Defining a problem or constructing a reference mode’
Khalid Saeed
Social Science and Policy Studies
Worcester Polytechnic Institute
Worcester, MA 01609, USA
saeed@ wpi.edu
April 1998
Abstract
This paper attempts to describe the process of defining a problem - called a reference mode in
Jargon - for building a system dynamics model. System dynamics models represent problems not
systems, hence a model cannot exist without first defining a problem. The characteristics of a
reference mode and how these might differ from a historical record are defined. A learning
model is used to the delineate steps entailed in constructing a reference mode, which would often
subsume past history as well as inferred future and which might also represent only a slice of a
complex historical pattern. It is proposed that a small number of archetypical patterns might be
developed to represent a large cross-section of problems encountered in experience. A
methodological task ahead is to crystallize these patterns.
Key Words: System Dynamics, reference mode, experiential learning, problem definition,
pattern recognition
“Helpful comments from Michael J. Radzicki over the course of preparation of this paper are
gratefully acknowledged.
Defining a problem or constructing a reference mode
Khalid Saeed
Social Science and Policy Studies
Worcester Polytechnic Institute
Worcester, MA 01609, USA
saeed@ wpi.edu
April 1998
Introduction
System dynamics models represent problems not systems per se. Therefore, the first step in the
modeling process is to define the problem the model will represent. This problem definition is
named the reference mode in Jargon. A reference mode is based on historical information and is
often described in a graphical form, although there are many misconceptions about what a
reference mode is. More often than not, a reference mode is perceived to consist of historical
data, whereas, historical data may only be a starting point for constructing a reference mode,
which is an abstract concept that must be developed very carefully from the historical data and
the inferred future trends it points toward. In fact, a reference mode is a fabric of trends
representing a complex pattern rather than a collection of historical time series. It may contain
concrete variables actually existing in historical data as well as abstract variables summarizing
qualitative information. I shall attempt in this paper to describe the process of constructing a
reference mode from historical data and illustrate this process with examples.
Reference mode represents a pattern of change
A problem is often perceived in a conventional sense as an existing condition which must be
alleviated. For example, in the economic development context, poverty, unequal income
distribution, high illiteracy rate, low infrastructure inventory and corruption are often defined as
problems. In a business context, low market share, overstaffing, poor quality, low productivity,
non-profitability, etc., are viewed as problems. In all such cases, the starting point for a policy
search is the acceptance of the existing condition defined as the problem. In system dynamics, a
problem is defined as an internal behavioral tendency found in a system. It may represent a set of
patterns, a series of trends or a set of existing conditions that appear resilient to policy
intervention. In other words, an end condition by itself is not deemed adequate as a problem
definition. The complex pattern of change implicit in the time paths preceding this end condition
would, on the other hand, represent the problem. Often, an end condition has not been reached
when a problem is being defined. In such cases, the future must be inferred through an intelligent
projection of the past patterns. Last, but not least, any time history may incorporate multiple
patterns experienced over a single time path, over different geographic locations or over different
periods of history. A penetrating problem description should collect these patterns so that a
simple model that contains the policy space for a pattern change can be developed to represent
the problem. Such a collection of patterns would constitute a good reference mode for building a
model.
Reference mode is an abstract concept
Even when treated as a historical pattern, a reference mode is often represented as a complex
time path, which can be a misleading way to start a modeling exercise. A complex time path can
often lead to a complex model that might track a history and project a future, but which may not
create any insights into the dynamics of the system under study since its behavior cannot be
understood. Any policy recommendations arising out of an examination of the projections made
by such a model are often disconnected from the microstructure of the model and constitute a
common sense grocery list of instruments, rather than an operational process that can be
implemented to modify existing decision rules. A reference mode, in fact, is not historical data, it
is an abstract concept that must be arrived at through a careful analysis of the historical data and
the future that can be inferred from it. Table 1 illustrates the differences between a reference
mode and its related historical evidence. While historical evidence can consist of time series data
or anecdotes, a reference mode is a pattem of behavior conceptualized from such evidence, and
the two may have quite different characteristics.
Table 1: Characteristics of historical evidence and reference mode
Historical Evidence Reference Mode
Quantitative Qualitative
Descriptive Intuitive
Complex Organized
Events Pattern
Past Past and inferred future
Isolated time series Integrated fabric
Noisy Noise-free
Snapshots Tendencies
System behavior Problem behavior
At the outset, while both historical behavior and a reference mode can be expressed in either
quantitative or descriptive terms, a reference mode is essentially a qualitative and intuitive
concept since it represents a pattern rather than a precise description of a series of events. A
reference mode also subsumes past history, extended experience and a future inferred from
projecting the inter-related past trends. It can be seen with the mind’s eye as an integrated fabric,
although it can be represented on paper only as isolated tendencies. A reference mode will also
not contain random noise normally found in historical trends, as it will represent a problem
behavior rather than the system behavior. A reference mode is an abstract concept far different
from the historical data and the qualitative descriptions concerning a problem. Let me illustrate
this with an example.
