ENHANCING CASH FLOW FORECASTING
BY THE USE OF SYSTEM DYNAMIC MODELLING TECHNIQUES
Chris Smart
Virginia Bryant
City University, School of Informatics,
Department of Business Computing
Northampton Square, London EC1V OHB, England
Tel: +44 71 477.8419/8409
Fax: +44 71 477 8586
Email: c.smart@uk.ac.city
ABSTRACT
This paper examines the use of a systems dynamics modelling technique to enhance the contribution made by
cash flow forecasts to decision makers’ mental models. It is argued that by making explicit and accessible the
dynamic complexity in cash flow relationships, systems dynamics can provide valuable insights for decision
making purposes. By permitting the exploration of behavioural responses to perceptions about the financial
position of the business, a richer picture of the decision outcome is developed leading to changes in decision
makers perceptions about the riskiness of a proposed course of action. A case study of a commercial
organisation is used to illustrate these insights.
INTRODUCTION
Many important business decisions are based on forecasts to evaluate the potential financial impact of
alternative courses of action on the organisation. Forecasts are usually based on analysis of static cash flow
models, consistent with accountancy and financial management practice. The dominant accounting revenue and
cost.behaviour models depend on simplifying assumptions, such as the classification of costs as fixed or
variable relative to changes in volume or activity (Horngren, 1977) and of output increasing/reducing by some
ntage. Spreadsheet models frequently reflect these. model limitations. In reality very few variables actually
jave as indicated by their accounting model relationships. It is commonly accepted that dynamic relationships
exist between many of the factors to be considered but forecasts based on static cash flow models cannot be
used to represent dynamic behaviour explicitly. Consequently dynamic behaviour is normally dealt with by
informal assessments of the causality and of the estimated impact of a change (Franks, Broyles and Carleton,
1985). This may be deemed adequate for predicting outcomes fom ongoing repetitive operations, as experience
can be drawn on to refine the model and for inter ion of the results, but is considered to be less satisfactory
as a basis for analysis where such experiential influence cannot exist. 5
THE PROBLEM
Decision making requires the development of ‘mental images - models of how an organisation and its
environment functions’ (Mintzberg, 1973). Decision makers often synthesise forecasts to create mental models
of alternative futures...Forecasts which focus on.cash flows are used to determine that the funding requirements
for each of the proposed options are in line with. acceptable liquidity levels and borrowing ability, together
termed ‘solvency’, and to determine the financial worth of the option in terms of the present value of future cash
flows. Cash flow, long recognised as fundamental to survival, is increasing been seen as central to decision
analysis. with the assumption underlying ‘shareholder value analysis’ (Rappaport,1985) that a ‘business is
worth the net present value of its future cash flows' (Barfield, 1991) .
Decision makers acquire an understanding of possible outcomes by developing mental models based on their
experience of similar or analogous situations, often supported by analytic models which vary in terms of their
ability to represent behavioural aspects of the decision. scenario. Static, spreadsheet based, financial models are
widely used for informing decision makers. Such models have limited capabilities for the explicit
representation of interrelationships between factors, leaving the decision makers to rely on their domain
knowledge and past experience to predict the behaviour arising from interrelationships. In the case of a major
investment decision for a small or medium-sized firm the scenario may have elements which are novel to the
decision makers. In order to develop an appropriate mental model they need to test.their assumptions about the
behavioural aspects of the decision scenario. In order to determine how systems dynamics modelling could be
used to explore behavioural assumptions where the significant outcomes are in terms of cash flows, we studied
a small firm who had recently been faced with making a financial decision. Adopting the interpretivist paradigm
we accepted the decision mi ' view of what the issues were and the use of solvency as the decision criteria.
SYSTEM DYNAMICS '93. 495
CASE STUDY
The client is a small partnership providing legal services. They occupy a leasehold building in the centre of
London. One of the partners had expressed a desire to expand the parmership by acquiring additional members
of staff and recognised the need for additional office space to accommodate them, another was keen to improve
the image of the firm by acquiring a more prestigious building, while a third partner wanted to take advantage
of the slump in property prices to acquire a freehold building as a’ long term investment for the ership.
