Financial analysis

Entrepreneurship as a science of the artificial

Saras D. Sarasvathy *

University of Maryland and R.H. Smith School of Business, 3322 Van Munching Hall,

College Park, MD 20742, USA

Abstract

This essay connects four key ideas from Herbert Simon�s ‘‘Sciences of the Artificial’’ to re- cent research on entrepreneurial expertise: (1) natural laws constrain but do not dictate our

designs; (2) we should seize every opportunity to avoid the use of prediction in design; (3) lo-

cality and contingency govern the sciences of the artificial; and, (4) near-decomposability is an

essential feature of enduring designs. The essay is based on a series of conversations and emails

with Simon about the empirical findings of my doctoral dissertation that involved a protocol

analysis study of expert founder-entrepreneurs.

� 2003 Elsevier Science B.V. All rights reserved.

PsycINFO classification: 2340; 3940

JEL classification: D21; D81; L20

Keywords: Entrepreneurship; Rational choice; Effectuation

1. Introduction

Most of us enter the world of research with perceptions of the scholarly life as a

quest for the holy grail. The grand myth of Parsival inspires us even when we might

be less than confident about our own potential to equal his achievement. The majesty

of being a part of the quest itself inspires us. . . until sooner or later, the terrible fate of Sisyphus, waking every dawn to push the boulder relentlessly up the mountain,

only to have it drop back at dusk and start all over again in the morning, begins

to loom as possible reality in our research ‘‘careers’’ – and it becomes more and more

difficult to seek meaning and fulfillment in our ‘‘quest.’’

* Tel.: +1-301-405-9763; fax: +1-301-314-9414.

E-mail address: [email protected] (S.D. Sarasvathy).

0167-4870/03/$ - see front matter � 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0167-4870(02)00203-9

Journal of Economic Psychology 24 (2003) 203–220

www.elsevier.com/locate/joep

I had the good fortune to work with Herb Simon, a Parsival if ever there was one

in the social sciences, from whom I learned that the key was not to strive to become

Parsival, but to ‘‘imagine Sisyphus happy,’’ as Camus compels us in his profound

essay. Many of my meetings with Simon, especially in the early stages of my relation-

ship with him, began by his asking, ‘‘So what do we know today that we did not know the last time we met?’’ That was not an easy question for a doctoral student

to answer every other week! But I took it as a challenge and began using the opener

to provoke discussions into a wide variety of topics such as: how entrepreneurs

think; the importance of ‘‘naming’’ things in how human beings learn; how embryo-

nic cells know how to become particular types of tissues such as lung tissue or bone

marrow; Borges� mazes and Lem�s constructors; and even deeply personal issues such as what love and death and money meant to him (and to me). Several strands of this

wonderfully messy conversation, however, were converging on the notion of the ‘‘ar- tificial’’ – the idea of ‘‘design’’ as opposed to ‘‘choice.’’ We had just finished a con-

ference paper based on this particular convergence (Sarasvathy & Simon, 2000), and

were in the middle of revising it for a journal, when the curtain came down on his last

act. 1

This essay is spun from the threads of that conversation. It revolves primarily

around four key themes that Simon outlined in ‘‘Sciences of the Artificial’’ and illus-

trates how they connect together in the world in one particular domain: Entrepre-

neurship. In the following pages, I attempt to tell the story of how we explored the connections as well as try to establish a thesis about the connections.

The story begins with my empirical investigations of entrepreneurial expertise for

my doctoral dissertation. Based on the findings of this investigation, I developed the

theory of effectuation (Sarasvathy, 2001a) and located its antecedents in Knight�s formulation of true uncertainty (Knight, 1921), Weick�s conceptualization of enact- ment (Weick, 1979), and March�s technology of foolishness (March, 1982). Only when Simon and I were invited to contribute a paper to a conference in 2000 and

he suggested that there might be a connection between my formulation of effectua- tion and his theory of near-decomposability that I began to see the various threads in

the Sciences of the Artificial that directly tied together with effectuation (Simon,

1996).

Simon�s logic in seeking a connection between the two theories was, as most of his seminal ideas have always been, very simple: ‘‘As near-decomposability is an astonish-

ingly ubiquitous principle in the architecture of rapidly evolving complex systems, and

effectuation appears to be a preferred decision model with entrepreneurs who have cre-

ated high growth firms, we should be able to link near-decomposability to the processes

these entrepreneurs use to create and grow enduring firms – whether in an experimental

situation or in the real world.’’

Delving into both theories with this new insight led me to realize that the connec-

tion lay in the roles that locality and contingency played in each. Locality here refers

1 When I asked him about death the day after his 80th birthday, he said he could accept that life, like a

play, would have a last act and then the curtains would come down.

204 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

to the fact that cognitive limitations on our rationality allow us to build artifacts that

achieve only local optima at best; yet, our artifacts can endure over time by learning

to adapt to contingencies and sometimes even exploit those contingencies for their

own survival and prosperity. But I am getting ahead of the story here – to tell it

properly, I will first briefly explain (1) effectuation and (2) near-decomposability and then explore the roles that locality and contingency play in each. If I do my

job right, the story will have the happy ending that effectuation, together with Si-

mon�s work on the artificial can explain the creation of high growth firms; and also, that several interesting sequels can be developed by envisioning entrepreneurship as

a science of the artificial.