Say we are investigating the change in worker productivity in a firm undergoing a re-engineering
effort that would, in the first instance, increase profitability by decreasing the firm's workforce
and cutting down on worker cost. Part A in Figure 1 illustrates what might appear to be the
historical trends concerning profitability, worker productivity, and worker morale following the
re-engineering effort. Profitability rises due to reduction of workforce, but worker productivity
and worker morale do not. Part B of Figure 1 extends further these trends subsuming an extended
experience of the process. After all, if worker productivity and worker morale are stagnant, the
profitability increase will have to taper off after worker costs have been cut down as far as
possible. Also, a declining rate of profitability growth, together with the worker attrition
resulting from re-engineering, would affect the internal environment of the organization, which
would lead to an erosion of morale and consequently also of productivity.
profitability -------- worker morale — — — worker productivity
extended inferred future further inferred
case history experience future
Figure 1 Constructing a reference mode from past and inferred future
Extending the trends further, one would expect that the eroding morale and decaying
productivity would eventually create a stagnation in profitability and precipitate an inevitable
decline as the feedback between profitability, productivity and worker morale continues to work.
This is illustrated in parts C and D of Figure 1. Finally, the three trends in part D must be
visualized together as facets of a whole fabric to get an intuitive appreciation of the problem.
Viewed as disjoint trends, they will not tell the whole story. The conceptualization of the
reference mode as a fabric, therefore, leads to an abstract concept that goes beyond what can be
represented in a graphical form on a two-dimensional block. Fortunately, we have an immense
experience of visualizing such a fabric due to the constant demand made on our perceptions to
convert limited perceptual images of reality into more comprehensive mental images. For
example, a two dimensional vision frame that our eyes construct can be perceived as a three
dimensional mental image by our mind [Abbot 1987].
Learning process as a framework for constructing a reference mode
As I have pointed out in Saeed (1997), the conceptualization of a reference mode requires the
same learning process as the development of a dynamic hypotheses, the construction of a model,
the creation of the model understanding and the design of a policy for system improvement. I
have also proposed that these steps could be built around Kolb’s model of experiential learning
illustrated in Figure 2 [Kolb 1984, Hunsacker and Alessandra 1980, Kolb, et. al. 1979, Kolb
1974, 1998].
Concrete
Experience
Thinking
concrete
Testing Active Passive Observations
Implications Doing Watching | and reflections
abstract
Feeling
Abstract
Conceptionalization
Figure 2 Kolb’s model of experiential learning
Four basic faculties - watching, thinking, doing and feeling drive Kolb’s experiential leaming
cycle. For the learning process to be effective, watching must result in the careful observation of
facts, leading to identifiable organized pattems. These patterns, then, must drive thinking, which
should generate a concrete experience of reality. The implications of the concrete experience
must be tested through experimentation conducted mentally or with physical and mathematical
apparatuses. Finally, this experimentation must be translated into abstract concepts and
generalizations through a cognitive process driven at the outset by feeling, which would, in turn,
create an enhanced framework for careful observation thus invoking another learning cycle.
The learning faculties, according to Kolb’s model, reside in two basic human functions, physical
and cognitive; each integrated along two primary dimensions, which are also illustrated in Figure
2. The first dimension, conceming the physical functions is passive - active. The second,
conceming the cognitive functions is concrete - abstract. Thus, watching is a passive physical
function, thinking a concrete cognitive function, doing an active physical function and feeling an
abstract cognitive function. Since the mental construction of reality and its interpretation must
filter unwanted information, each faculty must be guided by certain organizing principles to
affect learning. Additionally, the leamer is required to shift constantly between physical and
cognitive domains to create opportunities for refuting the anomalies arising between the two and
thus reconciling mental images with the physical reality.
Figure 3 interprets Kolb’s learning model in the context of constructing a reference mode. One
must begin by carefully examining historical information, both quantitative and qualitative, in
the passive physical domain to discern patterns that it incorporates. This process is followed by
system identification in the concrete cognitive domain, which returns system boundary in terms
of the variables that must be considered to describe the discemed patterns. These variables may
or may not be the same as in the historical data. Some of the variables in the data can be
aggregated while others substituted by more abstract concepts, depending on the problem focus,
the time horizon of interest and the policy space considered. The time horizon of reference mode
depends on the purpose of the model, but it would invariably be longer than the historical
information it is based on as it would include also information about the inferred future.
Next, an experimentation process carried out in the active physical domain calls for drawing the
trends and intelligently extrapolating them into the future on the basis of the knowledge about
the system as illustrated in Figure 1. Finally, the drawn trends must be conceptualized as a
multi-dimensional fabric - an abstract concept that represents the reference mode, which can be
readily related to the information in the micro-structure domain for formulating a dynamic
hypothesis.
SYSTEM IDENTIFICATION
a) system boundary
EXPERIMENTATION OBSERVATION
@) graphing pattems active | passive. a) historical data
physical | physical
a) reference mode
CONCEPTUALIZATION
Figure 3: Learning process underlying the construction of a reference mode
An illustration
Some time ago, I prepared a background paper for the United Nations Economic and Social
Commission for Asia and Pacific (UN-ESCAP) on environmental trends and their future
projection [Saeed 1994]. Since projections were required for an extended period of time and
could not possibly be based on a simple trend extrapolation, I adopted the process of constructing
a reference mode to create intelligent projections. I’ll illustrate here how the historical data was
used to determine the food production pattern in the region covering past as well as inferred
future behavior.
