EDMG611Wk6

Course Research: Using the Case Method to Build and

Teach Management Theory CLAYTON M. CHRISTENSEN

Harvard Business School

PAUL R. CARLILE Boston University

Some in the Academy have questioned the useful- ness of case studies in teaching sound manage- ment theory (Shugan, 2006). Our research and ex- perience suggest exactly the opposite—that case studies can unite the development of theory with the teaching of it in a single enterprise we’ll call course research. Conclusions such as those that Shugan and others have reached stem from miscon- ceptions about the relationship of research, theory, case studies, and teaching. In fact, the proper use of case studies in teaching can help faculty resolve a basic dilemma of academia: Promotion is often based upon our published research, and we find that responsibilities to teach detract from the mandate to publish. When approached properly, case studies can transform teaching into research, which enrolls students as “course researchers,” whose class participation can be exceptionally valuable in the theory-building process.

Conclusions such as those that Shugan and others have reached stem from misconceptions about the relationship of research, theory, case studies, and teaching.

Compartmentalizing our world into mutually ex- clusive realms of research and teaching wastes substantial intellectual energy and insight—our own, and that of our students. In the research side of our lives many of us have learned to train and trust students as junior partners in our theory- building efforts. We readily engage doctoral, MBA, and even undergraduate students as research as- sistants to collect and examine data. We discuss with these students the hypotheses that emerge from their analyses, and ask them to plunge back

into the data to refine these hypotheses. When we walk into the classroom with those same students, however, many of us mentally shift gears into a mode of conveying theory, not developing it. While we work together as trusted research partners out- side the classroom, we bifurcate as instructors and instructees inside the classroom. We rarely frame classroom activities as opportunities for theory building—for examining the phenomena, catego- rizing it, synthesizing hypotheses, and testing them to find anomalies that will yield improved theory. Yet classroom work can be structured as theory-building activity, just as out-of-classroom work can be, if case studies are used properly.

To show how research, course development, and teaching can be knit together in this way, we will first present a model of how theory can be built. Although there are other useful models of theory building, the simple one we employ here has proven helpful to us and many students and col- leagues as we have sought to build, teach, and publish better theory for management.1

A THEORY-BUILDING PROCESS

We define “theory” very simply—as a body of un- derstanding. The building of theory occurs in two major stages: a descriptive stage and a prescrip-

1 This model was developed first by synthesizing models of theory building that have been developed by scholars of this process in a range of fields. These include Kuhn (1962) and Popper (1959) in the natural sciences; and Campbell and Stan- ley (1963), Glaser and Strauss (1967), Stinchcombe (1968), Roeth- lisberger (1977), Yin (1984), R. Kaplan (1986), Weick (1989), Eisen- hardt (1989), and Van de Ven (2000) in the study of management and social science. To this synthesis we have added our own observations, derived from studying the research efforts of var- ious faculty members and doctoral students at Harvard, MIT, Stanford, and the University of Michigan.

� Academy of Management Learning & Education, 2009, Vol. 8, No. 2, 240–251.

........................................................................................................................................................................

240 Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express written permission. Users may print, download or email articles for individual use only.

tive stage (Carlile & Christensen, 2005). Within each, theory builders proceed through three steps. As we describe below, case-centered teaching can help students and faculty iterate through these three steps again and again, and collaboratively build better theory. In other words, theory building, teaching, and learning converge to comprise the same process. We first explain how these steps are used to build descriptive theory.

The Descriptive Stage of Theory Building

The descriptive stage of theory building is a pre- liminary stage because theory builders must pass through it in order to develop prescriptive theory. The three steps that researchers who are building descriptive theory generally pass through are ob- servation, categorization, and association.

Step 1: Observation

In the first step of theory building researchers care- fully observe phenomena and describe and mea- sure what they see. This is important because if subsequent researchers cannot agree upon the de- scriptions of phenomena, then improving theory will prove difficult. Early management studies such as The Functions of the Executive (Barnard, 1939) and Harvard Business School cases written in the 1940s and 1950s were primarily descriptive work of this genre. Observation of phenomena can occur anywhere along the continuum, from analy- sis of huge databases on the one end, to field- based, ethnographic observation on the other. This stage of research is depicted in Figure 1 as the base of a pyramid because it is a necessary foun- dation for the work that follows.

Researchers in this step often develop what we term constructs. Constructs are abstractions that help us rise above the messy detail to understand the essence of what the phenomena are and how

they operate. Some examples are Bower’s (1970) constructs of impetus and context, which describe how momentum builds behind certain investment proposals and fails to coalesce behind others in the resource allocation process. This has helped a generation of strategy researchers understand how investment decisions get made. The path- breaking work of Spear (2005) and Spear and Bo- wen (1999) on the Toyota Production System is built upon their constructs of activities, connections, and pathways. Economists’ concepts of “utility” and “transaction costs” are likewise constructs—ab- stractions developed to help us understand a class of phenomena they have observed. We would not label the constructs of utility and transaction cost as theories, however. They are part of theories— foundations upon which bodies of understanding about consumer behavior and organizational structure have been built.

Step 2: Classification

Once the phenomena is observed and described researchers in the second stage typically classify it into categories. In the descriptive stage of theory building, the classification schemes that scholars propose generally are defined by the attributes of the phenomena. For example, a categorization from the study of strategy, diversified versus focused firms. Lean and inventory-carrying manufacturing systems are categories that have emerged from research in operations. Burgelman and Sayles (1986) built upon Bower’s (1970) construct of context by categorizing two different types: organizational and strategic. Categorization helps researchers simplify the world in ways that highlight possibly consequential rela- tionships between the phenomena and the outcomes of interest. Management researchers often refer to these descriptive categorization schemes as frame- works or typologies.