2. Effectuation: A theory of entrepreneurial expertise

For my doctoral dissertation, I used the protocol analysis method that Simon and

his colleagues had used to study experts in a variety of fields (Ericsson & Simon,

1993). My subjects consisted of expert entrepreneurs, founders of firms ranging in

size between $200 million and $6.5 billion and including a variety of entrepreneurial

experiences in a wide variety of industries. Each subject had to solve a 17-page prob-

lem set consisting of 10 typical decision problems that occur in transforming an idea

into a successful firm. The logic underpinning the study was: Given the fact that the subjects are expert entrepreneurs, and have nothing else in common, is there anything

common in the problem solving processes they use? If so, I could then extract those

commonalties and create a base-line model of entrepreneurial expertise.

The intuition based on extant literature suggested there was no such thing. Several

studies trying to tie psychological traits of entrepreneurs with firm success did not

seem to lead anywhere (see Gartner, 1988, for a good review); entrepreneurs were

also found to range all over the risk-preference spectrum (Brockhaus, 1980; Palich

& Bagby, 1995); and some economists even theorized that entrepreneurial expertise was nothing but a statistical artifact (see Arrow�s comments in Sarasvathy, 2000).

But intuition based on extant scholarship is not the only type of intuition avail-

able to us. Another type of experiential intuition has kept entrepreneurship scholars

steadfast in their pursuit of the mythical beast ‘‘entrepreneur.’’ Meeting and talking

with entrepreneurs in person, and interacting with them on a daily basis suggests

there is indeed something that ties them together as a species – something in the lan-

guage they use, the stories they tell, and the way they approach and handle problems

and people. Of course, that could merely be a retrospective retelling of an essentially random set of experiences. Hence, the protocol analysis to delve into the black boxes

of their cognition.

By the time I got to the 20th entrepreneur in the analysis of the very first problem

(identifying the market for a new product), the coders began to agree that a clear

pattern had come to light about how entrepreneurs create markets and firms. The

key characteristic of this pattern was that it inverted the principles and processes that

we teach students in MBA programs on how they should go about identifying the

market for their ideas. Since these principles and processes are usually based on a

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 205

causal or predictive approach to reasoning, I termed the entrepreneurial approach

‘‘effectuation,’’ to emphasize its aspects of being an ‘‘inverse of causation.’’ To sum-

marize the results quantitatively, 74% of the participants in the study behaved in ac-

cordance with the effectuation model at least 63% of the time, and 44% of them, at

least 85% of the time (Sarasvathy, 2001b).

2.1. Brief outline of effectuation

The model of entrepreneurial expertise extracted from the protocols is complete in

the sense that it identifies a particular problem space, a solution process, a set of

principles and a unifying logic that ties it all together into a coherent whole. Here

I will briefly outline the theory. See Sarasvathy (2001a) for a detailed exposition.

Traditional models based on causal rationality operate in a small, comfortable clearing in the woods characterized by: (a) given, well-specified goals; (b) well-under-

stood causes and past histories that enable reasonably reliable predictions about the

future; and, (c) an independent environment (such as a ‘‘market’’) that serves to sep-

arate the wheat from the chaff of decisions made by individuals and firms. But all

around this cozy clearing stretches the vast, relatively unexplored jungle where goal

ambiguity, Knightian uncertainty, and endogenous markets dominate the landscape.

This is the problem space for effectuation and is best described by Simon in Sarasv-

athy and Simon (2000): ‘‘Where do we find rationality when the environment does not independently influence outcomes or even rules of the game (Weick, 1979), the future is

truly unpredictable (Knight, 1921), and the decision maker is unsure of his/her own

preferences (March, 1982)?’’

An empirical example of this Weickian–Marchian–Knightian problem space is the

‘‘suicide quadrant’’ in Fig. 1. Both expert marketers and experienced venture capital-

ists routinely avoid this space that involves introducing a new product in a new mar-

ket. Yet, experienced entrepreneurs know that this is the space within which great

Existing Market

New Market

Existing Product

New product

Suicide Quadrant

Fig. 1. Example of problem space for effectuation.

206 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

companies such as Edison�s General Electric, Apple Computers and Medtronics often emerge. The problems in this space, unlike problems involving causal rationality, do

not begin with clearly specified goals.

Some examples should clarify the distinction between the two types of processes.

Imagine the manufacture of a product. In the case of causal or decisional rationality, the blueprints of the product are provided in advance, together with its costs, and

estimates of market demand; the manufacturer needs simply to procure the raw ma-

terials and process and assemble them according to the predetermined plan. In the

case of effectuation, the manufacturer has a general idea that might lead to a product

that could be marketed profitably. Gillette was looking for something that customers

would have to purchase repeatedly (McKibben, 1998). While he was shaving one

morning, it occurred to him that a non-permanent razor might fit his specification.

He then had to develop a cheap, effective removable-blade razor, generate plans for creating an adequate initial market, search for sources of funds to get started, and so

on, always modifying his plans as he gained new knowledge from his initial experi-

ments and efforts. This example involves both causal and effectual approaches at dif-

ferent stages of firm creation.

But for the purposes of clear theoretical exposition, a simple, but highly dichoto-

mous example might serve to anchor the arguments better. Imagine the contrast be-

tween a chef to whom a specific menu is presented, and who only has to list the

needed ingredients, shop for them, and then cook the meal (causal decision); and a chef who happens to find some ingredients in his cupboards, and some utensils

in the kitchen, from which he imagines and produces a delicious meal (effectual de-

sign). In the one case, the givens are assembled, in the other case, they are con-

structed in a constrained environment through imaginative agents in an ebullient

pursuit of interesting (and hopefully valuable) possibilities.

