理组 Management and 研威 究管 Organization Review Management and Organization Review 10:3,November 2014,381-389 doi10.1111/more.12068 理組 研織 究管 Reflections on Choosing the Appropriate Level of Abstraction in Social Science Research Johann Peter Murmann UNSW Australia Business School,Australia ABSTRAcT Although researchers often do it subconsciously,every explanation involves choosing a level of abstraction at which the argument proceeds.The dominant North American style of research in Organization Theory,Strategy,and International Business encourages researchers to frame their explanations at the highest level of abstraction where country-level contextual factors are suppressed or ignored.Yet to provide powerful explanations for recent developments in China,researchers are drawn to a greater level of context specificity.This tension is evident in the Child and Marinova (2014)paper.One way to reduce the tension is to identify general causal mechanisms that combine in different ways to produce different results depending on context.This research strategy is more effective than seeking invariant,general patterns of development across all times and places. KEYwoRDs choosing level of abstraction,philosophy of social science,research design, research on industries,research on firms 关于在社会科学研究中选择适当的抽象水平的反思 摘要 尽管研究者都是潜意识而为之，他们在论述解释问题时都进行了不同程度的抽象。 组织理论、战略和国际贸易中主流的北美研究范式，都鼓励研究者最大程度地抽象 他们的论述，与此同时却压制或忽略了国家层面的情境因素的影响。但是为了充分 地解释中国近期的发展，研究者都执着于高水平的情境特性。这种矛盾体现在Cil d和Marinova(2014)的文章中。减少这种矛盾的一种方法是发现一些普遍的因果 机制，采用不同的方式结合这些机制，从而根据情境产生不同的结果。这种研究策 略比寻求不随时间和地点变化的、普遍的发展模式更为有效。 关键词：选择抽象的水平，社会科学的哲学，研究设计，行业研究，企业研究 INTRODUCTION The beauty of science is that it takes what are seemingly different objects,phe- nomena,or events and reduces them to something similar.The more general 2014 The International Association for Chinese Management Research
Reflections on Choosing the Appropriate Level of Abstraction in Social Science Research Johann Peter Murmann UNSW Australia Business School, Australia ABSTRACT Although researchers often do it subconsciously, every explanation involves choosing a level of abstraction at which the argument proceeds. The dominant North American style of research in Organization Theory, Strategy, and International Business encourages researchers to frame their explanations at the highest level of abstraction where country-level contextual factors are suppressed or ignored. Yet to provide powerful explanations for recent developments in China, researchers are drawn to a greater level of context specificity. This tension is evident in the Child and Marinova (2014) paper. One way to reduce the tension is to identify general causal mechanisms that combine in different ways to produce different results depending on context. This research strategy is more effective than seeking invariant, general patterns of development across all times and places. KEYWORDS choosing level of abstraction, philosophy of social science, research design, research on industries, research on firms 关于在社会科学研究中选择适当的抽象水平的反思 摘要 尽管研究者都是潜意识而为之，他们在论述解释问题时都进行了不同程度的抽象。 组织理论、战略和国际贸易中主流的北美研究范式，都鼓励研究者最大程度地抽象 他们的论述，与此同时却压制或忽略了国家层面的情境因素的影响。但是为了充分 地解释中国近期的发展，研究者都执着于高水平的情境特性。这种矛盾体现在Chil d和Marinova（2014）的文章中。减少这种矛盾的一种方法是发现一些普遍的因果 机制，采用不同的方式结合这些机制，从而根据情境产生不同的结果。这种研究策 略比寻求不随时间和地点变化的、普遍的发展模式更为有效。 关键词：选择抽象的水平，社会科学的哲学，研究设计，行业研究，企业研究 INTRODUCTION The beauty of science is that it takes what are seemingly different objects, phenomena, or events and reduces them to something similar. The more general bs_bs_banner Management and Organization Review 10:3, November 2014, 381–389 doi: 10.1111/more.12068 © 2014 The International Association for Chinese Management Research
382 J.P.Murmann and simple scientific statements are,the more they dazzle the human mind. Consider Newton's second law of motion stating that the force (F)on an object is a function of its mass(m)and its acceleration(a),leading to the simple formula F=m*a.It is formulated to apply to all places and all times.Inspired by this idea of science,management scholars-just as other social scientists-are drawn to formulate abstract theoretical statements that apply universally at all places and all times.As a paradigm,this universalist ideal pressures scholars to seek generalizability often at the expense of insightful knowledge.In sociology,Talcott Parsons is a prominent example of the quest to formulate a general theory that tried to encompass all social action(Parsons,1937).Many sociologists from the 1940s onward started to realize,however,that when theories in sociology are formulated at the most abstract level possible,rather than becoming more pow- erful,they often lose explanatory power.For this reason,Robert Merton (1949: 42)advocated creating'middle range theories'that lie 'beteen the nomothetic and the idiothetic,betieen the general and the altogether particular,betwveen generalizing sociological theory and historicism. My first step in this essay is to argue that the 'middle range'invoked by Merton is a very wide range,requiring scholars to make a decision whether their explanations will lie closer to the idiothetic (the singular)or nomothetic (the all- encompassing,general)end.Building on the work of Stinchcombe (1978)and Tilly (2008),I next argue that good social theory with explanatory power is developed by paying close attention to the details of contexts.This means that fruitful theorizing is initially much closer to the idiographic end of the spectrum. Having set the stage,I will then comment on the stimulating paper by Child and Marinova (2014)in light of the preceding methodological discussion.In my reflection on Child and Marinova (2014),I propose that China-based scholars and scholars