Step 3: Defining Relationships

In the third step, depicted as the peak of the pyra- mid in Figure 1, researchers explore the associa- tion between the category-defining attributes and the outcomes observed. Researchers recognize and make explicit differences in attributes that corre- late most strongly with the the outcomes of interest and the magnitude of those attributes. Techniques such as regression analysis can be useful in defin- ing these correlations. Often we refer to the output of studies at this step as models.

Descriptive theory that quantifies the degree of correlation between the category-defining at- tributes of the phenomena and the outcomes of

Predict

Anomaly

Confirm

Observe, describe & measure the phenomena

(constructs)

Categorization based upon attributes of phenomena

(frameworks & typologies)

Statements of association

(models)

Ded uc

tiv e pr

oc es

s Inductive process

FIGURE 1 The Process of Building Theory

2009 241Christensen and Carlile

interest is generally able to make probabilistic statements of association representing average tendencies. For example, Hutton, Miller, and Skin- ner (2000) examined how stock prices respond to the phrasing in earnings announcements. Their analysis enabled them to assert that on average across the entire sample of companies and an- nouncements, delivering earnings announcements in a particular way would lead to the most favorable (or least unfavorable) reaction in stock price. But managers of specific companies cannot know from descriptive theory whether following that average formula will lead to the hoped-for outcome in their particular situations. This requires the development of prescriptive theory, as we show below.

How Theory Is Improved Within the Descriptive Stage

When researchers move from the bottom to the top of the pyramid through observation, categoriza- tion, and association—and in so doing give us constructs, frameworks, and models—they have followed the inductive portion of the theory-build- ing cycle. Researchers can then improve theory by cycling from the top down to the bottom of this pyramid in the deductive portion of the cycle, seek- ing to “test” the hypotheses that had been induc- tively formulated. This most often is done by ex- ploring whether the same correlations exist between attributes and outcomes in a different set of data than the data from which the hypothesized relationships were induced. When scholars test a theory on a new data set (whether the data are numbers in a computer, or are field observations taken in a new context), they sometimes find that the attributes of the phenomena in the new data correlate with the outcomes as predicted. Such “tests” confirm that the theory is of use under the conditions observed. However, researchers who stop at this point simply return the model to its place atop the pyramid tested but unimproved.

It is only when an anomaly is identified—an outcome for which the theory can’t account—that an opportunity to improve theory occurs (Kuhn, 1962). As Figure 1 suggests, discovery of an anom- aly gives researchers the opportunity to revisit the foundation layers in the theory pyramid—to define and measure the phenomena more precisely, or to cut it into alternative categories—so that the anomaly and the prior associations of attributes and outcomes can all be explained. Every com- plete cycle up and down the theory-building pyra- mid consists of an inductive side and a deductive side. Theory-building efforts stall when research- ers drop the baton and declare victory having run

only an inductive or deductive half-cycle up or down the theory pyramid.2

Kuhn (1962) observed that confusion and contra- diction are the norm during descriptive theory building. It is characterized by a plethora of cate- gorization schemes because the phenomena have many different attributes. In the study of how tech- nological innovation affects the fortunes of firms, for example, an early attribute-based categoriza- tion scheme was radical versus incremental inno- vation. The correlations that were measured upon it concluded that established firms on average do well when faced with incremental innovation, but stumble in the face of radical change. But there were anomalies to this generalization: established firms that successfully implemented radical tech- nology change. To account for these, Tushman and Anderson (1986) offered a different categorization: competency-enhancing versus competency- destroying technological changes. This resolved many of the anomalies, but subsequent research- ers uncovered new ones for which the Tushman– Anderson scheme could not account. Henderson and Clark’s (1990) categories of modular versus architectural innovations; Christensen’s (1997) cat- egories of sustaining versus disruptive technolo- gies; and Gilbert’s (2001) threat-versus-opportunity framing each uncovered and resolved anomalies for which the work of prior scholars could not ac- count. This body of understanding has improved and become quite useful to practitioners and sub- sequent scholars (Adner & Zemsky, 2005; Daneels, 2006) because these scholars articulated theories that could yield anomalies.

Anomalies are valuable in theory building be- cause the discovery of an anomaly is the enabling step to less ambiguous description and measure- ment and to identifying and improving the categori- zation scheme in a body of theory. Researchers whose goal is to “prove” a theory’s validity often view discovery of an anomaly as failure, and there- fore, search for reasons to exclude outlying data points in order to get more significant measures of statistical fit. There typically is more information in the points of outlying data than in the ones that fit the model well, however, because understanding the

2 We assert in this model that all theory-building efforts begin at the bottom of the pyramid. This is because Feyerabend (1975) and others have noted that observational abilities tend to be shaped by cognitive structures, previous experience, or some theory-in-use. Although individual researchers might feel that their work is has started at the top of the pyramid, we believe that generally the hypotheses that deductive theorists test were inductively derived from observation, consciously or uncon- sciously, by themselves or others.

242 JuneAcademy of Management Learning & Education

outliers or anomalies is generally the key to discov- ering problems in definition and measurement, and in formulating better categorization schemes. Re- searchers who seek to surface and resolve anoma- lies, therefore, tend to advance their fields more pro- ductively than those that seek to avoid them.

Researchers who seek to surface and resolve anomalies, therefore, tend to advance their fields more productively than those that seek to avoid them.

What Types of Data Can Be Trusted to Yield Reliable Theory?