For a more complex example from entrepreneurship, we can contrast the actual

history of an internet company such as RealNetworks (leader in the real time audio

and video streaming industry on the Web) with how we teach entrepreneurship stu- dents to develop a business plan. If a student came up with the idea of starting an

interactive cable TV channel with progressive content (which was what Rob Glaser,

the founder of RealNetworks originally set out to build in 1994), we would advise

them to proceed as follows: carry out market research to estimate size, growth rate

etc. of key target segments; come up with financial forecasts; write a business plan;

raise funds needed; test market the product and then implement market strategies to

capture as large a market share of the target markets as possible. And the student

would most probably never come upon the idea of giving voice to the ‘‘mute’’ web. In contrast to this, a close examination of the actual history of Progressive Net-

works (later renamed RealNetworks) tells a fascinating story of effectuation with

contingent twists and unpredictable turns, relatively unplanned plunges into proxi-

mate markets, and a relentless endeavor to shape and control the standards in an

emerging industry. See Sarasvathy and Kotha (2001) for a detailed analysis.

Effectuation is not a process of choosing among given alternatives, but of gener-

ating the alternatives themselves, and simultaneously discovering and assessing desir-

able and undesirable qualities of several possible ends. In this sense, effectual

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 207

processes involve design (including the design of alternative goals), not just choice.

Entrepreneurship, involving effectuation, has proved an elusive target for economic

theories, mainly because those theories have, with rare exceptions, been limited to

choices among given alternatives, applying pre-specified criteria, to achieve predeter-

mined goals.

2.2. Means available for effectuation and the solution process

Entrepreneurs begin with three categories of what I have called ‘‘means.’’ They

know who they are, what they know and whom they know – their own traits, tastes

and abilities, the knowledge corridors they are in, and the social networks they are a

part of 2 . Starting with these means, the effectuator asks herself, ‘‘Given who I am,

what I know, and whom I know, what can I do? What types of effects can I create?’’ Contrast this with causal reasoning that focuses on questions such as, ‘‘Given the

particular goal I want to achieve, what ought I to do? Which particular path should

I take?’’ Causal reasoning tends to begin with a universe of all possible alternatives

and seeks to narrow the set of choices to the best, the fastest, the most economical,

the most efficient etc. Effectual processes seek to expand the choice set from a narrow

sliver of highly localized possibilities to increasingly complex and enduring opportu-

nities fabricated in a contingent fashion over time. One important example of this

process, that of entrepreneurial marketing, is represented in Fig. 2. Causal models of marketing prescribe that the entrepreneur begin with a market

defined as the universe of all possible customers; then divide this universe into rele-

vant segments based on rigorous market research; choose a target segment after an-

alyzing predicted returns and risks for each segment; and finally design marketing

strategies to capture the target market. The effectual model suggests the entrepreneur

should find a customer or a partner searching very locally, just someone from within

their personal social network or through garbage can processes; then generalize the

initial customer or partner into a segment; add segments over time in a contingent fashion; and eventually define the market for their product/firm.

Causal models are based on a predictive logic: To the extent we can predict the fu-

ture, we can control it. Being able to predict the size, growth rate and potential trends

of target segments, for example, allows the entrepreneurial firm to secure its own

financial future.

2.3. The logic of effectuation

Effectuation suggests a rather different logic for the choice process: To the extent

we can control the future, we do not need to predict it. How does one control an un-

predictable future? The answer to this seemingly paradoxical question lies in the re-

2 At the level of the firm, the corresponding means are its physical resources, human resources, and

organizational resources, a la the resource-based theory of the firm (Barney, 1991). At the level of the

economy, these means become demographics, technological capabilities, and socio-political institutions

(such as property rights).

208 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

alization that a large part of the future actually is a product of human decision mak- ing. By bringing on board key stakeholders who can ‘‘deliver’’ the future, the entre-

preneur need not waste time, resources, or effort on prediction. Of course, such a

view may express hopes rather than realities, and many entrepreneurs in the real

world do fail. This fact does not negate the hypothesis that they are often more con-

cerned with molding, or even creating, the part of the world with which they are con-

cerned than with predicting it and reacting to the prediction.

Fig. 2. Effectual market creation contrasted with causal marketing.

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 209

This particular logic of control is embodied in three principles that together form

the core of effectual reasoning:

1. Affordable loss rather than expected returns: Causal models focus on maximizing

the potential returns for a decision by selecting optimal strategies. Effectuation pre-determines how much loss is affordable and focuses on experimenting with

as many strategies as possible with the given limited means. It prefers options that

create more options in the future over those that maximize returns in the present.

The extreme case of this is the zero resources to market principle. This principle

destroys uncertainty by pre-digesting the down side.

2. Partners rather than competitive analyses: Causation models such as the Porter

model in strategy, emphasize detailed competitive analyses. Effectuation empha-

sizes partnerships through pre-commitments from stakeholders as a way to reduce and/or eliminate uncertainty and erect entry barriers. Pre-commitments from key

stakeholders make uncertainty irrelevant by ‘‘delivering’’ a future that looks very

similar to what was contracted for.

3. Leveraging contingencies rather than avoiding them: When pre-existing knowledge

such as expertise in a particular new technology forms the source of competitive

advantage, causation models might be preferable. Effectuation, however, would

be better at leveraging contingencies that arise unexpectedly over time.

This principle makes uncertainty a friend and an asset, eliminating the need to

overcome it.

