MARKET ANALYSIS AND FORECASTING AS A STRATEGIC BUSINESS TOOL
James M. Lyneis and Maurice A. Glucksman
Pugh-Roberts Associates, Inc.
65 Charlbury Road, Oxford OX2 6UX, U.K.
L INTRODUCTION
Pugh-Roberts has developed a number of simulation models to forecast the demand for products in spe-
cific markets. These models contain key feedback relationships which create growth, decline, and cy-
cles in demand. They are unique in their integrated representation of macro-economic, micro-economic,
and regulatory factors. Their broad scope makes them powerful enough to specifically show the relative
importance of industry factors such as manufacturer pricing policies, inventory and production policies,
capacity expansion policies, and the timing of new product introduction, in creating or magnifying mar-
ket cycles. These models are highly valuable because, unlike simple statistical models, they explain the
Toot causes of cyclical behaviour and can therefore more accurately predict the timing and severity of
market cycles. As a result, the models provide valuable information regarding:
* timing of "booms" and "busts"
* required production capacity and production rates
* “windows of opportunity" for new product introduction
* relative growth of market segments
* relative importance of growth versus replacement demand
* relative importance of price, availability, and technology in determining demand.
and market share
* effects of macro-economic, regulatory, and/or political changes on overall demand.
patterns.
These market models have been used successfully in the aerospace, automobile, computer, container
shipping, financial services, industrial coatings, insurance, health care, and mail order retailing indus-
tries, so they are quite generally applicable. To illustrate their value versus statistical models, two of the
existing aerospace models are examined in detail. With emphasis on the "Macro/Micro" issues, the basic
causal structure is described and the importance of feedback is demonstrated with sensitivity tests. Gen-
eral uses, and examples of other potential applications, are also discussed.
IL DESCRIPTION OF THE MODELS
2.1 Causal Structure
Exhibits 1 and 2 show the pattern of order rates for aircraft in the North American helicopter market and
the European commercial jet market over the recent historical period. The models produce. historically
accurate, but different, behaviour in each case. In spite of these differences, the two models have com-
mon feedback structures. These are depicted in Exhibit 3. The narrow arrows are ordering cycle feed-
backs and the thicker arrows are profitability, regulatory and flight frequency feedbacks.
Starting at the left of Exhibit 3, demand is a function GDP (or one of its derivatives, personal income or
corporate profits), population, experience, and fares. GDP and population are exogenous inputs. Expe-
tience and fares are computed internally from factors within the rest of the system.
Demand growth is extrapolated by operators to produce a projected demand. Projected demand is con-
verted into a desired fleet based on a "desired utilization rate": for commercial aircraft, this is Revenue.
Passenger Kilometres per year per aircraft (RPK); for helicopters, this is Flight-Hours per year per air-
craft (FH). Desired utilization rate is an exogenous input.
Comparing desired fleet to the existing fleet determines the number of "growth orders". The fraction of
growth orders for new aircraft, versus used, depends on: new versus used aircraft prices, delivery delay,
technology, and regulations.
137
EXHIBIT 4
NORTH AMERICAN HELICOPTER ORDER RATES
Simulated Orders
— Actual Orders
---> Regression-generated Orders
= VA
250,
1976 1978 1980 1982 1984 1986
EXHIBIT 2
EUROPEAN JET AIRCRAFT ORDER RATES
100. KA He 4
LE TACT
AT (A INA ‘
treme
138
EXHIBIT 3
CAUSAL RELATIONSHIPS IN AIRCRAFT MARKET MODELS
Desired Utilization
(Passenger Kilometres
Deregulation; Preference for ee AC)
Congestion — Eroquoncy vs. B ria Now AC
cos
Desired
SL Feet “Grown as aie
action
‘Growth Orders logy &'
Experience Projected New Regulations
Demand
\ \ 2 a New Arerat Sb aivery
ay
Profitabi
Demand. 1 ars .
{Personal sts Backlo
ersonal Income, . s
Gorporate Pratt) yw, Mig. rr
Population Ay | nemenpe Floplacement &
Total Costs 7 trade-ins"
“Fixed” Costs
per RPK or FH, per RPK or FH )
Used AC
Variable Costs es. Prices
Flight per RPK or FH from Used og
Frequency Fleet \
Fuel & other
Sales
Input Prices oe
Orders for new aircraft are added to manufacturers backlogs and then delivered after a delivery delay
determined by production lead times and capacity. As owners of new aircraft "trade-in", these used air-
craft become available for satisfaction of growth demand, or are sold abroad. Finally, after the normal
lifetime of the aircraft, it is retired. Replacement and retirements are modulated by relative aircraft
prices, technological changes, regulations, and, as discussed below, operator financial condition.