The first task seemed to be to organize the available numerical data so it could give information
about the whole region. A complementary effort was also launched to review the related
qualitative information. Some 300 time series, covering fourteen selected countries representing
the Asia and Pacific region over the past three decades, were constructed from published UN
sources to serve as a data-base for the analysis. In view of the many missing cells, differences in
units and definitions of data categories, the variability in the national policies across countries
and the possibility of the data from one country dominating an aggregate trend, it was decided
not to aggregate country data into any regional categories but to examine closely the selected set
of countries as a sample for comparable resource- and environment- related trends.
The selected countries were divided into three categories based on per capita income. Australia,
Japan, Korea and Singapore were placed in category (A) representing relatively high levels of
income; Malaysia, Thailand, Philippines, and Indonesia were placed in category (B) representing
middle levels of income; while China, India, Nepal, Pakistan, Sri Lanka, and Vietnam were
placed in category (C) representing relatively low levels of income. This classification was
incidentally consistent with the one proposed by the Asian Development Bank [Okita 1989]. It
also covered the variety of the countries in the Asia and Pacific region well, in terms of
geographic location, form of government and economic conditions. The presence of a particular
trend in the selected countries over all three categories provided the basis for the deduction that
the trend is pervasive in the region covered by the sample.
Although, above data format was partly necessitated by the quality of the data available, it
greatly facilitated making general inferences conceming the whole of the Asia and the Pacific
region. The individual differences between the data elements, in this case the country-specific
time series, allowed the data to be viewed as a sample representing the region it was drawn from.
The countries in the various categories of the sample were not viewed as special cases, but as
many outcomes (or multiple behavior modes) of the agricultural system of the region.
Time series plots for the various categories of countries were prepared for population, GDP and
GDP per capita to examine growth in the consumption base. The use of agricultural resources
was examined through a per capita food production index, fertilizer and pesticide application,
cultivable land and area under forests. The following observations were made with respect to the
growth of the consumption base and the condition of renewable agricultural resources.
Growth of the Consumption Base
Figures 4: a, b and c show population and GDP growth in the three categories of countries
selected for the analysis. Considerable population growth is shown over the three decades
covered by the data in all categories, although growth is much higher in the low-income
countries. GDP growth is the highest in the middle-income countries, while growth rates in the
high- and low- income countries are comparable. Consequently, as shown in Figures 5: a, b and
c, GDP per capita has grown at comparable rates in the high- and medium- income countries due
to moderate population growth in the former and high economic growth in the later. However,
high population growth rates and moderate economic growth have led to stagnation in GDP per
capita in the low-income countries.
According to the projections of UNCHS, shown in Figures 6: a, b and c, tremendous growth has
also occurred in urban populations across the board and the high growth rate is expected to
continue, although these rates are projected to taper off in the high-income countries. On the
other hand, rural populations have shown stagnating or declining trends in the higher income
countries and may be expected to decline further in the future. However, due to the overall
momentum of population growth, rural population has risen significantly in the medium- and
low- income countries, but is expected to taper off and begin to decline over the second decade
of the twenty-first century. As also shown in Figures 6: a, b and c, the total population is
expected to continue to rise in all countries well into the twenty-first century, although the rates
of projected population growth are negatively correlated with the levels of income - lower
income countries experiencing higher and continued rates of total population growth and
urbanization [UNHS 1987].
10
a) High Income Countries
POPULATION @pp
‘apes! MILLION ‘sscones: MILLION (U8 $)
= nuerRALA “= AUBTRALIA
fie ro00o00}| Ae wae
yooh] ORE fy tt HARA 2 BINGAPORE ‘Aus?
Fe IIHR AHI IIIA AE AOR mien ee
a0
[RST 10000
10] :
1000]
190
1000 190s 1908 TZ 1078 108019841080
TIME
b) Medium Income Countries
POPULATION GDP
59 MON ‘s00000. ‘MILLION (US $)
“= imooNEaia
Ae uacaran
Oe PHILiePINES
10000
c) Low Income Countries
POPULATION
MILLION
1000
Figure 4 Population and GDP growth
waco tee 9881072 wo7e 19801084
TIME
@DP
MILLION (US 8)
1000000;
= INDIA
oe WEPAL
41000005 | “K paxiscan
“+ BRILANKA
11
a) High Income Countries
GDP PER CAPITA
sa THOUSANDS UB §
= AUSTRALIA
127) 4 JAPAN
“he KOREA
107 | -o sincaPORE
woo 04) 988) 07278 = eB t884 1888
b) Medium Income Countries
GDP PER CAPITA
uss
10000;
= INDONESIA
MALAI
ie PHILIPPINES MALAYSIA
1000|,| “THAILAND
THAILAND
LIPPINES,
es Ura peeeeoonnnaastal
ONESIA
wooo 19841988072 876880841988
TIME
c) Low Income Countries
GDP PER CAPITA
Figure 5 GDP per capita growth
12
salqunoZ aurosuy MOT (9
saLnunog auoouy umipayy (q
UN P2010 “ROIEN JELGH,
9202 _O10F 002 066 OBL 0/81 OBEL O5BL
}85 VEWNH Lo 10084 1BqO!D EEL SHONN
a zounos:
salumog stwoouy ysiT (e
“eeeig AYIBIOANUN P1O1xO IQOIEN 182:00H
RI8G UEUNH UO 12008y EGO LEE SHONN
Nouwindod Tune
10M 901% “GOIEN TLGEH
9202 O10Z 0002 0681 OBE O61 OBL OBEL
{0S URLNH UO 1J000H 60015 2861 SHONN,
20a 'gouNoS
‘e903 4
“pasi weve,
°
2 f wodetg e
equews a ae o-—_e—»—5 86 88
etiean wien
don bad Oz -
sgueania
os er
: in
weUIe!A tiie
oa oo
2
TalEinea oot ae sara
ez
oe cos weep
ey
foe aeeop.1 ces ag
Tio
001 on zr
soon suomi 7)
Nouvingod Wune
JOAIUN BIO} *O "ROIBN IBUAEH
105 VBUNH VO LLOdey IECOID “L861 SHONN
siocetuis
eifteren be
or
pueleus
oo
sould
oe
oot
leisouopuy
jozt
jon
ost loz
‘evar ‘euOHIN corr
NowwIndod NvBEN
NowWindod Nvaun
NoUWINdod NYeUN
Projections for urban and rural population
Figure 6
13
In all cases, there is growth in the consumption base originating from two sources, growth in
population and expansion in economic activity. It remains to be seen how far the growth in the
consumption base can be sustained by the natural resource base and the environmental capacity.