Effectuation has at least one major implication for the success/failure of entrepre-

neurial firms. While firms created through effectual processes may not reduce the

probability of failure, they do reduce the costs of failure. They allow failures to hap-

pen earlier and at lower levels of investment, while keeping open the upside option of

making larger investments should early successes begin to cumulate.

That is because the logic of control overcomes the problems of prediction by keeping investments to the utmost minimum, continually negotiating with key stake-

holders, and learning to use contingencies to create new ends or adapt better to

achieve old ones. The idea of using a logic other than that of prediction is extremely

important for the creation not only of firms, but of any enduring human artifact. As

Simon puts it in his book, Sciences of the Artificial (1996, p. 147): ‘‘Since the conse-

quences of design lie in the future, it would seem that forecasting is an unavoidable

part of every design process. If that is true, it is cause for pessimism about design, for

the record in forecasting even such ‘‘simple’’ variables as population is dismal. If there is any way to design without forecasts, we should seize on it.’’

3. Entrepreneurship as a science of the artificial

Sciences of the Artificial is one of the most exciting pieces Simon has ever pub-

lished. In an oeuvre of over a thousand publications, that is saying a lot. But it is

also, in my considered opinion, one of the most irritating. It bursts at its seams with

210 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

brilliant ideas and mouth-watering possibilities for scholarship and pedagogy, but

does not develop many of these into something readers can sink their teeth into, es-

pecially in the domains of management and economics. One is left with a sense of the

enormity of work to be done, but not quite sure where to begin. So when we were

invited to write a paper for a technology entrepreneurship conference in 2000, and Simon suggested in a quietly provocative voice that we should try and link effectu-

ation with the notion of near-decomposability that he had outlined in the book, I

was rather skeptical. But rolling up my sleeves and digging into the fertile loam of

ideas that is Sciences of the Artificial made me realize how close my ideas generated

through my empirical work were to his ideas culled from a lifetime of trying to un-

derstand human artifacts.

3.1. Near-decomposability 3

Near-decomposability is a pervasive feature of the architecture of the complex

systems that we find in the world, both inorganic and organic, ranging from elemen-

tary particles to social systems (Simon, 1969). A complex system is nearly decompos-

able if it is comprised of a number of interconnected subsystems in such a way that

elements within any particular subsystem interact much more vigorously and rapidly

with each other than do elements belonging to different subsystems. There may be a

whole hierarchy of systems, subsystems, sub-subsystems, etc., where this same prop- erty holds between any two levels. In such systems, (1) the short-term (high-fre-

quency) behavior of each subsystem is approximately independent of the other

subsystems at its level, and (2) in the long run, the (low-frequency) behavior of a sub-

system depends on that of the other components only in an (approximately) aggre-

gate way.

We should note at this point that near-decomposability is not the same as com-

plete decomposability. The key to understanding near-decomposability is that in this

architecture, what constitutes a good design for a component is nearly independent of the designs of other components. With correlated components, good design in-

volves a space of kn possibilities (n being the number of components and k being the number of subassemblies). With complete decomposability, it involves search

in the space of n � k possibilities, and with near-decomposability, it involves approxi- mation in the space of n � k possibilities. The human body is a good example of a nearly, but not completely, decomposable system. On the modularity continuum,

nearly decomposable systems are stably balanced between synergistic specificity

and separability (Schilling, 2000). The structural property of near-decomposability has two implications for the evo-

lution of complex systems:

First: If we begin with a set of simple elements that are capable sometimes of form-

ing stable combinations, and if the combined systems thus formed are themselves

3 In this section, I borrow chunks of my conference paper with Simon (Sarasvathy & Simon, 2000) and

so at least parts of it embody his voice as well as mine.

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 211

capable of combining into still larger systems, then the complex systems we will ob-

serve after this process has proceeded for some time will almost all be nearly decom-

posable systems. The universe as we observe it today provides ample evidence for this

claim. The gradual evolution of the elements from primeval fundamental particles

and hydrogen, then the evolution of successively complex molecules and living organ- isms – has observably produced nearly decomposable systems with clearly defined

particle, nuclear, and atomic levels, and a whole sequence of molecular levels above

the atomic. Moreover, it has been shown that the time available for evolution of living

organisms on earth is sufficient to produce organisms of the complexity that is actu-

ally observed (say, bacterial complexity) if the organisms and their subsystems are

nearly decomposable, but not if they must be completed by an uninterrupted se-

quence of unions of elementary components (Simon, 1996, p. 189).

And second: If we begin with a population of systems of comparable complexity, some of them nearly decomposable and some not, but all having similar frequencies

of mutation, the nearly decomposable systems will increase their fitness through evo-

lutionary processes much faster than the remaining systems, and will soon come to

dominate the entire population. Notice that the claim is not that more complex sys-

tems will evolve more rapidly than less complex systems but that, at any level of

complexity, nearly decomposable systems will evolve much faster than systems of

comparable complexity that are not nearly decomposable.

The connection between near-decomposability and rapid evolution is simple and direct. In nearly decomposable systems, each component can evolve toward greater

fitness with little dependence upon the changes taking place in the details of other

components. Simple mathematics and recent simulations by Marengo, Frenken,

and Valente (1999) have shown that, if and only if these conditions hold, natural se-

lection can take advantage of the random alterations of components with little con-

cern for countervailing cross effects between them. Such a system is like a defective

safe that clicks whenever one of its dials is set correctly, independently of where the

other dials are currently set. In other words, in nearly decomposable systems failures may be contained as local

events, without disastrous system-wide consequences. Thus nearly decomposable

systems survive not because they make fewer mistakes, but because they can control

the damage locally. Yet, the system as a whole can cumulate the benefits of learning

over time precisely because it is not completely decomposable. It is the tension be-

tween the interdependence of parts and their approximate independence that gives

an evolutionary advantage to nearly decomposable systems. In this particular char-

acteristic, they echo the implications for success and failure suggested by the theory of effectuation.