A key measure of a theory’s quality is the reliabil- ity of the data that characterize the phenomena at the base of the pyramid. Many view numerical data as more trustworthy than qualitative data. But where does “objective” data come from? The data used in many research projects comes from compa- nies’ financial statements, for example. Is this objec- tive? Johnson and Kaplan (1987) showed quite con- vincingly that the numbers representing revenues, costs, and profits in financial statements are the re- sult of processes of estimation, negotiation, debate, and politics that can produce grossly inaccurate re- flections of true cost and profit. Even the “hardest” of numbers, such as those measuring prices and prod- uct performance, really are manifestations of the pri- oritizations, fudging measurements, exaggerations, and negotiations that occurred prior to a number appearing as a proxy for all of those things.

The healthiest mind-set for researchers is that nearly all data—whether presented in the form of a large quantitative data set on one extreme, or an ethnographic description of behavior on the other—are subjective. Numerical and verbal data alike are abstractions from a much more complex reality, out of which the researcher attempts to pull the most salient variables or patterns for examina- tion. Whereas the subjectivity of data from field- based, ethnographic research is glaringly appar- ent, the subjective etiologies of numerical data are hidden and cannot be reviewed. There should be no smugness among quantitative researchers about the alleged objectivity of their data, and no defensiveness amongst field researchers about the subjectivity of theirs. We are all in the same sub- jective boat, and are obligated to do our best to be humble and honest with ourselves and our col- leagues about where our data comes from as we participate individually within and collectively across the theory-building cycle.

There should be no smugness among quantitative researchers about the alleged objectivity of their data, and no defensiveness amongst field researchers about the subjectivity of theirs. We are all in the same subjective boat, and are obligated to do our best to be humble and honest with ourselves and our colleagues about where our data comes from as we participate individually within and collectively across the theory- building cycle.

More often than most of us realize, data about the phenomena comes packaged in the form of, or is extracted from, case studies. What is a case study? We define it simply as a multifaceted examination of a situation. Christensen’s (1992, 1997) work in the disk drive industry, by illustration, began with a com- plete, numbers-intensive history of the industry, which analyzed the technical specifications and components used in every product ever announced by any company in the world, from 1974 to the present. It examined every company that ever en- tered the industry, whether it sold any products or not, and evaluated data on sales by product category for all companies that did. That case was a complete census, not a statistical sample. Nested within that case were studies of six individual companies. These accounts relied on quantitative information supple- mented by some qualitative data extracted from in- terviews. Within each of those were accounts of in- dividual product development projects, built with some quantitative data and lots of qualitative infor- mation. These were also cases. And within each of these were case studies of individual people—built from qualitative, open-ended interviews.

The unit of analysis of a case may be a national economy, an industry, a period of trading on a stock exchange, a population of patients, a single patient, a company, a set of projects, the career experiences of a population of people, or a single experience of an individual. All are cases, and all examine data that are subjective abstractions or manifestations of a much more complex underly- ing reality.3 What is important, for the purpose of

3 This definition of a case is, we believe, broader than but not inconsistent with Yin’s (1984: 23) definition that “[a] case study is an empirical inquiry that 1) investigates a contemporary phe- nomenon within its real-life context; when 2) the boundaries between phenomenon and context are not clearly evident; and 3) multiple sources of evidence are used.”

2009 243Christensen and Carlile

building theory, is that “case studies, like experi- ments, are generalizable to theoretical proposi- tions” (Yin, 1984: 21). In the descriptive stage they are not, and need not be, generalizable to the pop- ulations or universes in which those situations are nested. Generalizability comes with prescriptive theory.

The Transition From Descriptive to Prescriptive Theory

The confusion that often stems from categorization conflicts in descriptive theory becomes resolved when researchers, often through detailed empiri- cal and ethnographic observation where n � 1, move beyond statements of correlation to define what causes the outcome of interest. This is de- picted in Figure 2 as a transition to the top of the pyramid of prescriptive theory, whose capstone is a statement of what causes the outcome of interest, not just what is correlated with it.

For many, the usefulness of a theory is its ability to predict the consequence of an action (van de Ven, 1989). Bazerman (2005) has noted that one rea- son why research in some fields has had so little influence on management is that many social sci- ence researchers choose not to be prescriptive. He shows that prescriptive theory is not only possible to develop in the social sciences, it is desirable.

Prescriptive theory, like its descriptive predeces- sor, still needs to be improved—and researchers

do this by following the same steps that were used in the descriptive stage. Hypothesizing that their statement of causality is correct, they cycle deduc- tively to the bottom of the pyramid to test the causal hypothesis: “If we observe these actions or events occurring, these should be the outcomes that we observe.” When they encounter an anom- aly, they then delve back into the lower levels of the pyramid. Most often, they resolve anomalies by revisiting the categorization stage—but this time with a different question. Rather than categorizing by the attributes of the phenomena, researchers identify the different situations or circumstances in which managers might find themselves. They do this by asking, when they encounter an anomaly, “What was it about the situation in which those managers found themselves, that caused the mechanism to yield a different result?” By asking this question as they cycle up and down the pyra- mid of prescriptive theory, anomaly-seeking re- searchers ultimately define a relatively complete set of the circumstances in which managers might find themselves when pursuing the outcomes of interest. This allows researchers to make contin- gent statements of causality—to show how and why the casual mechanism results in a different outcome, in each different situation. A prescriptive theory that is built upon well-researched catego- ries of circumstance can help managers predict accurately what actions will and will not lead to the desired result, given the circumstance in which

Anomaly

Predict

Ded uc

tiv e p

ro ce

ss

Confirm

Inductive process

Anomaly

ConfirmPredict

Categorization by the attributes of the phenomena

Statements of correlation

Ded uc

tiv e pr

oc es

s

Inductive processObserve, describe & measure the phenomena

Statement of causality

Observe, describe & measure the phenomena

Categorization of the circumstances in which we

might find ourselves

Prescriptive Theory

Descriptive Theory

FIGURE 2 The Transition From Descriptive Theory to Prescriptive Theory

244 JuneAcademy of Management Learning & Education

they find themselves. The circumstance-contingent predictability of prescriptive theory enables man- agers to know, in any given situation, what they ought to do.