working on China need to carefully select the appropriate level of abstraction.The chief aim of my essay is to persuade scholars who research China that even if the ultimate goal is to find general concepts, we develop better theoretical knowledge when our theoretical efforts are deeply informed by context,and when our concepts have proven to work well in a specific context.Stinchcombe (1978:21),reflecting on decades of work in sociology,makes this point well:Far from it being the case that the most powerful general theorists ignore details,the precise opposite is true. Social theory without attention to detail is wind;the classes it invents are vacuous,and nothing interesting follows from the fact that *A*and *B*belong to the class'. Before I begin,let me make one disclaimer.This essay is written based on the experience of a macro scholar'mainly concerned with the development of firms,institutions,and countries.I leave to our more micro colleagues to decide whether my arguments are also valid at the level of individual human behaviour. C2014 The International Association for Chinese Management Research
and simple scientific statements are, the more they dazzle the human mind. Consider Newton’s second law of motion stating that the force (F) on an object is a function of its mass (m) and its acceleration (a), leading to the simple formula F = m*a. It is formulated to apply to all places and all times. Inspired by this idea of science, management scholars – just as other social scientists – are drawn to formulate abstract theoretical statements that apply universally at all places and all times. As a paradigm, this universalist ideal pressures scholars to seek generalizability often at the expense of insightful knowledge. In sociology, Talcott Parsons is a prominent example of the quest to formulate a general theory that tried to encompass all social action (Parsons, 1937). Many sociologists from the 1940s onward started to realize, however, that when theories in sociology are formulated at the most abstract level possible, rather than becoming more powerful, they often lose explanatory power. For this reason, Robert Merton (1949: 42) advocated creating ‘middle range theories’ that lie ‘between the nomothetic and the idiothetic, between the general and the altogether particular, between generalizing sociological theory and historicism’. My first step in this essay is to argue that the ‘middle range’ invoked by Merton is a very wide range, requiring scholars to make a decision whether their explanations will lie closer to the idiothetic (the singular) or nomothetic (the allencompassing, general) end. Building on the work of Stinchcombe (1978) and Tilly (2008), I next argue that good social theory with explanatory power is developed by paying close attention to the details of contexts. This means that fruitful theorizing is initially much closer to the idiographic end of the spectrum. Having set the stage, I will then comment on the stimulating paper by Child and Marinova (2014) in light of the preceding methodological discussion. In my reflection on Child and Marinova (2014), I propose that China-based scholars and scholars working on China need to carefully select the appropriate level of abstraction. The chief aim of my essay is to persuade scholars who research China that even if the ultimate goal is to find general concepts, we develop better theoretical knowledge when our theoretical efforts are deeply informed by context, and when our concepts have proven to work well in a specific context. Stinchcombe (1978: 21), reflecting on decades of work in sociology, makes this point well: ‘Far from it being the case that the most powerful general theorists ignore details, the precise opposite is true. Social theory without attention to detail is wind; the classes it invents are vacuous, and nothing interesting follows from the fact that *A* and *B* belong to the class’. Before I begin, let me make one disclaimer. This essay is written based on the experience of a ‘macro scholar’ mainly concerned with the development of firms, institutions, and countries. I leave to our more micro colleagues to decide whether my arguments are also valid at the level of individual human behaviour. 382 J. P. Murmann © 2014 The International Association for Chinese Management Research
Choosing the Appropriate Level of Abstraction 383 CHOOSING THE APPROPRIATE LEVEL OF ABSTRACTION Even if we already agree with Merton that as social scientists we should be mainly constructing theories of the middle range,in every piece of research it is still necessary to make a decision about the precise level of generality at which we want to construct an explanation.The middle range between the nomothetic and the idiothetic end of the spectrum is wide.Let me offer an illustration of this point.Let's say we want to develop a theory of firm behaviour.Let us assume in this thought experiment that our universe is made up of 100 firms (FI to F100)and five countries(CI to C5).Each country has the same number of firms,which means that each country has 20 firms.If our theory is formulated at the most general level of abstraction,it will apply to all 100 firms.At the other extreme,if the behaviour of each firm is completely distinct from the behaviour of any other firm,that is, each firm is entirely idiosyncratic,we would need 100 distinct explanations for the behaviour of our 100 firms.A widely shared definition of science sees it as an enterprise trying to abstract from the particular and generalize.A completely general statement would apply to 100 firms,a completely idiothetic statement would apply to only one firm and thus not amount to 