3.2. Locality and contingency in near-decomposability and effectuation

In Sciences of the Artificial, Simon shows that the phenomena that constitute the

artificial are imbued with and driven by locality and contingency, both in structure

and movement.

212 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

In designing artifacts, human beings are confined within rather narrow local limits

in terms of space, time and knowledge – primarily because of the bounds on our cog-

nitive capacities and the natural limits on our internal information processing sys-

tem:

• First, we can attend only to a limited number of things at a time. • Second, our planning horizons tend to be short run rather than long run. • Third, the stock of knowledge at any given point in time exists dispersed across

individual experts and specialized knowledge corridors that are not always eas-

ily accessible to all decision makers.

Ergo, most artifacts for most purposes are only locally adaptive. In other words,

they survive well only within particular domains and short-run periods during which

the knowledge stock remains relatively unchanged. As Simon observes, ‘‘The world

of economic affairs is replete with local maxima. It is quite easy to devise systems in which each subsystem is optimally adapted to the other subsystems around it, but in

which the equilibrium is only local, and quite inferior to distant equilibria that can-

not be reached by the up-hill climb of evolution.’’ (1996, p. 47).

Enduring artifacts, however, have to incorporate a way to deal with changes in

local environments over time, whether these are changes in technologies or prefer-

ences or other contingencies that reshape the environment. That is why artifacts that

incorporate the property of near-decomposability in their structure endure better

since they allow parts of the structure to be modified, or even destroyed and rebuilt, while retaining the rest of the entity relatively unchanged. Nearly decomposable sys-

tems are very good at exploiting both locality (necessitated by the limitations of the

inner environment) and contingency (necessitated by the changing complexities of

the outer environment).

That brings us to the question of how effectual processes can create nearly decom-

posable artifacts. Here the analogy of a patchwork quilt is very useful. Using effec-

tual processes to create firms and markets is somewhat like making a patchwork

quilt. Quilters begin the process with a random assortment of fabric patches and seek to create a meaningful and pleasing pattern in the quilt they make with them. In the

beginning the quilter could try different combinations of patches that suggest possi-

ble patterns and pictures in the finished quilt. While the availability of the particular

assortment of patches constrains the design, it does not determine it. A good quilter

can create intriguing and even meaningful patterns with the most chaotic of initial

assortments. Furthermore, as the quilt begins to take shape, quilters might seek

out particular patches outside their initial endowments, say from friends and garage

sales. Contingent upon the patches they find, they might change their initial designs as new possibilities emerge and they imagine better visions for the finished quilt.

It turns out, therefore, that such effectually created patchwork quilts can be rather

good examples of nearly decomposable systems. While particular patches have to

work with other patches to create an interesting pattern, sections can be re-worked

without redoing the entire quilt as the quilt grows larger. A causal analogy to this

effectual quilt would be a jigsaw puzzle, where the picture is already there, and the

pieces are merely to be assembled ‘‘correctly.’’ The patchwork quilt, however, has

no pre-determined pattern and depends almost entirely on the imagination of the

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 213

quilter and his or her mother wit in transforming unexpected contingencies into op-

portunities. In general, while causal models are tethered to goals, effectuation is un-

moored from specific goals enabling the effectuator not only to change particular

goals, but to create multiple new ends that could not have been foreseen at the be-

ginning of the process. So too the effectual entrepreneur begins with who she is, what she knows, and

whom she knows, to discover at least one customer or partner who is interested in

a product or service she can offer. Thus the first stable configuration of product/

stakeholder/environment comes into existence (perhaps after several aborted starts).

But the first stable configuration changes the means now available to her – her

knowledge corridors expand, her social networks grow larger and even her identity

is enhanced, through reputational and legitimation effects for example. Depending

on who the first stakeholder is and what he or she is interested in, the effectuator starts expanding the initial configuration and adds new configurations in a contin-

gent (and usually path dependent) fashion. Throughout the process, she seeks to

tie the different pieces together through innovative yet meaningful themes that get

embodied in mission statements, business plans, marketing brochures and press kits.

While the bottoms-up building block by building block process reduces costs of fail-

ure, the continual effort to create a unified identity allows successes to cumulate,

learning to occur and competitive competencies to be forged.

In this way, effectuation too creates nearly decomposable artifacts. Firms cannot be completely decomposable or 100% modular, if they are to have a strong identity

that inspires loyalty and trust with internal stakeholders. Yet, they need to be some-

what decomposable, so negative feedback from a variety of stakeholders can be in-

corporated to re-work parts of the firm as it grows and endures in the marketplace. It

is this particular opportunity to perceive and harness advantages both from the in-

terdependence of parts and their independence that gives effectually created nearly

decomposable entities a peculiar edge in evolving faster and enduring longer.

In the spirit of one of Simon�s favorite storytellers, Borges (1980, p. 107), who said, ‘‘I�ve observed that people tend to prefer the personal to the general, the con- crete to the abstract,’’ I will now provide an extract from one of the protocols in my

study. The extract lucidly illustrates how effectuation stitches together nearly decom-

posable firms. I use the extract, not as evidence for the existence of effectual pro-

cesses, but merely as an illustration of how they may build near-decomposability

into economic artifacts. Here the subject has been asked his opinion as to the growth

possibilities for an imaginary firm that begins with a single imaginary product, a sim-

ulation game of entrepreneurship. Notice how he begins by not showing much faith in the product, but gradually imagines himself into the vision of a great company

(see phrases in bold font). Notice also that at least thrice during the protocol he

strives to tie together the different bits and pieces he is imagining through a common

theme or an ‘‘identity’’ of sorts (see italicized phrases).