The history of research into manned flight is a good way to visualize how this transition from descriptive to prescriptive theory occurs and how it is valuable. During the Middle Ages, would-be avi- ators did their equivalent of best-practices re- search. They observed the animals that could fly well and compared them with animals that could not. The vast majority of the successful fliers had wings with feathers on them, and almost all of those that couldn’t fly had neither of these at- tributes. This was quintessential descriptive the- ory. Pesky outliers such as ostriches had feathered wings but couldn’t fly; bats had wings without feathers and were very good at it; and flying squir- rels had neither and got by. But the R2 was so high that aviators of the time copied the seemingly sa- lient characteristics of the successful fliers in the belief that if they copied the characteristics of the “best-practices” fliers, they could fly, too. So they fabricated wings, glued feathers on them, jumped off cathedral spires, and flapped hard. It never worked. Disagreements emerged about the catego- rization scheme: debates of which of the birds’ attributes truly enabled flight, and which didn’t. For example, Roger Bacon wrote an influential pa- per asserting that the differentiating characteristic was birds’ hollow bones (Clegg, 2003). With solid bones, Bacon reasoned, man could never fly. He then proposed designs of machines that could flap their wings with sufficient power to overcome the disadvantage of solid human bones. But it still never worked. Armed only with the correlative statements of descriptive theory, aviators kept kill- ing themselves.

Then Daniel Bernoulli, through his careful study of fluid mechanics, identified a shape that we call an airfoil, which, when it cuts through air, creates a mechanism that we call lift. Understanding this causal mechanism made flight possible. But it was not yet predictable. In the language of this paper, the theory predicted that aviators would fly suc- cessfully when they built machines with airfoils to harness lift. Crashes were anomalies that Bernoul- li’s theory could not explain. Crashes, however, allowed researchers to revisit the categorization scheme. But this time, instead of slicing up the world by the attributes of the good and bad fliers, researchers asked, “What was it about the circum- stance that the aviator found himself in that caused the crash?” This then enabled them to im- prove equipment and techniques and articulate circumstance-contingent prescriptions: “This is

how you should normally fly the plane. But when you get in this situation, you need to fly it differ- ently in order to get the desired outcome. And when you get in this other situation, don’t even try to fly. It is impossible.”

Careful studies of anomalies allowed research- ers to identify the set of circumstances in which aviators might find themselves. When they then modified the equipment or developed piloting techniques that were appropriate to each circum- stance, manned flight became not only possible, but much more predictable. Discovery of the causal mechanism made flight possible. Categori- zation of the salient circumstances then enabled flight to be made more predictable.4

Establishing the Validity of Theory

Those who use management theory usually want to know whether to trust it, where it applies, and where it does not. To help us gauge whether and when we can trust it, Campbell and Stanley (1963) described two types of validity for a theory: inter- nal and external validity. In this section we dis- cuss how these concepts relate to our process model of theory building and describe how re- searchers can make their theories more valid on both of these dimensions.

Internal Validity

A theory’s internal validity is the extent to which (1) its conclusions are unambiguously drawn from its premises; and (2) the researchers have ruled out plausible alternative linkages of the phenomena with the outcomes of interest. Measures of statisti- cal significance and goodness of fit are common methods of assessing internal validity.

Researchers improve the internal validity of a theory when they examine the phenomena from as many perspectives as possible. As long as there’s a possibility that another researcher could say, “There’s a different explanation for why this hap- pened,” we cannot be assured of a theory’s internal validity. Scholars who examine the phenomena and outcomes of interest from the perspectives of other potentially relevant academic disciplines or, who take the viewpoints of different functional groups in a company, can either incorporate what they learn into their explanations of causality, or rule out other explanations so that theirs is the only plau- sible one left standing. Burgelman’s (2002) recon-

4 We thank Matthew Christensen for teaching us with this example.

2009 245Christensen and Carlile

struction of the workings of each functional group within Intel as the company exited the dynamic ran- dom access memories (DRAM) business and threw its resources behind its microprocessor strategy is an outstanding example of such research.

External Validity

The external validity of a theory is the extent to which a relationship that was observed between phenomena and outcomes in one context can be trusted to apply in different contexts as well. Mea- sures of statistical significance and goodness of fit are not relevant measures of external validity. To illustrate why, consider Christensen’s experience after publishing his prescriptive theory of disrup- tion, which had been inductively derived through empirical analyses of the history of the disk drive industry. Those who read his early papers instinc- tively wondered, “Does this theory apply outside the disk drive industry?” To address these con- cerns when writing The Innovator’s Dilemma, Christensen (1997) sought to establish the general- izability of the theory by “testing” it on data from as disparate a set of industries as possible, includ- ing hydraulic excavators, department stores, steel, computers, motorcycles, diabetes care, accounting software, motor controls and electric vehicles. De- spite the variety of industries in which the theory seemed to have explanatory power, executives from industries that weren’t specifically studied kept asking, “Does it apply to health care? Educa- tion? Financial services?” When Christensen pub- lished papers that applied the model to these in- dustries, the response was, “Does it apply to telecommunications? Regulated industries? The German economy?” A disk drive engineer even asked, “It clearly applies to the history of the disk drive industry. But does it apply to its future as well?” As these illustrate, it is simply impossible to establish a theory’s external validity by testing it on data. There will always be another set upon which it hasn’t yet been tested, and the future will always lie just beyond the reach of data.