Methods of acquiring additional space had been explored on several occasions. Reviews of the office property
market revealed many offices for sale and for rent, however, the freehold office buildings for sale would involve
capital costs obviously beyond the firm's borrowing ability, while most rentals were at substantially higher
costs than the existing premises, Towards the end of 1991 a freehold property was identified in the same street
which was priced to reflect the fact that it would require extensive building work, taking about 6 months to
make it usable as offices, It was established that sufficient capital for the purchase and some building work
could be borrowed by the partnership. However under the terms of the partnership agreement, the decision to
go ahead with this purchase must be unanimous, and one partner objected to the purchase on the grounds that
the partnership did not have sufficiently high predicted fee income to cover the costs which would result from
the proposed purchase. He felt that the slump in demand for property made the disposal of their existing leases
unlikely in the short term, requiring rental payments to be continued on these until expiry. He felt that there
was a high Probability of reaching the limit of their borrowing ability before enough cash could be generated
from their fees to ‘return the partnership to solvency. Since partners in a partnership do not benefit from the
limited liability enjoyed by incorporated businesses he felt that the risk was too great.
Much argument ensued, but the other partners could not commit themselves to higher billing targets and so he
would not a agree to the purchase. This placed considerable strains on the relationships between the partners
since some of them felt that if they went ahead with the purchase, they would be motivated to obtain new
business, render higher bills and keep other costs down; things which would all be in the long term interests of
the partnership.
Spreadsheet based cash flow forecasting was undertaken to determine the maximum amount of the funds
required and the point at which the cash flows would start to recover. These were based on various levels of fee
income resulting from both realistic and optimistic targets. All the spreadsheets showed that there would not be
sufficient cash generated by the ordinary activities to support the purchase and that the partnership would
exceed, or be uncomfortably close to, its borrowing ability for the foreseeable future.
However the partners who supported the purchase would not accept the decision to abandon the purchase.
They argued that the spreadsheet models did not adequately reflect how they would respond to being put in
the position of having to find more cash from customer work. They felt that they would be motivated to find
extra work, which would lead to more billing in the short term and to expansion from repeat business in the
longer term. Perceptions of a poor financial position could also raise the levels at which quotations and
subsequently bills are set for some types of work. Some partners also argued that they would reduce costs by
avoiding unnecessary expenditure. They felt that many of these points would make the difference between
success and insolvency, but that none of this was shown in the spreadsheet models.
The dissenting partner would not agree to the purchase and serious arguments ensued about the future of the
partnership. The decision, to purchase the new building or not, appeared to be borderline, since four partners
were for it and one against, all on the basis of the same analysis and shared knowledge of the significant
factors. The spreadsheets provided a focus, but not a basis for a decision. It was felt that this example could be
used to explore the role of the synthesis constructed by the decision makers of their own behaviour in response
to feedback about the perceived financial position of the firm. The factors were reviewed and a systems dynamic
model constructed centred around the cash balance and their likely behaviour in response to it.
Purchasing the property would require the following expenditure:
450000, @ (Ona
Property Purchase £ arter 3 of the simulation)
Building Costs £ 92000 Q3
£123000 fort
Other Costs £ 10000 Q3
This expenditure would be covered by:
Loan F569500 fox)
Partner Capital Injection £100000 Q3
The decision criteria for the client was whether the purchase would push borrowings beyond or uncomfortably
close to their residual borrowing ability of #100000 for more than a year. This would bring their total
borrowings to more than #750000 a level at which the client would be risking insolvency. This decision
496 SYSTEM DYNAMICS '93
criteria was determined by the attitude of the client's bank. The bank would lend to the client beyond their
Poon pal borrowing ability provided that they were convinced that borrowings would quickly return below
vel
The rest of the paper describes what was revealed by the system dynamic modelling process and the client's
responses to the models developed
MODEL DEVELOPMENT AND OVERVIEW
The model was developed with the client in three stages. The client had no experience of systems dynamics
approaches and needed to be satisfied that a systems dynamics model could represent their existing
understanding of the partnerships cash flow behaviour as exhibited by their spreadsheet models, The first stage
model was therefore built to mimic the behaviour of their spreadsheet models. During this stage the client
Tealised that a systems dynamic model could represent behaviour resulting from interdependencies and delays
better than the spreadsheet model. A second stage model was built, essentially an elaboration of the first stage
model, which included these features. During the third stage feedback loops were examined, incorporated into
the model, and decision alternative scenarios explored.
