Rational Choice Theory And Large-Scale Data Analysis

Hans-peter Blossfeld, Gerald Prein

• Publisher: Routledge
• Year: 2019
• Print ISBN: 9780367300463
• eBook ISBN: 9781000308952
• Topic: Social Sciences
• Subtopic: Sociology
• Index: Social Sciences

The relationship between rational choice theory and large-scale data analysis has become an important issue for sociologists. Though rational choice theory is well established in both sociology and economics, its influence on quantitative empirical sociology has been surprisingly limited. This book examines why there is hardly a link between the t

Part 1
Introduction

1
The Relationship Between Rational Choice Theory and Large-Scale Data Analysis—Past Developments and Future Perspectives

Hans-Peter Blossfeld and Gerald Prein

Introduction

Since the early 1980s, rational choice theories¹ have rapidly grown in complexity and sophistication in the social sciences. Although these approaches aim to improve our understanding of real world phenomena, it is surprising how little impact they have had on empirical social research in general (see, e.g. Green and Shapiro 1994) and on large-scale data analysis in particular (Goldthorpe in this volume).

In this book, we try to achieve two main goals: first, to explore in detail the various reasons for this development in the field of quantitative macro-sociological analysis; and, second, to make suggestions of how this gap between rational choice theory and large-scale data analysis can be bridged in future.

Bringing together experts from rational choice theory and practitioners of large-scale data analysis, this volume contributes to a deeper understanding of why rational choice theory needs large-scale data analysis, and why quantitative analyses of social structure can be improved by using rational choice models.

The contributors to this book share the view that successful empirical studies of a macro-sociological problem cannot concentrate on structural or micro approaches alone. Instead, theoretically powerful empirical sociological analyses must pay attention to both structural- and micro-level issues.

However, such a macro-micro framework must recognize from the outset that time matters in this context. It must achieve two goals (Tilly 1984):

1. it must identify the particular historical structures and processes which dominate the changes occurring in a given population, i.e. the sociologically important dependent and independent variables;² and
2. it has to specify the causal mechanisms that allow us to trace the encounters of intentionally acting individuals with the flow of history as a series of choice processes.

In this respect, the invaluable function of time is to offer a continuously changing point of reference by distinguishing, conceptionally and empirically, between a closed (but always changing) past, the present, and an intrinsically open future. Good sociological theories should therefore not only allow us to explain (or understand) a given outcome at present with reference to the closed past, but also help us to predict outcomes in an uncertain future with reference to the known past and present conditions. The crucial empirical test of sociological theories is not the extent to which they help us to explain a given outcome ex post facto, but the degree to which they also yield successful predictions of individuals' actions and their outcomes ex ante.³

Over the last 15 years, modern longitudinal social research has made great methodological progress in assessing causal inferences based on sociological theories.⁴ Life course and panel studies have collected timerelated data over substantial blocks of space and historical time. In methodological terms, the most important but still fairly unrecognized advancement that has been made is that longitudinal data can be studied by new statistical techniques in a stepwise time-related fashion. They allow us to follow up a great number of individuals belonging to different cohorts over longer spans of time and to differentiate at each point in time between a closed past, the present, and an intrinsically open future (Blossfeld and Rohwer 1995). Our more specific intention in this book is therefore to demonstrate how causal relations suggested by a dynamic combination of macro- and micro-theories can be represented in longitudinal statistical models and assessed with temporal data.

Why Rational Choice Theory Has Not Been Very Successful in Large-Scale Data Analysis? A Review of the Past

Let us first explore in this introduction some of the reasons for the obvious fact that the influence of rational choice theory on quantitative macro-sociology has been surprisingly small (see also Goldthorpe in this volume). Our perspective is that in the process of establishing rational choice theory as a "new" theoretical paradigm in sociology, its proponents have tended to pigeonhole their rivals, to caricature competing theories, to exaggerate existing theoretical cleavages, to overlook their own conceptional weaknesses, to downplay the difficulties of their empirical applications, and to neglect more recent and actually quite successful theoretically driven research programs based on longitudinal data. In this introduction, we will address these points step by step and then give an overview of the chapters that follow in this book.

Of course, there is not only one single rational choice theory but many different versions of it that share an umbrella label, but often little else. There has also been an important shift from earlier crude models with fairly unrealistic behavioral assumptions towards more recent theories elaborating the role of heterogeneous preferences, the effects of uncertainty, the impact of structural constraints, the relationship between norms and rational choice, and the possibilities of the non-existence of a rational choice. No doubt, this development has made rational choice approaches increasingly attractive for empirical practitioners. However, the result of the aspiration to establish rational choice theory as the only coherent and unified theoretical approach in sociology has been the creation of a scientific camp mentality and, unfortunately, the almost complete disregard of rational choice approaches in large-scale data analysis. Using a concrete example of our recent research on consensual unions, we want to demonstrate this problem in more detail.

Attacking Outdated and Empirically Unimportant Theories

Rational choice theorists have frequently been polemical against specific sociological conceptions which rarely had any but certainly almost have completely lost their influence on empirical studies today. For example, proponents of rational choice theory enjoy criticizing the homo sociologicus,⁵ extreme normative sociological conceptions⁶, crude functionalism, or more elaborated models of structural-functionalism⁷. But as far as we can see, all of these theories, though still discussed in theoretical sociological seminars, have turned out to be fairly irrelevant in contemporary empirical research.⁸ Thus, even if the theoretical critique of rational choice proponents is justified, it could not be very consequential for the conduct of concrete empirical social research today.