'science'on the aforemen- tioned definition.But notice that even if we could only find commonality among two of the 100 firms,we would have still engaged in abstraction and generalization. Developing a generalization across two out of 100 firms is to engage in a scientific process.The key point I want to make here is that scientific statements(or theories if you will)in this particular thought experiment could apply to as few as two firms and as many as 100 firms. Scholars comparing two or more countries have repeatedly discovered that national differences in economic development,institutions,and culture have a strong influence on firm behaviour (Hall Soskice,2001).This means a powerful theory explaining the behaviour of 20 firms in one country may not apply well to the other 80 firms in the other four countries.As mentioned before,science does not require one to construct theories that apply to all 100 firms.Too often scholars fall into the trap of thinking that one has to formulate a general theory at the outset, which often requires one to ignore the most interesting causal factors that drive the behaviour of firms in a particular country.The result is that little understanding is gained.If one realizes,however,that science even resides in developing generali- zations for a subset of firms,one will see clearly the fruitfulness of first developing generalizations that apply to a smaller set of firms (e.g.,to 20 firms from one country)before examining whether the generalization holds for a larger set of firms or even the entire universe of firms. To summarize,any researcher in this thought experiment is forced to decide whether the theoretical statement will apply to two or three or 100 firms.Because of the universalist ideal in much of macro management research,scholars are typically inclined to formulate their theoretical statements in the most general way, 2014 The International Association for Chinese Management Research
CHOOSING THE APPROPRIATE LEVEL OF ABSTRACTION Even if we already agree with Merton that as social scientists we should be mainly constructing theories of the middle range, in every piece of research it is still necessary to make a decision about the precise level of generality at which we want to construct an explanation. The middle range between the nomothetic and the idiothetic end of the spectrum is wide. Let me offer an illustration of this point. Let’s say we want to develop a theory of firm behaviour. Let us assume in this thought experiment that our universe is made up of 100 firms (F1 to F100) and five countries (C1 to C5). Each country has the same number of firms, which means that each country has 20 firms. If our theory is formulated at the most general level of abstraction, it will apply to all 100 firms. At the other extreme, if the behaviour of each firm is completely distinct from the behaviour of any other firm, that is, each firm is entirely idiosyncratic, we would need 100 distinct explanations for the behaviour of our 100 firms. A widely shared definition of science sees it as an enterprise trying to abstract from the particular and generalize. A completely general statement would apply to 100 firms, a completely idiothetic statement would apply to only one firm and thus not amount to ‘science’ on the aforementioned definition. But notice that even if we could only find commonality among two of the 100 firms, we would have still engaged in abstraction and generalization. Developing a generalization across two out of 100 firms is to engage in a scientific process. The key point I want to make here is that scientific statements (or theories if you will) in this particular thought experiment could apply to as few as two firms and as many as 100 firms. Scholars comparing two or more countries have repeatedly discovered that national differences in economic development, institutions, and culture have a strong influence on firm behaviour (Hall & Soskice, 2001). This means a powerful theory explaining the behaviour of 20 firms in one country may not apply well to the other 80 firms in the other four countries. As mentioned before, science does not require one to construct theories that apply to all 100 firms. Too often scholars fall into the trap of thinking that one has to formulate a general theory at the outset, which often requires one to ignore the most interesting causal factors that drive the behaviour of firms in a particular country. The result is that little understanding is gained. If one realizes, however, that science even resides in developing generalizations for a subset of firms, one will see clearly the fruitfulness of first developing generalizations that apply to a smaller set of firms (e.g., to 20 firms from one country) before examining whether the generalization holds for a larger set of firms or even the entire universe of firms. To summarize, any researcher in this thought experiment is forced to decide whether the theoretical statement will apply to two or three or 100 firms. Because of the universalist ideal in much of macro management research, scholars are typically inclined to formulate their theoretical statements in the most general way, Choosing the Appropriate Level of Abstraction 383 © 2014 The International Association for Chinese Management Research
