shown in the figure 7 and is represented as follows:
“—
pp =_CAE
Stage PMs Tanah
Logit
Mechanism
Auxiliary_9
GAMMA Annual equivalentcost
Figure 7. Model’s diagram using logit for decision making
The results obtained with the Logit model can be appreciated in figure 8. The market share of
compact fluorescent bulbs is larger than the one of incandescent bulbs, but do not penetrate the
whole market.
00 is 3 + £
50 4 +—Market Size
-2—F. Compact Bulbs
-3— Incandescent Bulbs
0 + + + + 1
0 20 40 60 80 100
Time
Figure 8. Technology propagation of light bulbs using the logit model
Financial decisions: When modelling financial decisions, such as net present value, one could
determine the discount rates at which compact fluorescent bulbs would be most attractive. The
developed model uses the same principle as the logit model, however net present value is used
instead of the annual equivalent cost, to incorporate the concept of risk, via discount rates (see
figure 9).
‘consumption_inc
FINANCIAL
DECISIONS CORE
hours per month
ectrical tarit
Figure 9. Financial Decisions mechanism
Figure 10 shows model results when financial decisions are used, for discount rates of 30% and
200% respectively. Results are as expected: to lower discount rates, larger penetration of the
corresponding technology. These discount rates could be modify by means of educational
programmes.
100.
1007 4—
80.
2 a
, 6 2
5 g F
2 50} ae |] Fe
—y— ine 40 é —> —y— ine
20.
of o> J
é za it 0 50 100
Time Time
Figure 10. Results of the financial model for a rate of discount of 30% and of 200%,
respectively.
When considering technology diffusion mechanisms (on top of financial decision), where
individuals are just followers, and never take their own decisions, we have the model of Figure
11.
Pop. followers Decision Making Population
© &
Growth propagation
Percentage population
with decisions
Figure 11. Decision propagation model
Model results can be appreciated in the figure 12. These results contrast the findings exhibit in
figure 10. This is due to the fact that only a small percentage of the population is capable of
making decisions because of their lack of knowledge or fear of change. However, we expect to
experiment a propagation process that will help defeat the incumbent technology.
100:
100 4 aie:
. \
2
60.
io)
e al afc p ajekt
ine 404 ine
Time Time
Figure 12. Technology propagation when 5% of the population's considers changing,
using discount rates of 30% and 200%.
6 Conclusions
By way of conclusion one can say that results are encouraging and that further work may
contribute to help changing the vision of policy making with respect to demand side
management. This investigation may reveal the importance of using models at the final user's
level, rather than some large aggregate models that are often used for policy assessment
nowadays.
References
Adam B Jaffe and Robert N Stavins. THE ENERGY -EFFICIENCY GAP, WHAT DOES IT
MEAN? Energy Policy 1994 22 (10) 804-810.
Bunn, D and Dyner, I. Systems simulation to support integrated energy analysis and liberalised
planning. International transactions in operational research, pp105-115, Vol 3, No 2, 1996.
EIA.”U.S. Electric Utility Demand-Side Management 1996”.
Elster, Jon. Tuercas y tornillos, una introduccion a los conceptos basicos de las ciencias
sociales. Ed. Gedisa. Barcelona. 1990a.
Elster, Jon. When Rationality Fails. The Limits of rationality. Edited by: Cook, Karen. 1990b.
GATELY, DERMOT (1980). “Individual discount rates and the purchase and utulization of
energy-using durables: Comment”. Bell Journal of Economics, 11, 1, 373-374.
HAUSMAN, JERRY (1979). “Individual discount rates and the purchase and utulization of
energy-using durables.”. Bell Journal of Economics, 10. 33-54.
Howart, Richard B. and Bo Andersson. MARKER BARRIERS TO ENERGY EFFICIENCY.
Energy Economics, October 1993.
Leon, Orfelio G. Analisis de Decisiones. Ed. McGraw-Hill. 1994. Impreso en Espaiia.
Morecroft, Jhon DW, The Int. JI of Mgmt Sci. Vol. 11. No 2, pp. 131-142. 1983. Great
Britain.
Simon, Herbert A. El comportamiento A dministrativo, estudio de los procesos decisorios en la
organizacion administrativa. Ed Aguilar Argentina S.A. Buenos Aires, Argentina. 1978.
Sioshansi, Fereidoon P. Demand Side Management, The Third Wave. Energy Police, Vol 23,
No 2, pp 111-114, 1995.
Schweers and Levi, Margaret. 1990. The universuty of Chicago Press.
Sun Tzu. SUN TZU ON THE ART OF WAR. 500 BC. Translated from the Chinese with
Introduction and Critical Notes BY LIONEL GILES, M.A. First Published in 1910.
http://www.kimsoft.com/polwar.htm.
CC BY-NC-SA 4.0