A significant side effect of the expansion of economic base is the growth in urbanization, which
is necessitated by the technological choices requiring concentration of production activity to
achieve economies of scale. The concentration of human activity, however, may also create
concentrated doses of pollution which the environment is unable to assimilate. The results of
this are manifest in a lowering of the water table, pollution of groundwater aquifers and acid rain,
all of which adversely effect agriculture. Urbanization also often encroaches on prime
agricultural land, which reduces production potential, especially in Asia where only marginal
lands remain uncultivated.
Condition of Renewable Agricultural Resources
Renewable resources considered include agricultural land and forests, which have traditionally
met the food, fuel and timber needs of society. Figures 7: a, b and c show past trends in food
production per capita and agricultural land per capita in the countries of the three designated
categories of the sample.
The food production index is not comparable across selected countries due to differences in the
criteria used for calculating the base figures, but represents only an internal measure of the
changes in food availability in each country. Some autonomous jumps also appear in the data
since it has been constructed from many sources which, although mostly published by the UN,
contains some inconsistencies in the definitions used to represent the various categories of data.
For the purpose of constructing a reference mode, however, long-term patterns of trends rather
than numerical values of the time series are to be compared across the countries of the sample.
Hence, the above problems can be tolerated.
14
a) High Income Countries
INDEX OF PER CAPITA FOOD PRODUCTION AGRICULTURAL LAND PER CAPITA
(1979 -1981 = 100) (HA/PERSON)
“expos KOREA —* SINGAPORE _—* AUSTRALIA|
140) ge ees te
{ gar N, 10 .
120) “ vie A portato oeseoesres
100 er
a ere +h+-KoREA foe
Octet cat ttt tch ttt te eet
60 * ea JS - “non
é 2 + KOREA / CONE EEE SINGAPORE. pete ee ee ga,
* Simm | | someon Renee
feoo wes to68 tara ro 00 000 wns SOPPEO geo ee wet
TIME TIME
Source:
source:
FAO 1989: FAO Quartely Bulletin of Statistica 1989, Vol.2. New York. UN (1970-1988): statistical Yearbook for Asia & the Pacilic (1970-1088). NY.
UN (1970-1988): Statistical Yearbook for Asia & the Pacific (1970-1088). NY. UN 1979: Demographic Yearbook 1978. NY.
b) Medium Income Countries
INDEX OF PER CAPITA FOOD PRODUCTION AGRICULTURAL LAND PER CAPITA
(1979 - 1981 = 100) (HA/PERSON)
[ eimooNesia —!- MALAVaIA “F-PMILLIPINES THAILAND
140 mu
MALAYSIA eee = a
179 vices a THAILAND
10 <1 os See Or aera
PHILLTPINES” 5 MALAYSIA
20. 03 aan
7 | ek + PHILLIPINES
°° — INDONESIA oz! la PINES
40 + MALAYSIA 4
+ PHILUPINES a
a THAILAND
gl es ee Sit OG ee
eso tw84 060 —te7e ‘ore cv Seco” joes teea=~=SCwra~—sto7o~=~C«HDS:CS a
TIME: TIME
source: Source
FAO 1989: FAO Quartely Bulletin of Statistics 1989. VoL2. New York.