384 J.P.Murmann implying that they would apply to the entire universe of firms(100 in our thought experiment)even if they only studied a small sample of firms in just one country (I have also been guilty of this!).While theoretical statements that are formulated to apply to all firms at all times have the appeal that they seem to be in the same league as achievements in the physical sciences made by people like Newton,they typically sacrifice precision and often do not increase our understanding.In short, the power to explain and illuminate firm behaviour may reside in choosing a level of generality that is much smaller than the entire universe of firms.Given the differences between countries,one may have a much more powerful and robust theoretical statement if one initially restricts the scope of the scientific claim to apply to the context from which the cases are drawn.Let me turn to the work of Stinchcombe(1978)to articulate the methodological reason why this is so. Stinchcombe's(1978)key argument is that powerful theoretical concepts are not developed by initially examining a great number of cases (i.e.,large sample quantitative studies)but by examining a few cases in detail,trying to build deep analogies between these cases.Stinchcombe(1978)explains that deep analogies are constructed by establishing that a great many statements true of case A are also true of case B.These deep analogies amount to a generalization across the cases, giving rise to general concepts that are not tied to a specific case.The more these analogous statements are of an important causal character,the more scientifically important the resulting concepts are.This leads Stinchcombe to a conclusion that may be surprising: But if conceptual profundity depends on the deep building of analogies from one case to another,we are likely to find good theory in exactly the opposite place from where we have been taught to expect it.For it is likely to be those scholars who attempt to give a causal interpretation of a particular case who will be led to penetrate deeper analogies between cases.(21-22) Stinchcombe suggests that the typical objections to case studies,namely that they are not representative,is misplaced when it comes to the development rather than the testing of theories that are already well-advanced.Theoretical advancement comes from building up generalizations case by case,ensuring that the generaliza- tions hold across many details of the cases and further our understanding. REFLECTIONS ON CHILD AND MARINOVA This concept of how powerful theoretical statements are developed has profound implications for research on the development and behaviour of firms and industries in China.Instead of assuming that theoretical statements about firms developed for other countries and/or for the different stages of development also hold in China,it is more productive to build up theoretical statements based on detailed 2014 The International Association for Chinese Management Research
implying that they would apply to the entire universe of firms (100 in our thought experiment) even if they only studied a small sample of firms in just one country (I have also been guilty of this!). While theoretical statements that are formulated to apply to all firms at all times have the appeal that they seem to be in the same league as achievements in the physical sciences made by people like Newton, they typically sacrifice precision and often do not increase our understanding. In short, the power to explain and illuminate firm behaviour may reside in choosing a level of generality that is much smaller than the entire universe of firms. Given the differences between countries, one may have a much more powerful and robust theoretical statement if one initially restricts the scope of the scientific claim to apply to the context from which the cases are drawn. Let me turn to the work of Stinchcombe (1978) to articulate the methodological reason why this is so. Stinchcombe’s (1978) key argument is that powerful theoretical concepts are not developed by initially examining a great number of cases (i.e., large sample quantitative studies) but by examining a few cases in detail, trying to build deep analogies between these cases. Stinchcombe (1978) explains that deep analogies are constructed by establishing that a great many statements true of case A are also true of case B. These deep analogies amount to a generalization across the cases, giving rise to general concepts that are not tied to a specific case. The more these analogous statements are of an important causal character, the more scientifically important the resulting concepts are. This leads Stinchcombe to a conclusion that may be surprising: But if conceptual profundity depends on the deep building of analogies from one case to another, we are likely to find good theory in exactly the opposite place from where we have been taught to expect it. For it is likely to be those scholars who attempt to give a causal interpretation of a particular case who will be led to penetrate deeper analogies between cases. (21–22) Stinchcombe suggests that the typical objections to case studies, namely that they are not representative, is misplaced when it comes to the development rather than the testing of theories that are already well-advanced. Theoretical advancement comes from building up generalizations case by case, ensuring that the generalizations hold across many details of the cases and further our understanding. REFLECTIONS ON CHILD AND MARINOVA This concept of how powerful theoretical statements are developed has profound implications for research on the development and behaviour of firms and industries in China. Instead of assuming that theoretical statements about firms developed for other countries and/or for the different stages of development also hold in China, it is more productive to build up theoretical statements based on detailed 384 J. P. Murmann © 2014 The International Association for Chinese Management Research