External validity can only be created through categorization, and there is a process by which theory progresses in this direction. When an un- derstanding of causality first emerges and it is used to make ex post or ex ante predictions, re- searchers uncover anomalies—instances where the prediction failed. Then by asking, “What was it about the circumstance that caused this unex- pected outcome?” researchers come to understand the categories of circumstances. A prescriptive the- ory that is based upon a robust statement of causal- ity is externally valid when the categories of circum-

stance are mutually exclusive and collectively exhaustive. Mutually exclusive categorization al- lows managers to say, “I am in this circumstance and not that one.” And collectively exhaustive cate- gorization would assure us that all situations that managers might find themselves in with respect to the phenomena and outcomes of interest are ac- counted for in the theory. In concept (though never in practice) as researchers asymptotically approach a complete understanding of causality and circum- stance, the theory’s predictions would be accurate 100% of the time, because the outcomes that the the- ory predicts are contingent upon a set of collectively exhaustive, mutually exclusive situations in which managers might find themselves.

Several scholars have examined the improve- ment in predictability resulting from the transition from the attribute-based categorization of descrip- tive theory to the circumstance-based categoriza- tion of prescriptive theory. Consider Lawrence and Lorsch’s (1967) “contingency theory,” for example. They showed that the best way to organize a com- pany depended upon the circumstances in which the company was operating. In our model, contin- gency is not a theory per se. Rather, contingency is the categorization scheme, and is a crucial ele- ment of every prescriptive theory. Glaser and Strauss’s (1967) treatise on grounded theory, when viewed through the lens of our process model, ac- tually is a book about categorization—although they use different words. Their term substantive theory corresponds to the attribute-bounded cate- gories of descriptive theory. And their concept of formal theory corresponds to the categories of cir- cumstance in prescriptive theory.

Efforts to identify “the best practices of success- ful companies” almost always skip the categoriza- tion step in the theory process. Researchers in this mode study a set of successful companies, find that they share certain characteristics, conclude that they have seen enough, and then assert that if all managers would imbue their companies with the characteristics of these successful companies, they would be similarly successful. There is rarely a standard remedy that applies to all companies in all situations. By not taking the care to define the circumstances in which their statement of causal- ity would lead to success and when it would not, such researchers discredit their work.

TOWARD A PRESCRIPTIVE, CIRCUMSTANCE- CONTINGENT THEORY OF TEACHING

Few of us who instruct business students have framed what we do as employing a theory of in- struction, but we all do, at least unconsciously. The

246 JuneAcademy of Management Learning & Education

theory that most seem to employ is still in the descriptive stage. Shugan (2006), for example, cat- egorizes by the attributes of the courses: lecture- based versus case-based. The correlative state- ment he makes at the peak of his theory pyramid is that the lecture method may be a better way to convey theory, for a host of reasons.

The issues Shugan raises aren’t limited to busi- ness schools. The debate about problem- or case- based teaching versus lecture teaching is a hot one in medical schools, for example. As with all theories in the descriptive stage, educators are employing many categorization schemes—not just case versus lecture—in debates about which meth- ods of instruction are better, because there are many attributes of classes. Controversies center on whether distance learning can be as effective as in-person instruction; whether students in smaller classes learn better than those in larger ones; and whether project-based learning might be superior to subject-focused instruction. Researchers at the third step of the pyramid of these descriptive the- ories do their best to measure, on average, which methods of teaching work better than others. And true to form for bodies of understanding that are still in the descriptive stage, the evidence isn’t conclusive. There are pesky anomalies in the form of teachers, courses, or schools that seem to beat the odds, producing extraordinary results through methods that, on average, aren’t as efficacious.

In a very preliminary way, we hope to offer a prescriptive theory to faculty members who are attracted by the opportunity to engage in course research: to design courses and teach them in ways that feed research rather than detract from it. There are three circumstances—which essentially are positions along a continuum—in which faculty members might find themselves when needing to design and teach a course:

1. When wanting to teach a course on a topic around which a body of theory has not yet coalesced, and the teacher–researcher hopes to play a key role in its development;

2. When a significant body of understanding ex- ists, and the teacher wants to improve the the- ory by taking it from the descriptive to the prescriptive stage, or by sharpening the cate- gorization scheme by seeking and then resolv- ing anomalies;

3. When the body of theory to be taught is so well established that the marginal returns on ef- forts to improve it are minimal.

Course Research That Inductively Builds Theory

On occasion faculty find themselves needing to design and teach a course on a subject about

which a body of theory has not yet emerged. Many faculty members found themselves in this situa- tion when the Internet was emerging in the late 1990s. Others find themselves in this circumstance today, needing to teach students about the col- lapse of our financial markets. In such circum- stances, designing the course to move up the in- ductive side of the theory pyramid can be very productive. For example, in 1998 Harvard Business School Professor Kent Bowen observed that nearly 40% of his school’s graduates ultimately found themselves owning and running small businesses. These weren’t flashy venture capital-backed tech- nology start-ups. Rather, they were printing com- panies, service bureaus, retail stores, and so on. He decided to create a course, called Building and Running a Small Business (BRSB) to better serve students who were headed toward that destina- tion. Then discovering that little had been written about how to run low-tech, slow-growth compa- nies, he tackled the problem with an inductive course design.