As the size of the fleet increases, "fixed" costs per RPK or FH also increase. As demand increases, unit
fixed costs decrease. Variable costs depend on intensity of use (which is calculated internally) multi-
plied by fuel and other input prices, (which are exogenous inputs). The sum of fixed and variable costs
‘equals total cost, which is an important determinant of fares. Fares then feed back to affect derhand.
Revenues are determined by demand and fares, while costs depend on demand and total unit costs.
Revenues less costs equals profitability. Profitability affects orders for new aircraft, and both replace-
ments and retirements of older aircraft.
Two other important feedbacks are also depicted, The first is a feedback which says that as the size of
the fleet increases, flight frequency increases, and this acts to stimulate demand growth. The second is
an input which specifies the preference for frequency versus size. This preference determines the size of
the fleet required to serve a given level of projected demand. The Fesserce is a function of deregula-
tion and congestion. Deregulation encourages growth in the number of flights; congestion constrains
growth in the number of flights.
139
2.2 Level Of Detail
Within the structure discussed above, the model represents several demand and aircraft categories. The
categories are chosen to minimize the number of behaviourally distinct groups. For example, demand
in our commercial aviation model is divided among domestic, intra-European and inter-continental.
This selection is based on the idea that demand categories should reflect certain average flight charac-
teristics, passenger demographics and competitive factors. The selection of aircraft size categories, of
which there are five in the commercial aviation model, is influenced by the product positioning interests
of our clients, and by the selection of demand categories (some aircraft are unsuitable for certain appli-
cations or may serve specific market segments exceptionally well). Often the number of categories is
further limited by the degree of detail in the data. Similar objectives/constraints determined the demand
and aircraft categories in the helicopter model.
Categorisation of this kind is important for several reasons:
* Parameterisation is simpler and more "reasonable".
* Cross influences among categories can be explicitly included.
* Distinct markets and market niches are more accurately represented
* Tracking shifts in the distribution of aircraft types is possible.
* Diagnosing and explaining behaviour is simplified.
t Cycle
Before discussing specific examples it is useful to envision how a market cycle could evolve. Again,
following through Exhibit 3:
Suppose demand should increase. In the short term, this increase raises load factors, reduces unit costs,
and improves operating margins. Reduced costs lead to reduced fares, which further stimulate demand,
Operators extrapolate this demand growth into the future and conclude that they need additional aircraft.
Because their financial condition has improved, operators are more willing and able to order aircraft.
Improved financial condition also stimulates orders through early replacement of existing aircraft.
Given manufacturers’ delivery delays, operators experience several years of high load factors before the
new aircraft arrive. But when the new aircraft start to enter service, the dynamics start to reverse direc-
tion. An increase in the size of the fleet causes costs to rise and margins to decline. As costs rise, so do
fares, which in turn slows demand growth.
The combination of slowing demand growth and deteriorating financial condition causes operators to
cut back on new aircraft orders. But previously ordered aircraft continue to be delivered, unit costs rise
further, fares are increased, and demand growth slows or even becomes negative.
With reduced demand growth, the operators realize they have ordered too many aircraft. Moreover,
fares generally lag the increase in unit costs such that, with reduced demand growth, revenues do not in-
crease as fast as costs. Financial condition deteriorates further, operators cut back drastically on new
orders, delay orders for replacement aircraft, and delay the retirement of existing aircraft.
Given the backlog of aircraft on order and any overexpansion of capacity, several years of depressed
conditions can result before the situation turns around. But once it does, unit costs and fares will begin
to fall, demand increase, and another cycle starts.
I. VALUE ADDED BY A MICRO/MACRO MODEL
Models containing macroeconomic and microeconomic (industry) relationships add value over purely
statistical macro-models in two ways. First, they are better able to predict and explain the timing and
seventy Of industry cycles; and second, they offer sufficient detail in specific market segments to be
valuable for strategic decision-making (for example, capacity planning, product targeting, competitive
positioning). The predictive value added is illustrated with two examples in this section. The strategic
value added is discussed in section IV.