Choosing the Appropriate Level of Abstraction 385 studies of Chinese firms.At a later stage one can examine whether these theoretical statements can be generalized across different countries.Reading the paper by Child and Marinova(2014),I think that the authors on one level agree with this view of how effective theory is developed.Yet on another level,they seem to suggest that good theories should be formulated at the outset at the highest level of generality,ensuring that they apply everywhere and at all times. Let me first offer evidence that Child and Marinova(2014)agree with the idea that powerful theory comes from understanding details of a contextual setting. With the title of their paper,The Role of Contextual Combinations in the Globalization of Chinese Firms,the authors seem to announce that the purpose of their paper is to construct a theory that can explain the globabilization of Chinese firms and not firms in general.As one reads the paper,it becomes clear that the empirical data from which the conclusions are drawn are exclusively Chinese.Initially,the authors are careful to emphasize that their framework is tailor made to explain patterns of the outward foreign direct investment(OFDI)of Chinese firms that cannot be explained by the existing theories of foreign direct investment(FDD). They announce in their opening pages: It is our aim to propose a framework that develops this analysis and also to indicate how it can enhance our understanding of the advantages and disadvantages attending Chinese OFDI,as well as the contingent adjustments that foreign investing Chinese firms may have to make.(Child Marinova,2014:349) In the next sentence,they speculate that their framework generalizes to contexts that are similar to China: We believe that such a framework could have wider relevance,especially for OFDI from other emerging economies characterized by strong government intervention.(Child Marinova,2014:349) But on the next page they make a move so common in the management literature,suggesting that the framework applies to all countries and at all times. While we now develop the argument with specific reference to China,we also suggest that it can be applied more generally to the implementation of OFDI from any country.(Child Marinova,2014:350) This statement seems to lend support to the notion that only when a framework explaining the performance of OFDI can in principle be applied to all firms in all countries(all 100 firms in our earlier thought experiment)have we engaged in good scientific work.This reading gains even more support by the authors'graphical respresentation of their key argument in Figure 1,which is clearly not limited to China or similar countries,but suggests that it applies to any country in the world. 2014 The International Association for Chinese Management Research
studies of Chinese firms. At a later stage one can examine whether these theoretical statements can be generalized across different countries. Reading the paper by Child and Marinova (2014), I think that the authors on one level agree with this view of how effective theory is developed. Yet on another level, they seem to suggest that good theories should be formulated at the outset at the highest level of generality, ensuring that they apply everywhere and at all times. Let me first offer evidence that Child and Marinova (2014) agree with the idea that powerful theory comes from understanding details of a contextual setting. With the title of their paper, The Role of Contextual Combinations in the Globalization of Chinese Firms, the authors seem to announce that the purpose of their paper is to construct a theory that can explain the globabilization of Chinese firms and not firms in general. As one reads the paper, it becomes clear that the empirical data from which the conclusions are drawn are exclusively Chinese. Initially, the authors are careful to emphasize that their framework is tailor made to explain patterns of the outward foreign direct investment (OFDI) of Chinese firms that cannot be explained by the existing theories of foreign direct investment (FDI). They announce in their opening pages: It is our aim to propose a framework that develops this analysis and also to indicate how it can enhance our understanding of the advantages and disadvantages attending Chinese OFDI, as well as the contingent adjustments that foreign investing Chinese firms may have to make. (Child & Marinova, 2014: 349) In the next sentence, they speculate that their framework generalizes to contexts that are similar to China: We believe that such a framework could have wider relevance, especially for OFDI from other emerging economies characterized by strong government intervention. (Child & Marinova, 2014: 349) But on the next page they make a move so common in the management literature, suggesting that the framework applies to all countries and at all times. While we now develop the argument with specific reference to China, we also suggest that it can be applied more generally to the implementation of OFDI from any country. (Child & Marinova, 2014: 350) This statement seems to lend support to the notion that only when a framework explaining the performance of OFDI can in principle be applied to all firms in all countries (all 100 firms in our earlier thought experiment) have we engaged in good scientific work. This reading gains even more support by the authors’ graphical respresentation of their key argument in Figure 1, which is clearly not limited to China or similar countries, but suggests that it applies to any country in the world. Choosing the Appropriate Level of Abstraction 385 © 2014 The International Association for Chinese Management Research