Bowen first researched and wrote a series of cases that simply described what managers in these sorts of companies worry about and do. The purpose of each case discussion was to help the professor and students to understand the phenom- ena thoroughly. After a few classes, Bowen paused and orchestrated a class discussion to define pat- terns in the phenomena—to begin categorizing by type of company, type of manager, and type of problem. They next explored the association be- tween these types and the outcomes of interest. This portion of Bowen’s course had an inductive architecture that moved up the theory pyramid. In the next portion of his course, armed with their preliminary body of theory, Bowen and his stu- dents cycled down the deductive side of the pyra- mid to examine more companies in a broader range of circumstances. This allowed them to dis- cover things that their initial models could not explain. This helped them to improve their con- structs, refine their classification schemes, and im- prove their understanding of what what actions tended to yield what results, and why.5

Course Research That Deductively Improves Theory

The second circumstance is where well-researched theories pertaining to a field of management al-

5 The information about Bowen’s course comes from the first author’s personal interviews with Professor Bowen and his ob- servation of Bowen’s activities.

2009 247Christensen and Carlile

ready exist: theories that need to be improved through anomaly-seeking research. To illustrate how cases can be usefully employed to teach and build theory in this circumstance, we’ll offer a per- sonal case study of the first author’s attempts to build a course around his research on innovation. Along with Bowen’s work, Christensen’s experi- ences with this effort were pivotal cases in the distillation of this theory of course design and teaching.

For 50 years the default mode of course architec- ture at the Harvard Business School has been in- ductive in character. As described above, there was little management theory when the case method of teaching was initially developed. Just as with Professor Bowen’s course, this worked well; it was the right method, given the circumstance. In the manner described by Schein (1987), a strong culture emerged at HBS that reasoned case-based instruction in an inductive architecture is the best way to teach.

In 1997 Christensen was assigned to head Har- vard’s required MBA course in general manage- ment. Guided by Harvard’s culture, he architected the course inductively and infused it with cases that he and like-minded colleagues had written that described the problems of innovation, strat- egy, and operations from the point of view of gen- eral management. Case by case, he attempted to lead his students inductively to discover well- documented frameworks and models that scholars already had discovered and published in aca- demic journals. The course was an unmitigated disaster. Students could tell that Christensen al- ready had the “answer.” In his attempts to orches- trate a sequence of inductive case discussions, it seemed to them that the professor was asking the students to guess what was on his mind. The stu- dents hated it, and as a consequence, the faculty hated teaching it.

In the midst of this misery Christensen realized that he was in a different situation than the ones in which Bowen and Harvard’s early masters of case instruction had found themselves. There already was a strong body of theory in the field. This sug- gested that a deductive course architecture, where the course design started at the top of the pyramid and cycled deductively downward could be more appropriate. Cases could be used to help the in- structor and students together test theories in hopes that anomalies would emerge. So Chris- tensen took the same material that had been used inductively and disastrously in the general man- agement course and deployed it in a course named building and sustaining a successful enterprise (BSSE), which had deductive architecture appro-

priate to the circumstance. The BSSE course has become quite successful with its deductive architecture.

For each class, students must read two docu- ments to prepare for the discussion. The first is a paper that summarizes a normative, circumstance- contingent theory about a dimension of a general manager’s job. This is the primary reading. The second is a case about a company facing a prob- lem that is relevant to that theory. In class discus- sions, students look through the lens of the theory to see if it adequately explains what historically happened in the company. They use the theory as the basis for recommending what management ac- tions will and will not lead to the desired out- comes, given the circumstance the company is in. And they search for phenomena in the case that the theory cannot explain.

For the next class, students are given a paper summarizing another theory that applies to a dif- ferent dimension of management and a case about a company whose predicament is relevant to that theory. Students first examine the case through the lens of the reading assigned for this class. Then they must pick up the theories that they used in prior classes, and put them on, like a set of lenses, to see what additional insight they can get about the cause of the company’s problems and what can be done to resolve them. The students do this again and again, so that by the end of the class they have a significant arsenal of theories that they can employ to understand a company’s prob- lems in all of its dimensions and can predict the results of alternative courses of managerial action. And they have a lot of practice using the theories, as the analysis and discussion of the cases simu- lates their real-world application.

As an example, for one class session BSSE stu- dents read a book chapter summarizing the model of disruptive innovation as the primary reading (Christensen, 2003). A case is also assigned: “Con- tinuous Casting Investments at USX Corporation” (Christensen, 1991). The case and reading describe how steel minimills such as Nucor entered at the bottom of the market, selling rebar, and then moved upward. USX retreated from the attack, market tier by market tier. Until he began teaching in this way, Christensen’s writings on the topic had asserted that the mechanism causing disrup- tion was that the leading companies had to re- spond to the needs of their best customers in order to survive. In the first year of leading a class dis- cussion around this pair of readings, however, one student said, “When you say that listening to your customers is the cause, I don’t think you’ve gone deep enough. I think the fundamental mechanism

248 JuneAcademy of Management Learning & Education

that causes the disruptors to move up-market is the pursuit of profit. And what causes the disruptees to flee up-market is the pursuit of profit.” She was right. The student taught the professor a profound insight about his own theory.

A couple of years later, Christensen asked his students why USX was trying to “cram” minimill technology into its existing mills, rather than building a greenfield minimill as Nucor had done. Christensen asked this question sincerely: He didn’t know the answer. He put a chair in the mid- dle of the room to represent the USX CEO and another representing Nucor’s, and asked, “What dimensions of the world look different from these points of view?” Again, the discussion produced an astonishing insight. Both CEOs had been taught in their finance and economics courses that they should ignore sunk costs and just consider the marginal cost and profits from alternative courses of action. By following this rule, the marginal cost of leveraging his existing mill infrastructure was much more profitable to the USX CEO than was the full cost of building a greenfield minimill. To Nu- cor, however, there was no existing mill infrastruc- ture to leverage at the margin, so building a new mill was judged to be most profitable. The class realized that the doctrine of ignoring sunk costs and focusing on marginal costs and profits was being taught in their finance and economics courses as a universally applicable principle— there was no circumstance-based categorization scheme in that theory. So the students and profes- sor together developed a categorization scheme for it: When the capabilities and infrastructure you’ve put into place to succeed in the past are those you will need in the future, then the marginal-cost doc- trine pertains. But if new capabilities and infra- structure are needed to thrive in the future, then the marginal-cost analyses that cause managers to leverage the old are inappropriate. This insight formed the basis for a highly regarded article that the BSSE faculty subsequently published (Chris- tensen, Kaufman, & Shih, 2008).