The development of the systems dynamics model from a static spreadsheet model by a process of slowly
unfolding the behavioural features of the clients cash flow system was successful in maintaining the credibility
of the model with the client. The second development stage-of allowing the client to elaborate on the static
model before explicitly focusing on dynamic behaviour brought to light consequences of the purchase decision
which had not been considered in earlier analysis, such as the loss of partner time to manage the project and
interest rates increasing as the overdraft increases in size. As each elaboration of the model increased costs the
client was ready to explore how they could influence the behaviour of the cash flow system and the transition
to the third stage, examining feedback loops, followed naturally.
The systems dynamics model was built using i-think v1.01. The basic structure of the main sector of the model
is shown considerably simplified in figure 1. The client bills customers for work done and cash is received
over the next three quarters, less 3% bad debts. The market in which the client operates is characterised by a
e number of small partnerships of similar size to the client. Predictable business growth occurs organically,
estimated by the client at 3.5% p.a., or by acquiring fee earning staff who bring business with them.
‘Overdraft Payments
Figure 1: Cash Sector, considerably simplified
SYSTEM DYNAMICS '93 497
The legal profession is an “information” business and personnel (staff salaries and partner drawings),
accommodation and taxes are major costs (75%). Operating-expenses, excluding accommodation costs, are
driven by business volume, drawings are largely determined by existing personal financial commitments,
and the main business tax (known as Schedule D tax) is determined by partnership profitability.
Many of the significant cash flows have delays associated with them. Receipts are delayed. by up to three
quarters. Schedule D tax is based on the previous financial year and paid biannually. Increases in operating
expenses are delayed by a year and VAT by one quarter.
The.model contains four feedback loops as shown in Figure 2.
Overdraft Expenses
Payments
* &.
A “
cad Perceived
cash Financial
Billing ———> —— Position
Soe
Figure 2: The main feedback loops
L1) Cash - Overdraft Payments - Cash
This feedback loop cannot be directly influenced by the client, but has a significant effect on cash flow”
behaviour. A positive loop, any negative cash position is compounded by interest payments, the interest rate
increasing as the cash position worsens.
2) Cash - Perceived Financial Position - Expenses - Cash
The client believes their perception of the financial position of the partnership, as indicated by the cash,
position, will direct them to exercise greater cost control over expenses, a negative feedback loop. Their ability
to reduce costs, however, is highly constrained. Discretionary spend is a small percentage of operating costs
and the client considered the maximum practical reduction in operating expenses would be 5% of the total
operating costs excluding salaries and accommodation. The client's perception of their own financial position
was discussed in great detail and a number of alternative models explored.. Based on past experience it was
finally aj that a cash position of £-20000 causes concern within the partnership and initiates corrective
action. degree of concern is proportional to the ratio of negative cash to the total borrowing ability of the
pee. The total borrowing ability of the partnership is determined by the value of unsecured assets held,
yy the Pannen: The reduction in expenses is not linear with increasing concern as each additional saving is,
harder to find.