UN GB70-1080)- Stalslieal wurbook for Ante'& the Pecit (1070-1086), NY
UN (1970-1088): Statistical Yearbook for
Asia & the Pacific (1970-1988). NY
c) Low Income Countries
INDEX OF PER CAPITA FOOD PRODUCTION AGRICULTURAL LAND PER CAPITA
(1979 - 1981 = 100) (HA/PERSON)
| oma ws =n
120 7 NEPAL ©. PAKISTAN, = SRILANKA, + VIETNAM
Ate ot aa ue sa -
ol tumon Tea, ee
pacts pate 2g ee
fol or en ice
»
Bulletin of Statistics 1989, Volz. New York.
al Yearbook for Asia & the Pacific (1970-1988).NY
Figure 7 Food Production per capita and agricultural land per capita
15
It is observed that food production per capita exhibits a rising trend in all cases in spite of
considerable population growth, while agricultural land per capita shows a declining trend,
except in Australia, where it has been possible to maintain it at a steady level. This indicates that
increases in food production have been obtained largely through increasing the intensity of
cultivation and application of chemical fertilizers and pesticides. Indeed, as indicated in Figures
8: a, b, and c, fertilizer application has drastically increased in all countries of the sample over
the past three decades. The application of pesticides also seems to have increased in the countries
where data is available. The pesticides data, however, is inconsistent since in some cases it
refers only to DDT while in others it covers all pesticides.
Irrespective of the increases in yield, the absolute quantity of cultivable land has not increased
much in most of the countries of the sample, except in Australia, where it has been possible to
commission large tracts of unused land. This is shown in Figures 9: a, b, and c. It is observed
that, in general, where cultivable land did increase, it was at the cost of the forest area, which is
already very small in the countries with a stagnant level of land under agriculture. Some jumps
again appear in the plotted data, due to the changes in the definitions of the forest area and
agricultural land categories used.
Unfortunately, deforestation not only reduces valuable timber and fuel wood resources, it is also
known to cause soil erosion, water loss, flooding or drought, desertification and silting of
irrigation reservoirs, depending on the particular function of a forest in the complex organic
relationships existing in the ecological system [Bowonder 1986]. In spite of this knowledge,
about half of the area under forests in the developing countries was cleared between 1900 and
1965. At current rates of deforestation, the rest is likely to disappear in 50 years [UN-ESCAP
1986].
Excessive use of land resources has also been known to depreciate soil quality. Soil degradation
has occurred in the countries of the sample and elsewhere because of erosion, chemical
deterioration, loss of texture, water logging and salinity, all resulting from efforts to intensify
agricultural activity [Bowonder 1981]. Given the over-taxing of land resources, the per capita
food production index may be expected to decline in the future across the board. Declining
trends have already appeared in Nepal and Bangladesh, as shown in Figure 10.
16
a) High Income Countries
N FERTILIZER APPLICATION
(THOUSANDS TONS/YEAR)
< . = 100000
+ SINGAPORE -* AUSTR:
10000
1000
100
10
1 \ ee een in SINGAR ORE, / ;
woo wee e8sSwTZ=StO7GSC«NOHOC«éNGOASCSC#OO.
TIME
Source:
UN (1970-1988): Statistical Yearbook for
‘Asia & the Pacitic (1970-1988). NY.
iaia & the Pacific (1970-1908) NY
INSECTICIDE APPLICATION
(TONS/YEAR)
JAPAN oe
+ KOREA akean
© AUSTRALL
AUSTRALIA
Note:
1960-1071: ODT & Related Compound
1972 onward: Other Insecticide
b) Medium Income Countries
N FERTILIZER APPLICATION
(THOUSANDS TONS/YEAR)
10000 —— ——
20
1000 INDONESIA ~~
. 1
Hy
A Pagar or pe eBa 3
eeeee ‘
sof THAILAND —~_ WHDOWESIA |
wok + watavaia tg
+ paituieines
THAILAND
theo woe oa wre” wo7e toe toue 88
TIME
Source Source
UN'(1070-1008). Statistical Yearbook for
‘Asia & the Pacific (1970-1988). NY.
UN (1970-1988) St
‘Asia & the Paciti
INSECTICIDE APPLICATION
(TONS/YEAR)
toe |
Note:
1960-1071; DOT & Related Compound
1972 onward: Other Insecticioe
iatical Yearbook for
70-1988). NY.
c) Low Income Countries
N FERTILIZER APPLICATION
(THOUSANDS TONS /YEAR)
INDIA -
CHINA NEPAL 1 m
PAKISTINA = SRILANKA t= vteTWAM | ‘
100000 Sy rs
10000
oy
1000) BAKITAN
oe ost vetham #
1007 tt “SRILANKA
10 ‘
1
opt
1000 Wes t900 TRS w8O ee
TIME
source: Source:
UN (1970-1088): Statiatical Yearbook for
‘Asia & the Pacific (1970-1088). NY
Figure 8 Fertilizer and Pesticide application
17
UN (1070-1988) Stati
‘Asia & the Pacific (1970-1988). NY.