The required macroeconomics course at the Har- vard Business School, BGIE, also uses cases in a deductive architecture. Students look through the lenses of well-researched theories to examine case studies of different countries’ economic problems and achievements—to analyze why things hap- pened as they did and what policies will and will not lead to the desired outcomes. Even in an intro- ductory course, students can be taught what the theory is, and then learn how to use it through the deductive method of case learning.

DISCUSSION: COURSES AS THEORY-BUILDING ENTERPRISES

Note that both the courses described above enlist the students in building better theory and using it. They differ only in their starting points. Bowen’s BRSB course uses cases inductively to build a body of understanding in a field where little has been written. The BSSE course starts at the top of the pyramid with already-researched models as its starting point. In both courses, the cases at the foundation of the pyramid are deliberately multi- faceted, written to highlight multiple perspectives on a problem and outcomes for which the reigning theories might not account. Indeed, they feature problems that the professors themselves do not fully understand. As a result, students often dis- cover an anomaly in these complicated cases that faculty researchers have not seen. This enables the faculty and students to revisit the crispness of definitions, the categorization scheme, and the as- sociated statements of correlation or causality in the theory.

There is a third circumstance where a theory is so completely developed and tested that there is little potential improvement possible. Lecture- based course architectures might be appropriate to this circumstance. Rarely is this the case in man- agement education, however. We have found that even when well-accepted theories are employed in a deductive architecture, case discussions that en- roll students as research associates can surface puzzling anomalies. Questioning the sufficiency of the categorization schemes in the theories we study is almost always a productive path for build- ing better theory. “When and why might this not work?” and “Under what conditions might this gos- pel be bad news?” are simple questions that can yield breakthrough insights. In the process, the students can help the instructor refine categoriza- tion schemes to yield more accurately predictive circumstance-contingent statements of causality. The instructors often feel they learn more from the class discussions than the students do.

Entire courses need not be structured in the same way. Junior faculty members, for example, might design four to six class sessions for course research relating to their research interests, and populate the rest of the course with material pub- lished by others. Regardless of the source of mate- rial, however, professors can always learn from students if course research is their mind-set.

A prescriptive theory is a statement, based upon a body of understanding, of what causes what, and why. This means that consciously or uncon- sciously, our students are going to use theory every

2009 249Christensen and Carlile

time that they make plans and take action as man- agers. A deductively architected course helps them learn how to use theories that stem from the re- search and writing of others. Inductively archi- tected courses can help them be capable develop- ers of theory, because, like it or not, whenever we draw a lesson from our experience, we are build- ing a theory. Our hope is to prepare our students to be capable builders and discerning consumers of theory—to build an instinct for knowing when they can and cannot trust that the recommendations of an author, consultant, or subordinate will produce the needed impact, given the circumstance that they are in. The skill to learn the right lessons from experience and to apply them in circumstances where they are appropriate is perhaps the most valuable ability that our students might take from our tutelage.

SUMMARY

We have come to believe that instead of compart- mentalizing research and teaching into separate realms, it might be more productive to think of teachers and students as partners in a collective enterprise of building, improving and using man- agement theory. We hope that this paper can con- stitute the beginnings of a pyramid of its own: a prescriptive theory of course design. Case studies can comprise a solid foundation for theory-build- ing courses because they can be structured to dis- play the phenomena in all its complexity. Instead of a descriptive categorization of case-based ver- sus lecture courses, we propose that faculty find themselves in one of two circumstances. The first comprises situations where instructors need to de- velop theory. The second circumstance is when instructors hope to teach theory, improve it, and help students practice its use. We propose that it might be helpful to begin referring to our class- room activities in these two circumstances as course research.

At the peak of the pyramid that we propose is a statement of causality: What enables managers to be effective is the ability to use the appropriate the- ory or model, given the circumstance in which they find themselves, that will allow them accurately to predict the result of the actions they plan to take. They will be more effective managers if they are discerning consumers of theory—if they can tell an invalid and unreliable theory from a valid, reliable one, and know how the tools they employ to become successful must differ by circumstance.

For those steeped in the gospel of sample size, it will seem audacious that we propose a theory of course architecture that is extracted from a few case

studies, all of which are tainted by the subjective- ness of personal experience and observation. But as noted above, all data are subjective, and what makes a normative theory externally valid isn’t sam- ple size, but getting the categories right.

We aren’t finished, of course. We simply hope that these concepts can provide willing faculty members with a start. As they use these state- ments of cause and effect, we hope that most of them find that the statements of causality actually do lead to the creation of better theory and more productive learning by students and faculty. We can expect, however, that anomalies will arise, as researchers follow these methods and fail. When this happens, we invite them to publish a paper about the anomaly, focusing on what it was about the circumstance of their research that caused the causal mechanism we have described here not to work. It is only by doing so that we can approach a better understanding of what methods of research and teaching can most predictably yield the most insightful theories and the best-prepared students.

REFERENCES

Adner, R., & Zemsky, P. 2005. Disruptive technology and the emergence of competition. Rand Journal of Economics, 36(2): 229–254.

Barnard, C. I. 1939. The functions of the executive. Cambridge, MA: Harvard University Press.

Bazerman, M. H. 2005. Conducting influential research: The need for prescriptive implications. Academy of Manage- ment Review, 30(1): 25–31.