L3) Cash - Perceived Financial Position - Billing - Cash
Customer billing in the legal profession has a significant subjective element. Closer scrutiny of work carried
out can uncover legitimate justifications for increasing customer bills. It was estimated by the client that.
customer billings could be increased by a maximum of 10% by this process. However, this measure should!
only be undertaken when the financial position of the partnership is poor as they would then be billing at’
higher rates than their competitors. This negative feedback loop was implemented using a similar control
mechanism to the expenses feedback loop described above, the only difference being the ceiling imposed on the
effect of the feedback,
498 SYSTEM DYNAMICS '93
LA) Cash - Perceived Financial Position - Billing - Cash
A second billing feedback loop was identified which reflected the partnership's extra efforts to attract new
business. When financial concern is triggered the partners will try and make new customer contacts. Over time
some of these contacts will become regular customers to the extent that billing expectations can be raised to
include ’extra work from them. After much deliberation it was concluded that this effect was independent of the
internal level of concern within the partnership, but a function of effort put to finding new business. It was
estimated that each Leese that financial concem was triggered sufficient new customer contacts would be made.
to translate into a 1% increase in business volume after one year.
Alternative Scenario
Although the client had no immediate plans to increase the number of fee earning staff, if they found an
individual who would fit into the partnership, and would bring new business with them, they wor
employing them. The client felt that recruiting such an individual was a viable option if the partnership needed
to significantly increase business volume. This option was explored as an alternative scenario with the
following agreed characteristics. It is likely that any new fee earner would not join until Quarter 6 of the
simulation, after building works had been completed and the partnership had moved to the new premises.
Salaries: would be increased by an extra £10000 per quarter, effective Q6. The new fee earner would bring
‘cumulative business volume it increases of £10000, and £8000 in Q7,Q8 and Q9 respectively.
MODEL FINDINGS AND INTERPRETATION.
Figures 3a to 3h show the client's cash position simulated over 20 quarters for the decision scenarios. All
feedback loops ite as discussed above except where detailed below. The client's residual borrowing ability
is a constant £1 after purchasing the property. The quarterly oscillation in cash position is due to the
biannual payment of Schedule D taxes,
Figure 3a shows the cash position assuming the property is not purchased and suggests there not any intrinsic
strains on their cash position.
Figure 3b shows the cash position assuming the property is purchased incurring the additional costs detailed
above. This suggests that if they do purchase the new property, borrowings will be pushed to close to the
maximum for 5 quarters after Q7 and thereafter slowly climb out of danger. Although the client expected
borrowings to be high for some time, their expectation had been for a much swifter recovery to the borrowing
levels shown in Figure 3a. In order to examine whether a faster recover could be achieved the relative impact of
each feedback loop was explored.
Figures 3c to 3e assess the relative impact of feedback loops L2 to L4, Figure 3c shows the relative influence
of feedback loop L2 by comparing the reduction of operating costs by different percentages. The client
estimated the maximum practical reduction in operating expenses would be 5% of the total operating costs
excluding salaries and accommodation. This figure shows the effect of reductions of 1) 0%, 2) 2.5% and 3) 5%
and indicates that although expenditure control will reduce borrowings by a small amount it will not speed
recovery. Figure 3d shows the relative influence of feedback loop L3 by comparing the effect of increasing
customer bills by different percentages. The client considered that the subjective element of customer billing
would enable them to justifiably increase customer billing by a maximum of 10%. This figure shows the effect
of increases of 1) 0%, 2) 5% and 3) 10%. This result indicates that feedback loop L3 is critical. Increasing
customer bills by 5% or less risks the positive loop L1 becoming dominant and the cash position spiralling out
of control through ever increasing overdraft payments.
Figure 3e shows the relative influence of feedback loop L4 by comparing different success rates in attracting
business from new customers. The client estimated that each sustained effort at makin, ng ne customer contacts
would translate into an extra 1% business volume after 1 year. This figure shows the effect of 1) no extra
business volume, 2) 1% extra business volume and 3) 2% extra business volume. This result shows that
feedback loop LA is also critical though the deterioration of cash position is not as dramatic as is the case when
L3 is neglected. Unlike L2 and L3 the client could not establish a ceiling for this feedback loop and this
figure shows that increasing the volume of new customer business to 2% per quarter still involves a recovery
period of over 2.5 years. Since Met: presced growth without this feedback loop is less than 1% per quarter,
an additional 2% is unlikely to be achieved. It also became apparent that irrespective of the volume of new
customer business the impact of this feedback loop does not result in fast enough recovery.