INSECTICIDE APPLICATION
(TONS/YEAR)
Pie |
seco 1972«W076 80
TIME
Note:
1960-1971: ODT & Related Compound
1972 onward: Other Insecticide
leat Yearbook for
a) High Income Countries
AGRICULTURAL LAND FOREST AREA
(THOUSANDS HECTARES) (THOUSANDS HECTARES)
Sa See See ee [ea honen = awanrone = Aden]
1000000 p--—
Pee AUBTAAMIA © ooo oe Bee AUSTRALIA |
00000) anenner ail
nee
1000)
i 1000
100
: 100
= SINGAPORE
10) fia
FORE eee ee 16
| NY foe SINGAPORE ee
Fy aresrareren eres iin PUR rierenrerirereresrer sre aware? qb a nc del tt
180 Wee ES WW72 7G EO twS4 1888 feco wes 900 1072~=«W7S~SCHOStwe tw
TIME, TIME
Source: Source:
UN (1070-1988): Statiatical Yearbook for UN (1970-1988): Statistical Yearbook for
‘Asia & the Pacitic (1970-
188). NY. Asia & the Pacific (1970-1088). NY.
b) Medium Income Countries
AGRICULTURAL LAND FOREST AREA
{THOUSANDS HECTARES) {THOUSANDS HECTARES)
2 a = — woonesa SS MALAYOIA
INDONESIA InDonEBIA PRILLIPINES THAILAND
20+) + MALAYSIA —s i " ee
x PRILLIPINES ee: a8 _ _ _
yg || & ramano a [ oo :
y ee eel INDONESIA
Hi ae PHILLIPINES 4. b 100}
a pe | sot
: - gupsqee be BO 2 60
6 eet MARAYSIA. 3°) tymuano MALAYSIA
20) pwiuiieines: £ +
{eco wes 1008 tor2—tw7e 1980 +—«teea 1980 feo wes
TIME
Source source:
UN (1970-1988): Statistical Yearbook for UN (1970-1988): Statistical Yearbook for
Asia & the Pacitic (1970-1088). NY. Asia & the Pacific (1970-1988). NY.
c) Low Income Countries
AGRICULTURAL LAND FOREST AREA
(THOUSANDS HECTARES) (THOUSANDS HECTARES)
1000000 — . r me sce a =
: —E | crana wot + NEPAL
F ) = enna + INDIA + NEPAL | no. panistaN ~ SRILANKA VIETNAM
[| pantsran > SrILANKA “VIETNAM | b a
“SSS eae 4 an — -
100000] * Te oo a: |
PAKISTAN | 10000) . cee
39-9 698 6 OE BT ‘
a UCL. Jowee) i pee METNM ee eee :
wetter $8886 NIEINAS
Ene errr reece or)
Fp | recs Nae pp ereeoecree rere rer vor 1000 5
1960 01964 1968 1972 1976 © 1980 1964 1988 Ww aca
TIME
Sourct Source:
UN (1970-1088). Statistical Yearbook for
‘Asia & the Pacitic (1970-1988). NY.
Figure 9 The competition between cultivable and forest land
18
“INDEX OF PER CAPITA FOOD PRODUCTION
% (1979 = 1981 = 100)
“160
140F-
120F
100F
Bangladesh .
80F
60Fr
40F-r
20r —e——- Nepal
—t+- Bangladesh
re) L 4. 1 { 1 i L + J. i A. + [on oe o + ia L 1 + L L 4 + 4 L 1
1960 1964 1968 1972 1976 1980 1984 1988
Year
SOURCE:
UN: Stat. Yb. for Asia & the Pacific. N.Y.. Various issues.
Figure 10 —_ Declining food per capita trends in Nepal and Bangladesh
19
Reference mode
The observed trends in the data taken from a geographically, economically and politically
diverse set of countries show that in all cases, increases in agricultural production - clearly a
private gain whether pursued by individuals or collectives - has been achieved in the first
instance by making an intensive use of the land resources viewed as capital inputs rather than as
an environmental system. It is also quite evident that expansion in agricultural land has been
achieved by consuming forests - another environmental system which is important to the
maintenance of agricultural land as a sustainable resource, but which is viewed by individuals
and collectives involved with agriculture as an unused endowment. A variable implicit in the
above description is soil fertility, which is partly dependent on the volume of standing forests
and partly on the intensity of cultivation. Soil fertility can be propped up by fertilizer application,
but would eventually decay if the pressure on soil and deforestation continues.
The projections obtained from digesting the above information indicate a tragedy of the
commons in making, which is analogous to a reference mode that might be constructed from the
assimilating above information in a learning framework described earlier. In this reference mode,
inferred future of food production per capita will show an overshoot and decline behavior which
would be followed by a similar trend in population. Land under forests and soil fertility would
decline to a low stagnant level and land under cultivation would rise to a high stagnant level.
Figure 11 illustrates the pattern arrived at.
It should be noted that the reference mode constructed in Figure 11 is a pattern encompassing the
past and future, rather than merely being a historical record and is based on both past history and
inferred future. It contains both concrete variables appearing in the historical record and abstract
concepts like soil fertility that sum up several pieces of quantitative and qualitative information.
The time horizon of a reference mode may depend on the purpose for which a model is
constructed, but it will invariably be longer than the historical record. In this case, it must be
much longer than the historical record if the purpose is to search for a sustainable future as the
time constants of the processes being considered are long.
20
histrorical trends history and inferred future
oorts
° x.
x
‘
\
3 \
gf
af
of
ae
NN
N
Le .
ae ae N
aoe —
----- consumption base — — —soil fertility
food production index — land under agriculture
—— -— land under forests
Figure 11 Reference mode describing food production patterns constructed from
numerical and qualitative historical data
A reference mode is best visualized as a fabric subsuming the trends drawn on paper and would
often fall into recognizable categories of patterns, in this case - the overshoot and decline pattem
describing a tragedy of the commons. This raises an important issue: why have we not attempted
to recognize widely occuring patterns of behavior as reference mode archetypes? A related
problem is that in a messy world, problems may not present themselves as neat archetypes, but as
complex time histories. How should such complex time histories be partitioned to arrive at
problem archetypes? Let me address the second question first.