We want to demonstrate this point with a concrete research example studying the question of why people living in consensual unions marry if the woman becomes pregnant (Blossfeld et al. 1995a). In an extreme version of the normative model, one would have to assert the dominance of normative constraints and simply deny the importance of a choice at all. Thus, with the occurrence of a pregnancy in a consensual union, norms would have the effect of cutting down the feasible set of actions to a single point, i.e. marriage. Today's cohabiting men and women would therefore be portrayed as mindlessly repeating or imitating what their ancestors did in the past in similar situations (Elster 1989b). But as a serious theory of marriage action this normative model is obviously too wrong to merit any empirical analysis. Hence, it is not surprising that, to the best of our knowledge, there is no empirical researcher who has shaped his/her analysis of this problem on the basis of such a strong normative conception.

Rejecting an extreme normative approach in empirical studies however does not mean that norms (and culture) do not matter. Rather it is our view that norms are extremely important. Social interactions are intrinsically symbolic relationships that have a meaning and can therefore not be understood without reference to cultural settings.⁹ Yet in their attempt to push for a duality between an intentionally acting person guided by instrumental rationality and a passive executor of social norms (or culture), rational choice theorists often lost sight of the fact that norms (culture) and rational choices are important in empirical applications.¹⁰ The important theoretical issue for empirical analysis is therefore not whether social norms (culture) or instrumental rationality provide the motivation for actions¹¹, but how they can be conceptionally integrated so that we are better able to understand real life situations.¹²

For example, in the research example mentioned above, values, social norms, and traditions will certainly have an important impact upon people's marriage behavior in the case of a pregnancy, but in most cases this influence is likely to be mediated through the intentional actions of individuals. It is therefore plausible that there is a changing, frequency-dependent¹³ coexistence of norm-guided behavior and rational, self-centered behavior with regard to the decision to marry when a pregnancy occurs in a consensual union (cf. Elster 1989b). In such a hybrid model of empirical application, social norms do not dictate the marriage behavior, but allow a deliberate, reflective imitation of traditions. This is because social norms normally offer considerable scope for interpretation and manipulation. Social norms' main function might be to focus and coordinate expectations (Elster 1989b). Thus, when the norm "If the woman gets pregnant, then marry" is still shared within a community, people will to a certain degree expect each other to do that. In particular Elster (1989b) has stressed that the coordinating function of norms is mainly due to the strong emotions that their violations can trigger in the violators themselves and also in other people. Thus, the social norm "If the woman gets pregnant, then marry" is likely to be sustained by the feelings of embarrassment, anxiety, guilt and shame that a person suffers at the prospect of violating it. Of course, such emotions may or may not help unmarried couples to reach a marriage decision when the woman is pregnant, and they certainly will rarely be the only motivation for such an important and long-term decision. But this concrete research example illustrates that our understanding of the relationship between norms and intentional action is only in its rudimentary beginnings and that rational choice theory simply must provide better analytical concepts in this regard, if it is to serve as a powerful instrument in empirical applications. This book tries to achieve progress in this direction.

Aspirations to Provide Universal ("Time-Less") Theories

Some rational choice theorists have been adopting a rigid version of the modern philosophy of science (Green and Shapiro 1994). They have been striving for general causal theories and universal laws (e.g. Hechter 1987; Esser 1996b) and have been thinking little about "historicistic analyses" and "inductive generalizations" (e.g. Kiser and Hechter 1991). Regardless of how one values the merits of historicism and induction in sociology¹⁴, by no (logical) means are sociologists forced to accept only a choice between statements about universal laws or statements about accidental, contingent historical relationships. Rather, sociologists can legitimately try to establish causal mechanisms of limited generality tailored to a specific range of historical situations (e.g. Gambetta 1987; Elster 1989c; Hedström and Swedberg 1995 and in this volume; Erikson and Ragin in this volume). This perspective will be developed in more detail in the following chapters.

Yet the strong aspiration to formulate general theories and universal laws of social action may be one of the reasons why many sociologists consider rational choice theory to be empirically unappealing. In particular, the mainstream version of this approach "explains" people's rational actions with regard to universal and stable general human preferences and considers constraints as exogenously given. However, if preferences are universal and stable among individuals (Stigler and Becker 1977; Becker 1981; Lindenberg 1985a, 1990a), they must turn out to be simply irrelevant for explaining differences in peoples' behavior (Gambetta 1987).¹⁵ In this model, individuals' particular "tastes" or their marginal utilities, for specific courses of actions can therefore only be derived from the constraints (or oppor...

Table of contents

• Cover
• Half Title
• Title
• Copyright
• Contents
• About the Editors
• About the Contributers
• Preface
• PART ONE Introduction
• PART TWO Linking Rational Choice Theory and Quantitative Sociology
• PART THREE Large-Scale Data Sets, Rare Events, and Qualitative Research
• PART FOUR Empirical Analyses Using Rational Choice Theory
• PART FIVE Rational Choice Theory and Longitudinal Research
• PART SIX Future Perspectives
• References
• Subject Index
• Name Index