‘‘This company could make a few people very rich, but it cannot. . . I dont think it could ever be a huge company. The basic concept is a business sim-

ulator. . . startup simulator. . . so. . . in the same way in a jet simulator you

214 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

can hop in and fly something electronically and not blow it up. . . so you can hop into a business situation and practice and get a lot of reflexes

built up and thought processes built up up front. So. . . a successful launch of the first product with a big marketing sales push to penetrate

as many different markets as we could. . . might have a successful second product. For example, you could have a product which is how to succeed,

prosper, grow and get promoted within a large company. Making an

equivalent product for the quote organization person as opposed to the

entrepreneur would give you market of everybody with aspirations at

IBM, AT&T, Exxon etc. etc. so. . . That product could be a follow-on product. . . the research would be similar, the product development would be similar, and so the production part would be equivalent and some of

the same marketing channels would also work. You could make another product, would be, for students. How do I graduate in the top 10% of

your class at Stanford, or Harvard or Yale. And there. . . you could sim- ulate the learning process in the classroom. and research traits that tend

to make you successful or not. study habits that tend to make you suc-

cessful or not. and. . . a lot of how to be a good student is teachable. A lot. In my case for example, I took – So there are studying habits that

I�m aware of and you can do research on successful students and you could develop a profile that the. . . marketing pitch of which should be. . . students who graduate in the top 10% of a college class aren�t just smart in an accident. They have different habits and ways of doing busi-

ness that cause them to be successful and those are neither genetic nor in-

telligence related. . . they are learnable. So there�s your. . . now you got a product that can. . . you can sell to every student in the country. uhm. . . so we talked about entrepreneur business, big business, students, so we’re really talking about any learning in an interactive situation where simula-

tion is a benefit. So you got. . . next there is negotiation. . . there are books on negotiators. . . how to negotiate. . . famous books. . . here you could. . . in reading a book about negotiation would be less effective than having

an interactive 3D game about negotiation. So there you could practice

being a good negotiator. And that would work. There�s not a salesman in the United States who wouldn�t buy one of those. How to sell you know so you got you know another learning situation where how you

act and how you push people can can help you sell better. so. . . there is sales. So I guess you could go on and on and then you could generalize the thing to any situation which requires some sort of technical knowl-

edge. . . technical knowledge of negotiating. . . technical knowledge of bio-molecules. . . which also involves human organization. . . people you have to deal with. . . both outside the company to get them to help. . . to work with them and inside the company to get them to understand

what is the company�s methods objectives etc. So an organization in a learning situation with technical requirements. That simulation that had

those traits so now you can. . . I gave four five endeavors. . . you can

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 215

expand that so. . . maybe I�m gonna change my opinion about the growth potential for the company. . . The company could.. it is easy to see how within an hour you could name ten products and the ten products would

address huge markets like all employees in Fortune 500 companies that..

who are rich enough to pay hundred dollars for it. So now all of a sudden you can see it�s a software that could be a. . . could be a hit on the scale of Lotus.. what Lotus was to the spreadsheet world. And therefore you could

see a several hundred million dollar company coming from it.’’

To summarize the exposition so far, both effectuation and near-decomposability exploit locality and contingency in the evolution of the artifact. Just as effectuation

creates rapidly evolving artifacts that leverage the interdependence of parts to exploit

locality and contingency, so near-decomposability in the structure of such systems

leverages independence of parts to exploit the same locality and contingency. While

effectuation stitches together pieces of entrepreneurial fabric into economic quilts

that continue to make sense in an interactive and dynamically changing environment,

near-decomposability identifies lines of ‘‘tearing’’ so that pieces can be re-worked

in synchrony with the overall pattern as the needs imposed by the environment change.

Together they can provide a convincing explanation for the creation and growth

of the firms that we see in the real world. One way to substantiate such an explana-

tion would be to analyze the historical evidence already available to us. For example:

Wedgwood Pottery (Koehn, 1997), General Electric (Baldwin, 1995), U-Haul (Silver,

1985) and AES Corp (Waterman, 1990) all contain evidence as to how effectuation

processes have built large and rapid-growth firms with built-in near-decomposabil-

ity in their organizational structures. More general histories of the spread of ‘‘divi- sional’’ architectures through American industry can be found in Drucker (1947)

and Chandler (1962). Today, we can see numerous new examples of companies that

grow through franchising, joint ventures, and more recently, through ‘‘affiliate’’ pro-

grams pioneered by internet companies such as Amazon.com.

3.3. A vision for the effectual artifact

My research had already shown that entrepreneurs (rightly or wrongly) did set out

to design firms and to a considerable extent, even markets for their firms through the

logic of control. In this sense, entrepreneurship is a science of the artificial. But as

Simon points out in his book (1996, p. 113), ‘‘The previous chapters have shown that a science of artificial phenomena is always in imminent danger of dissolving and van-

ishing. The peculiar properties of the artifact lie on the thin interface between the

natural laws within it and the natural laws without. What can we say about it? What

is there to study besides the boundary sciences – those that govern the means and the

task environment?’’