21
Dealing with complex time histories
I have discussed the process of partitioning a complex problem in Saeed (1992) and Richardson
(1997). Since models cannot be made overly complex if they are to remain understandable,
complex problems must be sliced into smaller parts so that the parts meet the requirements of the
intended policy design. This calls for separating the multiple modes contained in a complex
historical pattern in a rather special way.
The term multiple modes is not new to system dynamics, although it is used a bit loosely. Not all
classes of behavior implied by multiple modes may be relevant to creating a model for an
effective policy design. In fact, many intuitively sensible schemes of partitioning a system may
create models that do not incorporate policy space for investigating the possibilities of change.
The multiple modes relevant to a problem may refer to the simultaneously existing components
of a complex pattern of behavior that is exhibited by a system over a given period; they may
represent patterns experienced over different periods of time in a system of relationships; or even
patterns experienced in similar organizations that are separated by geographic space. The
conceptual space in which multiple modes can be found is, therefore, three dimensional as shown
in Figure 12.
When multiple modes contained in a complex historical series are the focus of a modeling effort,
the complex modal space will be sliced as shown in Figure 13. The simultaneous modes
constituting the complex historical pattern will be subsumed in a selected partition while the
variety of pattems in the temporal and geographic dimensions are ignored. Such a problem
slicing process will create situational theories and forecasting models that may explain a unique
and complex pattern, and also extrapolate it into the future, but that do not shed any light on the
possibility to change it.
22
Time Separated Modes
Mode Space in
a Complex System
Simultaneous Modes
Figure 12 Multiple mode space
Source: Saeed (1996)
Time Separated Modes
situation specific
problem slice
Simultaneous Modes
Figure 13 _— Problem slices for developing forecasting models and situational theories
Source: Saeed (1996)
23
On the other hand, when a model is intended for exploring policy options for system change, the
complex modal space must be sliced as shown in Figure 14. The partition selected for modeling
will subsume multiple modes that are separated by time and geography since only then its
underlying structure would contain the mechanisms of modal change. It may not necessarily
incorporate multiple modes that exist simultaneously in system behavior since interaction
between the mechanisms creating these modes may not provide any additional policy space,
although this may enhance a model's ability to track history accurately. When policy exploration
rather than tracking history is the primary purpose of a modeling effort, simultaneously existing
multiple modes and their underlying structure can be separated and addressed in different models
for limiting complexity contained in a single model.
Time Separated Modes
situation independent
problem slice
Geography Separated Modes
Simultaneous Modes
Figure 14 —_ Problem slices for exploring policy design
Source: Saeed (1996)
24
Representing a complex system as a number of submodels that produce behavior different from
what appears in the historical data will require defining thereference mode differently from
historical behavior. For example, each of the two complex time histories shown in Figure 15
contains a trend simultaneously existing with a cyclical tendency. To be able to address the two
issues conceming the cycles and the trends, this problem may be represented by two models:
One subsuming the multiple modes existing in the two trends, the other subsuming the cyclical
mode existing in both of them. The two models so created will keep together the symbiotic
processes underlying the potential multiple pattems thus providing the policy space to attempt a
design for change. Also, the two components of the design so created can be pursued quite
independently.
simultaneous modes
>
composite decomposed
yD |IAn
~~ | LT bh
sepout peyertedes Ayderioeb pure aun
Figure 15 Decomposing multiple modes for slicing a complex problem
Source: Saeed (1996)
25
Reference mode archetypes - a task ahead
The concept of archetypes representing generic systems pervasively found in experience is quite
old in system dynamics, but somehow, the starting point for the archetype seems to have been
the system structure not a problem pattern it might represent. Even the nomenclature used to
define system archetypes is inconsistent. Archetypes found in the literature have been named
both with respect to structure and behavior and multiple archetypes may display the same
behavior. These anomalies arise apparently from not using the reference mode as a starting point
for defining archetypes.
Jay Forrester has often pointed out that a small number of systems can represent about 90% of
the problems encountered in experience. It is only appropriate that we attempt to define these
problems in their generic form, which has been the motive behind defining archetypes. Table 2
shows ten of the most commonly stated archetypes implicitly referring to a reference mode in
system dynamics. Unfortunately, many of them represent truncated patterns not useful for
modeling policy issues as they do not subsume inferred futures.
Table 2 commonly stated reference modes
Sluggish adjustment
Exponential growth
S-shaped growth
Overshoot and decline
Oscillation
Escalation
Erosion
Resource misallocation
Policy resilience
Multiple modes
26
Sluggish adjustment and inability to arrive at a designated goal is a pervasive problem both in
physical and social systems with inadequate feedback tracking discrepancy, although this
phenomenon occurs together with other patterns. Y et it represents an important basic pattem
subsuming a widely occurring reference mode.