Comparison of Figures 3c, 3d and 3e shows that feedback loop L3, Figure 3d, has the dominant influence in
the critical period of 2 years following the purchase. Although effective in the short term the dominance of L3
has serious long term implications for the client's business. Increasing billing by squeezing more fees out of
SYSTEM DYNAMICS '93 499
00s
€6, SOINVNAG WALSAS
200000.00:
|
-200000.00 +
0.00 8.00 10.00
Quarters
Figur 3b: Purchasing New Premises
200000.00.
200000.00:
0.00 5.00 10.00
Figure Se: Expenses Sensitivity Analysis (1=0%, 22.5%, 9:5%)
20.00
0.00 5.00 10.00
Quarters
Figure 3d: Customer Billing Sensitivity Analysis (1=0%, 2=5%, 3=10%)
€6, SOINVNAC WALSAS
Los
200000.00: 2000.00:
CASH 0.00 CASH 0.005
i H i
-200000.00 + | -200000.00 T ¥
0.00 8.00 10.00 15.00 20.00 0.00 5.00 10.00 15.00 20.00
Quaters Quarters
Figure 3e: New Customer Sensitivity Analysis (1=0%, 2=1%, 3=2%) Figure 3{: Additional Fee
Earner
eongon.o9 200000..00:
CASH 0.00 CASH 0.00% nnd nian
L
"
-200000.00 - T i -200000.00 T T
0.00 x 10.00 15.00 20.00 .00 5.00 10.00 15.00 20.0
Quaters Quarters
Figure 29: Additional Fee Earner, Customer Billing Sensitivity
Analysis:
ieee ‘Sh: Additional Fee Earner, Alternative Control mechanisn
the same work would place the client at a cost disadvantage compared to their competitors. This would not be
x } in the short term as customer relationships and service differentiation would be strong enough to
avoid losing business. However, the model suggests that increased billing in this way will have to be
maintained for at least 3 years by which time the client would be competitively very vulnerable.
The findings described above all show a cash position recovery which was too slow for the client. The main
alternative decision scenario was recruiting an additional fee earner under the conditions described in the model
overview. The model was modified to incorporate this alternative scenario and Figures 3f and 3g show the
impact of the additional fee earner. Figure 3f is equivalent to Figure 3b in the base scenario, Comparison of
Figures 3b and 3f shows a deeper trough but faster recovery with an additional fee earner. In the first quarter
after the new fee earner is recruited they are an extra cost to the client worsening the cash position, Over the
next three quarters the new fee eamer increases total business volume by £26000 per quarter accelerating the
recovery. rate. The cash position behaviour shown in Figure 3f was close to the clients perception of an
acceptable recovery rate, but they were still concerned about the effect of L3 on their long term competitiveness.
Under this scenario they would still have to squeeze customer bills for 2 years, an unacceptably long period of
time. Figure 3g, however, shows that the recovery is still very sensitive to increasing customer bills.
In order to explore whether they could reduce this period an alternative control mechanism for L3 was
implemented. The client felt that the maximum period of time they could squeeze customer bills without
serious long term competitive damage was 1 year. After discussion with the client the control mechanism of L3
was modified so that bills were increased for only four quarters from Q5 and by 10% irrespective of their cash
Bosition. Figure 3h shows that with this alternative control mechanism on L3 recovery is once more slow and
painful.