Bower, J. 1970. Managing the resource allocation process. Engle- wood Cliffs, NJ: Irwin.

Burgelman, R. 2002. Strategy is destiny. New York: The Free Press.

Burgelman, R., & Sayles, L. 1986. Inside corporate innovation. New York: The Free Press.

Campbell, D. T., & Stanley, J. C. 1963. Experimental and quasi- experimental design for research. Boston: Hougthon Mifflin Press.

Carlile, P. R., & Christensen, C. M. 2005. The cycle of theory building in management research. Harvard Business School Working Paper Series #05057.

Christensen, C. M. 1991. Continuous casting investments at USX corporation. Harvard Business School Case, #9-697-020.

Chistensen, C. M. 1992. The innovator’s challenge: Understand- ing the influence of market environment on processes of technology development in the rigid disk drive industry. Unpublished DBA Thesis, Harvard Business School.

Christensen, C. M. 1997. The innovator’s dilemma: When new technologies cause great firms to fail. Boston, MA: Harvard Business School Press.

Christensen, C. M. 2003. The innovator’s solution. Boston, MA: Harvard Business School Press.

Christensen, C. M., Kaufman, S., & Shih, W. 2008. Innovation

250 JuneAcademy of Management Learning & Education

killers: How financial tools destroy your capacity to do new things. Harvard Business Review, January: 98–105.

Clegg, B. 2003. The first scientist: A life of Roger Bacon. New York, NY: Carroll & Graf Publishers.

Daneels, E. 2006. Dialogue on the effects of disruptive technol- ogy on firms and industries. Journal of Product Innovation Management, 23: 2–4.

Eisenhardt, K. M. 1989. Building theories from case study re- search. The Academy of Management Review, 14(4): 532– 550.

Feyerabend, P. 1975. Against method: Outline of an anarchistic theory of knowledge. London, UK: Humanities Press.

Gilbert, C. G. 2001. A dilemma in response: Examining the newspaper industry’s response to the internet. Unpublished DBA Thesis, Harvard Business School.

Glaser, B., & Strauss, A. 1967. The discovery of grounded theory: Strategies of qualitative research. London, UK: Wiedenfeld and Nicholson.

Henderson, R. M., & Clark, K. B. 1990. Architectural innovation: The reconfiguration of existing systems and the failure of established firms. Administrative Science Quarterly, 35: 9–30.

Hutton, A., Miller, G., & Skinner, D. 2000. Effective voluntary disclosure. Harvard Business School working paper.

Johnson, H. T., & Kaplan, R. 1987. Relevance lost. Boston, MA: Harvard Business School Press.

Kaplan, R. 1986. The role for empirical research in management accounting. Accounting, Organizations and Society, 4: 429– 452.

Kuhn, T. 1962. The structure of scientific revolutions. Chicago, IL: University of Chicago Press.

Lawrence, P. R., & Lorsch, J. W. 1967. Organization and environ- ment. Boston, MA: Harvard Business School Press.

Poole, M., & Van de Ven, A. 1989. Using paradox to build man- agement and organization theories. Academy of Manage- ment Review, 14: 562–578.

Popper, K. 1959. The logic of scientific discovery. New York: Basic Books.

Porter, M. 1991. The competitive advantage of nations. New York: The Free Press.

Roethlisberger, F. 1977. The elusive phenomena. Boston, MA: Harvard Business School Press.

Schein, E. 1987. Organizational strategy and culture. San Fran- cisco: Jossey Bass.

Shugan, S. M. 2006. Save research—Abandon the case method of teaching. Marketing Science, 25: 109–115.

Spear, S. C. 2005. Fixing health care from the inside—Today. Harvard Business Review, 83(9): 78–91.

Spear, S. C., & Bowen, H. K. 1999. Decoding the DNA of the Toyota production system. Harvard Business Review, 77(5): 96–106.

Stinchcombe, A. L. 1968. Constructing social theories. New York: Harcourt, Brace & World.

Tushman, M. L., & Anderson, P. 1986. Technological discontinui- ties and organizational environments. Administrative Sci- ence Quarterly, 31: 439–465.

Van de Ven, A. 1989. Nothing quite so practical as a good theory. Academy of Management Review, 14(4): 486–489.

Van de Ven, A. 2000. Professional science for a professional school. In M. Beer, & N. Nohria (Eds.), Breaking the code of change. Boston, MA: Harvard Business School Press.

Weick, K. 1989. Theory construction as disciplined imagination. Academy of Management Review, 14: 516–532.

Yin, R. 1984. Case study research. Beverly Hills, CA: Sage Pub- lications.

Clayton M. Christensen ([email protected]) is the Robert and Jane Cizik Professor of Business Administration at the Harvard Business School, with a joint appointment in the Technol- ogy & Operations Management and General Management fac- ulty groups. Christensen holds a BA in economics from Brigham Young University and an MPhil in economics from Oxford Uni- versity, where he studied as a Rhodes Scholar. Christensen re-

ceived an MBA from the Harvard Business School in 1979. He was awarded a DBA from the Harvard Business School in 1992. His research and teaching interests center on the management of technological innovation, developing organizational capabil- ities, and finding new markets for new technologies.

Paul R. Carlile ([email protected]) is an associate professor of man- agement and information sys- tems at Boston University’s School of Management. Carlile holds a BA in philosophy and a master’s in organizational be- havior from Brigham Young Uni- versity. He received his PhD in organization behavior from the University of Michigan. The gen- eral focus of his research is on how work “practices” structure knowledge differently and the

consequences this has on effectively developing and valuing knowledge in cross-domain innovation. In particular Carlile examines the characteristics and design requirements of boundary processes between knowledge domains.

2009 251Christensen and Carlile