The exponential growth pattern is experienced widely in systems fueling their own reproduction,
although exponential growth would almost never go on for ever. Thus, exponential growth is
only a relatively short-term phenomenon, which invokes concems about overshoot beyond
sustainable levels followed by an inevitable decline. It might be uninteresting to investigate
exponential growth per se from a policy perspective. Only when it is coupled with its inferred
future, a possible smooth transition to equilibrium in the form of an S-shaped growth, ora
dysfunctional overshoot and decline, does there appear to be an interesting reference mode.
Oscillation is another widely experienced pattem, which might appear together with other
patterns, but sliced from the rest would have a unique policy implication of creating a damping
process. Hence, oscillation cannot be divorced from the extent of damping present in it.
Escalation implies increasing commitment to a failing cause, both in the physical and
metaphorical sense. Escalation, like exponential growth may not represent a complete reference
mode category, since any escalation trend must end into a catastrophe.
Erosion is another inadequately defined pattem as an eroding trend must culminate with the
tuming around or demise of an organization. Resource misallocation on the other hand represents
a tendency towards a misallocation goal, and even though it may appear to be a static pattern, the
past tendency precipitating it has significant policy implications. Policy resilience and multiple
modes can manifest in a variety of pattems and hence cannot be views as specific patterns.
At the outset, there seem to be three broad categories of problem behavior modes possible:
1) tendency towards a single and unique equilibrium
2) tendency towards multiple equilibria
3) patterns of growth and instability.
27
Within the first category, the system would tend to come to an equilibrium characterized by
problems like low productivity, poor efficiency, unequal income distribution, low market share,
etc. Such modes might give the appearance of a static rather than a dynamic problem, however,
the points of interest in them are the end conditions per se but how they are reached, irrespective
of the initial conditions. The second category of problems concern processes that may not lead to
a single and unique equilibrium, but might display multiple equilibria depending on the inputs
and the environmental conditions. Activating growth mechanisms in such systems might also
transform a functional equilibrium into a dysfunctional one. The last category of problems
concem growth, overshoot and instability. Often a smooth transition to a sustainable state is
desirable, overshoot and instability are to be avoided.
Further work is needed to identify specific patterns within each category that should fit pervasive
problem patterns. Once the reference mode archetypes are evident, generic systems underlying
them, both in terms of feedback loops and simple stock and flow structures should be delineated.
Conclusion
I have attempted in this paper to define the characteristics of a reference mode and how it is
distinguished from historical data, both qualitative and quantitative. A reference mode is an
abstract concept subsuming past as well as inferred future behavior. It can best be visualized as a
fabric collecting several patterns as well as the phase relationships existing between them. It may
contain concrete as well as abstract variables that are different from the data it is based on. It
may also represent only a slice of the complex time history it emulates and may thus look very
different from the history itself. A reference mode is an end product of a learning process that is
similar to the process involved with building a model and analyzing it. There seems promise for
constructing generic forms of problem patterns which might fit a large cross-section of the
dynamic problems encountered in the world. Constructing these generic forms is not attempted
in this paper but is seen as an important part of the methodological progression expected.
28
References
Abbott, E. A. 1987. Flatland London: Penguin
Bowonder, B. 1981. The Myth and Reality of High Yield Varieties in Indian Agriculture.
Development and Change. 12(2).
Bowonder, B. 1986. Deforestation in Developing Countries. Journal of Environmental Systems.
15.
Hunsacker, P. L., and Alessandra, A. J. 1980. Learning How to Learn. In The Art of Managing
People. Englewood Cliffs, NJ: Prentice Hall. pp. 19-49
Kolb, D. A. 1974. On Management and Learning Process. In Organizational Psychology: A book
of Readings, 2nd Ed. Englewood Cliffs, NJ: Prentice Hall. pp. 27-42.
Kolb, D. A. 1984. Experiential Learning. Englewood Cliffs, NJ: Prentice Hall
Kolb, D. A., Rubin, I. M., and McIntyre, J. M. 1979. Learning Problem Solving. In Organization
Psychology: An Experiential Approach, 3rd ed. Englewood Cliffs, NJ: Prentice Hall. pp.
27-54.
Okita, S., et. al. 1989. The Asian Development Bank in the 1990s: Report of a Panel. Manila:
Asian Development Bank
Richardson, G. P. 1997. Modelling for Management Vol. II. Brookfield, Vt: Dartmouth
Publishing Co. Ch. 15
Saeed, K. Sustainable Development, Old Conundrums, New Discords. Jay Wright Forrester
Award Lecture. System Dynamics Review. 12(1). 1996.
Saeed, K. 1992. Slicing a Complex Problem for System Dynamics Modeling. System Dynamics
Review. 8(3).
Saeed, K. 1994. Industry and Environment in Asia and the Pacific Region UN-ESCAP,
Bangkok
Saeed, K. 1997. System Dynamics as a Technology of Learning for New Liberal Education. FIE
conference proceedings. Pittsberg, PA
Saeed, K. 1998. Towards Sustainable Development, 2nd Edition: Essays on System Analysis of
National Policy. Aldershot, England: Ashgate Publishing Company.
UN/ESCAP. 1986. Environmental and Socio-Economic Aspects of deforestation in Asia and
Pacific. Proceedings of the Expert Group Meeting. Bangkok: United Nations
UNCHS. 1987. Global Report on Human Settlements 1986. Habitat Nairobi: Oxford University
Press
29
CC BY-NC-SA 4.0