‘There was a danger that by relying on the billing squeeze they could find that they were jeopardising their
future billing capability. Some customers would respond to higher fees by taking their business elsewhere,
forcing the firm to continue with greater billings squeezes and/or billing squeezes for a longer period. For a firm
who regard their market niche as providing commercial legal services, i.e. ‘city services’ at lower than ‘city
prices’, the risk of being regarded as an expensive supplier was unacceptable. By representing these
interrelationships explicitly, the fact that the purchase would require them to engage in policies that contradict
their strategy became clear. The risk of the short term fix becoming a damaging necessity had not occurred to
them until this was revealed by a systems dynamics model.
It became ay it to them that recovery within an acceptable time period depended on expanding by taking on
anew member of staff and by implementing a substantial and sustained billings squeeze of about 10% . It also
became apparent that these actions would not bring the level of indebtedness down to an acceptable level unless
they occurred (new member of staff) or started (increases in billings) sooner than they considered possible or
had thought necessary.
Finally, the systems dynamics model made clear the size of the business expansion necessary to effect recovery.
Since much of the argument had centred around their beliéfs about the possibility of finding new business, and
this was the area most explored by the spreadsheets, it was surprising to find that the increases required for
recovery was (at about 8,25% pa ) more than twice their realistic estimate of 3.5% pa. The client learnt that
their financial recovery from purchasing the new premises was reliant on the long term application of a short
term measure. Exploring their assumptions in this way changed their perceptions about the riskiness of what
they were proposing to do; from acceptable risk to unacceptable risk,
DISCUSSION.
Systems dynamics modelling enabled the exploration of the decision maker's behavioural assumptions
concerning their ability to take actions which would steer the firm towards back to solvency within an
table time period . The insights gained changed the decision inakers-perecptions about what would be
required to meet the cash flow needs if the new building was purchased. They had believed that it would be
enough to just squeeze expenses (downwards) and billing (upwards) for a few months. This view had been
strongly held. The nature of the spreadsheet models had allowed them to retain these beliefs, since the relative
capabilities of the possible policies identified were not apparent. By exploring the relationships between the
feedback loops, it was discovered that they had been planning to rely on a billings push policy that could have
Jed them to develop behaviour consistent with that of the system archetype ‘fixes that don't work’ (Senge,
1990). By demonstrating the interrelationships between these policies in terms of expected cash flow, the
systems dynamics modelling made the decision makers realise that it would be unwise to depend on squeezing
costs and on pushing billing as instruments for recovery. This realisation meant that they were forced to move
away from relying on actions which required relatively small changes to well understood processes, and
towards considering actions which required more radical changes, as a basis for recovery.
502 SYSTEM DYNAMICS '93.
In order to provide appropriate tool support for decision making, (Alter , 1992) has proposed that, rather than
focusing on the technology, the decision making process pa to be better understood. The way in which
mental models of the alternative futures resulting from decision options are developed by decision makers is
central to this understanding. The systems dynamic approach supported the exploration of the decision makers’
assumptions about the nature of behavioural responses, without challenging the validity of those assumptions.
This was important in obtaining commitment to, and belief in, the systems dynamics models. We found that
despite understanding the nature of the behavioural assumptions on which their mental models were based; the
decision makers had difficulty with synthesising, for the purpose of mental model development, the effect of
these behavioural assumptions in terms of impact on cash flow. The spreadsheets did not show any recovery
but, by continuing to support the purchase, the decision makers were overriding the spreadsheet forecasts and
responding on the basis of their mental models. The strong role played by mental models in determining
decision making outcomes, highlights the need for tools to aid in the development of richer mental models.
By enabling the exploration of the impacts of complex behavioural interrelationships, the systems dynamics
modelling used here contributed to the development of richer mental models in the decision makers studied, for
the purpose of the decision outlined in the case study.