He then goes on to explain what the contents of a science of the artificial might

consist of, ‘‘The artificial world is centered precisely on this interface between the in-

ner and outer environments; it is concerned with attaining goals by adapting the for-

216 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

mer to the latter. The proper study of those who are concerned with the artificial is

the way in which that adaptation of means to environments is brought about – and

central to that is the process of design itself.’’ It is here that entrepreneurship builds

on Simon�s formulation of a science of the artificial and moves it toward new hori- zons. I would like to argue that the design of entrepreneurial firms in general might involve something more than adaptation of the inner environment to the outer; it

might involve negotiation between the two. That is because, more often than not,

the environments of entrepreneurial firms (as well as markets in general) consist of

the contingent decisions of other human beings.

Without belaboring the point too much, I am not alone in my thesis that markets

are not ‘‘natural phenomena’’ based upon economic inevitability or even human ne-

cessity. However much economists might argue that de gustibus non disputandum est,

there is considerable historical and other types of evidence that marketers and entre- preneurs do succeed in their efforts to shape the preferences and tastes of their cus-

tomers. As early as 1939, Schumpeter pointed out, ‘‘It was not enough to produce

satisfactory soap, it was also necessary to induce people to wash’’ (Schumpeter,

1939, p. 243). More recently, Carpenter and Nakamoto (1989) theorize based on em-

pirical data that the practice of branding is essentially the formation of new prefer-

ence structures in the psyches of consumers. In a recent book, Koehn (2001)

chronicles several entrepreneurs from Wedgwood to Dell who created highly success-

ful enduring brands. Other economists too have argued against our assumptions of markets as something exogenous to the economic process, or something to be as-

sumed as a ‘‘given’’ in our analyses. Olson and Kahkonen (2000, p. 1) put it as fol-

lows, ‘‘The fourth primitive of economic thought – and of most lay thinking on

economics – is so elemental and natural that it is usually not even stated explicitly

or introduced as an axiom in formal theorizing. It is the half-conscious assumption

that markets are natural entities that emerge spontaneously, not artificial contri-

vances or creatures of governments.’’ Finally, Arrow (1974, p. 8) admits, ‘‘Although

we are not usually explicit about it, we really postulate that when a market could be created, it would be.’’

Therefore, if we do not take markets as completely exogenous to the economic

process, and view them instead as preferences being formulated and decisions being

made by a set of human beings that can be influenced by the actions of the entrepre-

neur, the effectual artifact of entrepreneurship (e.g. the firm) does not just adapt to

its external environment (‘‘the market’’). Instead it has the option of negotiating with

its environment, to shape the environment at least partially in its own image, just as

it adapts other aspects of its internal self to effectively reflect the environment. One might of course argue that negotiations too are a form of adaptation. That

would raise the question, ‘‘Adaptation to what?’’ Effectuation is fundamentally dif-

ferent from other forms of adaptation such as those involved in biological evolution.

While the effectuator does adapt to changing circumstances outside his control, he

also actively seeks to reshape his environment through those parameters that do sub-

mit to his control. Effectuation, therefore, includes adaptive techniques such as im-

provization, socio-psychological techniques such as bracketing and enactment, and

aggressively effectual techniques of negotiation such as lobbying the government,

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 217

participating in standards bodies, and obtaining pre-commitments from influential

stakeholders etc.

Understanding the role of such negotiations between inner and outer environ-

ment, whether as part of an adaptive process or in parallel to it, should form one

of the core areas for research in entrepreneurship. Another key area for research sug- gested by the formulation of entrepreneurship as a science of the artificial consists in

the role of firm failures in allowing entrepreneurs to understand limiting properties

of the artifacts they create in relation to the environments with which or within

which they negotiate.

In sum, the theory of effectuation suggests that entrepreneurship is indeed a sci-

ence of the artificial and that it builds on at least four key ideas in Sciences of the

Artificial:

1. Natural laws constrain but do not dictate our designs – i.e., within the constraints

of natural law, our designs are contingent on our imagination; there is nothing

intrinsically ‘‘inevitable’’ about them. Implication for entrepreneurship: Given

who we are, what we know, and whom we know, we can build a variety of effec-

tual artifacts by focusing on what we can do, rather than continually worrying

about what we ought to do, given pre-determined goals.

2. We should seize every opportunity to avoid the use of prediction in design. Impli-

cation for entrepreneurship: Designing without final goals allows us to free our- selves from the pitfalls of prediction so we can use other mechanisms such as

the scientific method, or the garbage can (Cohen, March, & Olsen, 1972), or

the effectual logic of control.

3. Locality and contingency govern the sciences of the artificial. Implication for en-

trepreneurship: Contingencies can be viewed as opportunities to be exploited

rather than as misfortunes to be avoided; while successes and failures are always

local, cumulative learning is still possible.

4. Near-decomposability is an essential feature of enduring designs. Implication for entrepreneurship: Effectual processes that exploit locality and contingency through

both interdependence and independence of parts are more likely to result in en-

during firms.

4. Conclusion

When I came up with the first draft of the conference paper, Simon had some dif- ficulties with my cooking and quilting metaphors. He was used to watchmakers and

clicking safes. But eventually as we talked and emailed back and forth, and partic-

ularly when we started discussing the role of locality and contingency in the two the-

ories, he began to see that the quilting example was particularly apt for what we were

trying to establish in the paper and admitted to me in an email that he was ‘‘more

persuaded than before of the effectiveness of the cookery and quilting metaphors.’’