140
it factors in determining the demand for helleopters GNP, as a measure
is indicative of long term demand; oil prices drive offshore oil exploration and
production, which is a major market for helicopter services. Perhaps a simple least squares regression
employing GNP and/or oil price could explain the changes in helicopter sales. We tried various combi-
nations of GNP and oil price as the basis for a simple regression and found the best to be a “two factor"
model: Helicopter Sales = [GNP Growth lagged one year, Oil Price Growth lagged one year].
Regression produces a correlation coefficient (R2) with the data of 0.4, which does not inspire great
confidence. The time series output of this regression is shown on Exhibit 1. (We use R2 here because it
is sufficient for illustrative purposes. A more rigourous statistical test such as the Theil statistic, which
separately measures phase correlation, standard deviation and mean errors, would reinforce the conclu-
sions drawn here.)
The output of the simulation model (Exhibit 1) was subjected to the same statistical test. The correlation
coefficient in this case was 0.84. This much higher correlation coefficient coupled with the obvious
"visual" correspondence indicates significant explanatory power. Since GNP and oil price are the key
macro-economic inputs to the dynamic simulation model, the difference in correlation coefficients is a
measure of the "value added" by the more complex structure of the dynamic model. Some of the indus-
try specific factors accounting for the dynamic model’s superior performance are:
* Operators tend to extrapolate demand trends several years ahead when formulating
their investment plans.
* Helicopter manufacturers accept used helicopters as trade-ins and agree to be the
financiers of last resort.
* Advancing helicopter technology encourages replacement sales.
* Operator profitability reduces the intensity of use and increases the rate of
retirements,
* Helicopter aviation experience further stimulates demand for helicopter services.
The existence of feedbacks that seem to have an overriding influence on aircraft sales suggests that be-
haviour might be somewhat insensitive to changes in the basic economic inputs. If this is so then it
might have been possible to predict the collapse in the helicopter market.
To see whether the in sales could have been anticipated, we resimulated with different eco-
nomic inputs starting in 1980. We substituted two GNP scenarios: 1) continued steady growth in GNP
at 3.5% per year; and 2) a business cycle that declines less in 1981-82 and recovers much more
strongly during 1983-85 than, in fact, was the case. As seen in Exhibit 4 the basic pattern of collapse
and stagnation occurs, even when one assumes a steadily growing economy. Diagnosis reveals that the
principal causes are on the supply side and not from a major change in the final demand for helicoy
services. The collapse is primarily the result of overexpansion of the helicopter fleet in the 1971
period, and is only secondarily affected by the economic recession of 1982. Even if planners in the in-
dustry had been far too optimistic about GNP in 1980, they could have seen the collapse coming.
The same is true for alternative oil price "forecasts". We tried three scenarios: 1) oil price remains con-
stant through 1986 at the 1980 price level; 2) oil price rises significantly during the following year (to its
actual 1981 peak), and then remains constant through 1986; and 3) oil price peaks in 1981 Eat then de-
clines, but by only half as much as was actually the case. As seen in Exhibit 5 the differences in the es-
sential collapse and stagnation pattern are quite small. This reinforces the conclusion that the “supply
side" dynamics determine the basic behaviour, and that the market collapse was predictable!
2- ‘ial Aircraft in Ei
Orders for commercial aircraft were historically very cyclical. Referring back to Exhibit 2, the corre-
spondence between simulation and what actually happence is quite remarkable. Again, it could be sup-
posed that some very simple economic inputs could explain this behaviour, but as in the first example
this is not really the case. Exhibit 6 shows a comparison between three simulations of the same model.
‘THenry Weil and Richard Park helped to develop this example.
141
EXHIBIT 4
HELICOPTER ORDER RATES FOR VARIOUS GNP SCENARIOS
GNP - Historical (1 200-2000)
—— — GNP - Weaker Downtum (1200,2000)
am — GNP - Steady 3.5% Growth (1200,2000)
Orders - Historical (0,1000)
Orders - Weaker Downturn (0,1000)
2000, —— — Orders - Steady 3.5% Growth (0,1000)
1080.)
St
1800 S| \ \ 2
758.
°
1976. 1981. 1986.
TIME
EXHIBIT 5
HELICOPTER ORDER RATES FOR VARIOUS OIL PRICE SCENARIOS
———— Orders - Historical
—— —— Orders - Oil Price Constant at 1980 Level
— Orders - Oil Price Constant at 1982 Peak
cod ‘Orders - Oil Price Declines Half as Much as Historical
1976. Ase. 1986,
142
The first is exactly the same as the simulation results shown in Exhibit 2. The second shows what
would have happened if the key economic inputs had been smooth; these include GDP, fuel price, infla-
tion and interest rates. Clearly the basic cyclical behaviour is the same -- much as we would expect
given the conclusions drawn from example 1.