We were interested in, testing. the use of systems dynamics modelling for exploring the behavioural
relationships where the impact, in terms of corporate solvency, of a decision was being considered. The possible
effects of a decision on long term solvency was not amongst the generic corporate problem modes identified by
Graham, (1988). However, for long range forecasts, the solvency profile is more important than the individual
figures. System dynamics has been suggested as particularly suitable for insight into the co-ordination of
fing policies, (Graham, et al, 1990), and for exposing counterintuitive behaviour in systems (Sterman,
1589). ‘e thought that it would be interesting to consider the role of ‘systems dynamics in enhancing
intellectual manageability for one-off/occasional financial decisions in small organisations; perhaps leading to
the development of tool/method support specifically for that environment. Much effort has been expended in
designing suitable decision support systems for the repeating decision e.g. bank lending, but, understandably in
view of their nature (and the adoption of spreadsheet tools) , occasional financial decisions have received less
attention from the decision Sapport systems community. Following on from this study we intend to consider
the development of method and tool support for occasional financial decisions based on the systems dynamics
modelling approach.
CONCLUSION
By developing a systems dynamics model to support decision making around a cash flow recovery profile , we
have illustrated ‘the application of systems dynamics modelling, as an alternative to the widely used
Sprache We see this as demonstrating the usefulness of the approach to financial management, where
ifficulties in developing forecasting models which incorporate complex behavioural responses has been
recognised as a factor limiting their value in the development of appropriate mental models,
REFERENCES AND BIBLIOGRAPHY
Alter S, 1992. Why Persist with DSS when the Real Issue is Improving Decision Making 7
In: Proceedings of IFIP8.3 Working Conference on Decision Support Systems: Experiences and
Expectations,
Barfield R. 1991 Shareholder Value: Managing for the Long Term. Accountancy. Oct. pp100-101.
Checkland P, 1985. From Optimizing to Learning: A Development of Systems Thinking for the 1990s.
Journal of the Operational Research Society. Vol.36, No9. pp757-767
Faber J and Scheper W. 1992. Taming the Decision Tide Within Organisations: some theoretical considerations
ad ud expenmental results. In: Proceedings of the 1992 International Systems Dynamics Conference.
iL.
Forrester Hee Beyond Case Studies - Computer Models in Management Education. MIT Publication D-
Franks J. Broyles J. Carleton W. 1985. Corporate Finance: Concepts and Applications. Kent. Boston.
Graham A. 1988. Generic Models as a Basis for Computer-Based Case Studies. In: Proceedings of the 1988
International Systems Dynamics Conference. La Jolla. CA.
SYSTEM DYNAMICS '93 503
Graham A. Senge P. Sterman J. and Morecroft J. 1990. Computer-Based Case Studies in Management
Education and Research. MIT Publication D-4039.
Homgrea C. 1977. Cost Accounting: A Managerial Emphasis. Prentice Hall International.
Kim D. 1990 Toward Learning Creanisotions: Integrating Total Quality Management and Systems Thinking.
MIT Publication D-4036-4.
Lane D. 1992. 1952! Modelling As Learning: A consultancy methodology for enhancing learning in managment
teams. European Joumal of Operational Research.
Machuca J. 1992. Are we losing one of the best features of systems dynamics ? Systems Dynamics Review 8
No.2 ppi75-177.
Mintzberg, H. 1973. The Nature of Managerial Work. Harper and Row. New York.
Morecroft J. 1988. System dynamics and microworlds for policymakers. In: European Journal of Operational
Research. 301-320,
Rappaport. 1985. Creating Shareholder Value. Free Press.
Senge P. 1990, The Fifth Discipline. Doubleday. New. York.
Sterman J. 1989. Modelling Management Behaviour: Misperceptions of Feedback i in a Dynamic Decision
Making Experiment. Management Science. Vol. 35. No.3.
Molaeaboine E. 1982. System Dynamics in Perspective. Journal of Operational Research Society. pp547-
Wolstenholme E. and Coyle R. 1983. The Development of Systems Dynamics as a Methodolgy for System
Description and Qualitative Analysis. Journal of the Operational Research Society. Vol.34, No.7
pp569-581.
504 SYSTEM DYNAMICS '93
CC BY-NC-SA 4.0