It was one of those treasured moments when I was able to surprise him in answer

to his familiar question, ‘‘So what do we know now that we did not know the last

218 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220

time we met?’’ He will always live in that question for me, and I am filled with grat-

itude I got to explore it with him so many times. Rest assured, Herb, I will keep try-

ing to catch you by surprise one of these days.

Acknowledgements

I thank Mie Augier for inviting me to write this essay and being patient with my

tardiness on delivery. I thank Jim March for his conversation and encouragement in

several of my scholarly efforts. I am grateful to Anil Menon for his comments and

feedback on the paper and particularly for suggesting the relevance of the myths

of Parsival and Sisyphus for scholarly work. I thank Martin Schultz and Rob Wilt-

bank for comments on an earlier version of this paper.

References

Arrow, K. J. (1974). Limited knowledge and economic analysis. The American Economic Review, 64(1), 1–

10.

Baldwin, N. (1995). Edison: Inventing the century. Hyperion.

Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of Management, 17(1), 99–

121.

Borges, J. L. (1980). Blindness. In: Seven nights. Translated by Eliot Weinberger. New York: New

Directions.

Brockhaus Sr, R. H. (1980). Risk taking propensity of entrepreneurs. Academy of Management Journal,

23(3), 509–520.

Carpenter, G. S., & Nakamoto, K. (1989). Consumer preference formation and pioneering advantage.

Journal of Marketing Research, 24(3), 285–298.

Chandler, A. D. (1962). Strategy and structure: Chapters in the history of the industrial enterprise.

Cambridge, MA: MIT Press.

Cohen, M. D., March, J. G., & Olsen, J. P. (1972). A garbage can model of organizational choice.

Administrative Science Quarterly, 17(1), 1–25.

Drucker, P. F. (1947). Big Business: A study of the political problems of American capitalism. London:

W. Heinemann Ltd.

Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal reports as data. Cambridge, MA: The

MIT Press.

Gartner, W. B. (1988). Who is an entrepreneur is the wrong question. American Journal of Small Business

(Spring), 11–32.

Knight, F. H. (1921). Risk, uncertainty and profit (1933 ed). New York: Houghton Mifflin.

Koehn, N. F. (1997). Josiah Wedgwood and the first industrial revolution. In T. K. McCraw (Ed.),

Creating modern capitalism (pp. 17–48). Cambridge, MA: Harvard University Press.

Koehn, N. F. (2001). Brand new: How entrepreneurs earned consumers� trust from Wedgwood to Dell. Boston: Harvard Business School Press.

March, J. G. (1982). The technology of foolishness. In J. G. March, & J. P. Olsen (Eds.), Ambiguity and

choice in organization. Bergen: Universitetsforlaget.

Marengo, L., Frenken, K., & Valente, M. (1999). Interdependencies, nearly decomposability and

adaptation. In T. Brenner (Ed.), Computational techniques for modelling learning in economics. Boston:

Kluwer Academics.

McKibben, G. C. (1998). Cutting edge: Gillette�s journey to global leadership. Boston: Harvard Business School Press.

S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220 219

Olson, M., & Kahkonen, S. (2000). A not-so-dismal science: A broader view of economics and societies.

Oxford: Oxford University Press.

Palich, L. E., & Bagby, D. R. (1995). Using cognitive theory to explain entrepreneurial risk-taking:

Challenging conventional wisdom. Journal of Business Venturing, 10(6), 425–439.

Sarasvathy, S. D. (2000). Report on the seminar on research perspectives in entrepreneurship. Journal of

Business Venturing, 15(1), 1–58, Participating authors: K. Arrow, M. Kamien, M. Olson, D. Sexton, H.

Simon, & S. Venkataraman.

Sarasvathy, S. D., & Simon, H. A. (2000). Near decomposability, effectuation, and the speed of growth of

entrepreneurial firms. Presented at ‘‘The first annual technology entrepreneurship research policy

conference’’, May 19 and 20, 2000. Robert H. Smith School of Business, University of Maryland.

Available at <www.effectuation.org>.

Sarasvathy, S. D. (2001a). Causation and effectuation: Towards a theoretical shift from economic

inevitability to entrepreneurial contingency. Academy of Management Review, 26(2), 243–288.

Sarasvathy, S. D. (2001b). Effectual reasoning in entrepreneurial decision making: Existence and bounds.

Best paper proceedings, Academy of Management 2001. Washington, DC, August 3–8, 2001.

Sarasvathy, S. D., & Kotha, S. (2001). Dealing with Knightian uncertainty in the new economy: The real

networks case. In J. Butler (Ed.), Research on Management and Entrepreneurship: Vol. 1 (pp. 31–62).

Greenwich, CT: IAP Inc.

Schilling, M. A. (2000). Toward a general modular systems theory and its application to interfirm product

modularity. The Academy of Management Review, 25(2), 312–334.

Schumpeter, J. A. (1939). Business cycles: A theoretical, historical, and statistical analysis of the capitalist

process (2 Vols.). New York: McGraw-Hill.

Silver, D. A. (1985). Entrepreneurial Megabucks. John Wiley & Sons.

Simon, H. A. (1969). The architecture of complexity. Sciences of the artificial. Cambridge, MA: MIT Press.

Simon, H. A. (1996). The architecture of complexity. Sciences of the artificial (third ed). Cambridge, MA:

MIT Press.

Waterman, R. H. (1990). Adhocracy: The power to change. Knoxville, TN: Whittle Direct Books.

Weick, K. E. (1979). The social psychology of organizing. Reading, MA: Addison-Wesley.

220 S.D. Sarasvathy / Journal of Economic Psychology 24 (2003) 203–220