The third simulation shows what would have happened if all_cyclical economic inputs remained but
some key feedbacks are removed. In particular, we removed the influence of airline operating margin
‘on orders, the influence of changes in celery delay on projected demand, the influence of aircraft
“launch" announcements on orders, and the influence of changing fares (fares still follow the general
trend experienced in history but they are not as volatile), This results ina drastically different pattern of
behaviour. The overall quantity of aircraft ordered is about the same, but none of the severe peaks and
valleys in the order pattern are evident.
To be fair, in removing the above feedbacks we have eliminated some of the economic influences at the
same time. So it may be inappropriate to conclude that the internal dynamics of the aircraft market have
a vastly stronger influence on the pattern of aircraft orders than economic inputs. Nevertheless, it seems
clear that the internal dynamics are more important.
These examples of specific output from two of our aircraft models demonstrate that factors internal to an.
industry can be very important determinants of industry cycles. Furthermore, since the strength of
relationships and selection of time constants is validated with historical data, essential behaviour is valid
well past the historical period -- especially when external factors remain relatively stable. In fact, the
larger the influence of internal factors, the more reliable forecasts are. This is the value-added: robust
predictive power.
EXHIBIT 6
JET AIRCRAFT ORDER RATE SENSITIVITY TESTS
——— Simulated Orders ———— Orders Without Feedbacks
00 LI Wah Smostod Econom nputs
150. nN
\
. A\ Na,
N Sy * edie rts }
al ah me (ies.
op Ney
Te 7 76. 7. 80, 82, 84, 66.8
143
IV. USES OF THE MODELS
The overall objective in developing and using these market models is to support strategy formulation
and planning. These models allow companies: (1) to gain an understanding of the mechanisms at work
in a market; (2) to produce forecasts; and (3) to explore the sensitivity of forecasts to changes in as-
sumptions. As a result, they are extremely useful for exploring planning issues that involve timing
and/or market targeting. Examples of such issues for various potential users are:
Manufacturers:
* Product design -- addressing needs of emerging markets
* Product introduction -- timing during market upswings
* Product upgrades -- focusing on expanding markets, timing during market
Up Wings or 4 5
* Product phase out -- winding down production of potentially unattractive or
obsolete products.
* Sales support -- convincing customers that a product is designed to address their
future needs
* Capacity investments -- preparing for overall demand
* Production planning -- preparing for demand cycles
* Workforce policies -- having trained labour available when it is likely to be needed.
* Subcontracting policies -- efficient division of fixed and variable costs depending
on market cyclicality
Operators:
* Fleet soguisition -+ before peaks in demand/order cycles and/or anticipating fleet
mix shifts
* Fleet sales -- when used prices are relatively high and/or anticipating fleet mix
sl
* Fleet retirements -- early in market lulls
* Purchasing strategies -- predicting the availability of new versus used aircraft
* Maintenance strategies -- timing major maintenance before market upswings to
minimize downtime during periods of intense use
* Pricing policies -- impact on market cyclicality/growth
Investors:
* Buying low and selling high -- investing before market turns up, selling before
market peaks.
* Speculating on delivery positions.
Although the examples listed here refer to the aircraft industry, these same issues are important in other
markets as well. Some examples are:
* Offshore oil exploration and drilling equipment
* Farm equipment
* QOcean-going tankers and bulk carriers
* Commercial transportation equipment
* Construction equipment
* Automobiles and components
* Suppliers of raw materials or subcontracted products to any of these
* Insurance industry
these results show that feedback models have very significant advantages over statistical
models using equivalent economic inputs. Nevertheless, as in any modeling approach, there are certain
caveats. models are an abstraction of the real system and therefore cannot account for "everything".
In particular, although the general period and amplitude of market swings is very accurately predicted in
these models, precise timing and severity of market swings often depends on external events. Therefore,
this element of uncertainty must be addressed with sensitivity tests.
The primary purpose of these models is to support overall strategic planning, especially where timing
and market targeting are important. In this role these models have proven their value in many markets,
But apart for overall planning, a rich variety of other roles has emerged. These include consensus
building, competitive positioning, sales support, production/capacity planning, product design,
management training and investment support.
CC BY-NC-SA 4.0