Delinquency Research

An Appraisal of Analytic Methods

Travis Hirschi,Hanan C. Selvin

• Editorial: Routledge
• Año: 2017
• ISBN:
  9781351288941
• Edición: 1
• Categoría: Sozialwissenschaften
• Categoría: Sozialarbeit

This remarkable guide to delinquency studies was co-winner of the 1968 C. Wright Mills Award for the best book in the field of social problems. The work is in effect three books in one: a forthright account of how to analyze survey data, a penetrating critique of delinquency research, and a set of original essays on methodology. It is a landmark work that continues to serve as an essential tool for those who both study and want to learn about deviance. In the new introduction, Travis Hirschi describes the setting in which 'Delinquency Research' was written, noting that it exudes a confident optimism that well-conducted research and analysis will quickly lead to important advances in the field. Hirschi maintains that twenty-eight years after 'Delinquency Research' was first published the validity of its optimistic view has been confirmed by the fact that the field of criminology is among the leading producers of high quality research. As a result, we know more about crime and delinquency than ever before. 'Delinquency Research' forms the basis for present and future studies of criminology and is a necessary addition to the libraries of sociologists, criminologists, scholars in the area of delinquency, and students interested in research methods.

Part i The Nature of Methodological Criticism

On Approaching Methodology

“Methodology” is a scare-word. Among the subjects studied during the education of a sociologist, only statistics arouses more apprehension in advance and more dismay during the course. We think these feelings are unfortunate and unnecessary. They are unfortunate because they keep students from understanding and enjoying what should be a central part of their education; they are unnecessary because they rest on a misunderstanding of the nature of methodology, at least in modern usage. In this brief introductory chapter we want to explain what methodology means to us and to offer some advice on understanding the methodological discussions in this book and on applying their lessons to other sociological reading.

In the fields of logic, education, and statistics, methodology has special, technical meanings. All of its sociological uses, however, conform to the dictionary definition of the “science of method.” The significance of this definition obviously depends on the root word method. There are at least four distinct meanings of method in sociology. When Emile Durkheim wrote The Rules of Sociological Method in 1895, he was thinking of the content of sociological theory, of the explanations of social phenomena that could legitimately be called sociological. This usage has virtually disappeared; sociologists dealing with these problems today are more likely to use the term meta-theory. The most common meaning of method today is a set of recipes for the everyday conduct of social research: how to gain rapport with a respondent in an interview, how to construct codes for classifying responses, how to percentage a table. Related to this is the use of method to mean statistical techniques, especially in university catalogs of ten or twenty years ago.

To us, these techniques, strategies, and rules of thumb are the tools of the trade, not its intellectual core. This core lies in the relation between data and theory, the ways in which sociologists use empirical observations to formulate, test, and refine statements about the social world. Or, in the words of Lazarsfeld and Rosenberg:

The term methodology . . . implies that concrete studies are being scrutinized as to the procedures they use, the underlying assumptions they make, the modes of explanation they consider satisfactory.¹

The task of the methodologist, then, is to explain what the investigator is doing and to determine whether or not these procedures will lead the investigator to the kinds of statements he wants to make. Implicit in this definition is a set of principles or criteria by which the methodologist can judge the adequacy of the procedures. The following chapters include many such principles and criteria—for example, criteria for judging if an observed relation between poor housing and delinquency can reasonably be taken as causal. Two chapters examine the criteria that other methodological critics have used in assessing delinquency research: Chapter 2 considers several broad attacks on the entire field of delinquency, and Chapter 8 takes up other criteria of causality implicit in the attacks of some critics on previous causal statements in delinquency research.

The methodological critic is thus subjected to the same kind of criticism that he applies to empirical investigators. These criteria and principles play a key part in assessing the work of empirical investigators and methodologists alike. Good criteria lead to useful criticism and better research. Bad criteria, such as those discussed in Chapter 8, result in pointless criticism and inhibit needed research. How, then, is one to know which criteria are good, acceptable, or valid, and which are bad, unacceptable, or invalid?

Methodological Criteria

Some criteria of adequacy rest on logical or mathematical reasoning. For instance, it is easy to show mathematically why a low level of random error of measurement is desirable in studies of the relations between variables: The greater the random error, the more the observed correlations are reduced (“attenuated”) below what they would otherwise have been. For problems like this, the methodologist needs only a basic knowledge of mathematics and statistics.² Sometimes not even that is necessary. In Chapter 8 we show, without any formal mathematics or statistics, that these “bad” criteria of causality have devastating implications: To accept any one of them is tantamount to denying the possibility of any causal relations.

It is rare to be able to justify one’s criteria in this way. Indeed, the validity of most of the criteria in this book is not demonstrable; for example, the three criteria of causality in Chapter 3 cannot be validated by any kind of logical derivation. They rest, instead, on the consensus of sociologists (and others), a consensus that has grown stronger over the years as these criteria have survived many different critical attacks. In the end, even mathematical derivations rest on such consensus, for all derivations depend eventually on undemon-strable assumptions that people have agreed to accept.

Since all notions of validity rest on consensus among the workers in a field of study, what is acceptable at one time may not be acceptable at another. For over fifty years sociologists were content to infer statements about the difference in suicide rates between Protestants and Catholics from the suicide rates in countries with varying proportions of Protestants and Catholics. Only after the publication of Robinson’s classic paper on ecological correlations in 1950 did sociologists come to realize that even the great Durkheim had been guilty of invalid reasoning.³

Methodology, then, is not everlasting, revealed truth. It is a living body of ideas that changes with time. There are many methods, and what is right and valid today may be wrong and invalid tomorrow. It would be nihilistic to use this mutability of methodological criteria as an argument for doing away with all methodological judgments. At any given time one has to decide by the standards of that time. True, standards may change so as to redeem a study now considered unacceptable, but this has happened to few studies in the past. Change in the other direction, from acceptable to unacceptable, is the more usual course. This is a fact of life in all empirical sciences: Sooner or later, every generalization perishes. An investigator can only hope that his generalization will live beyond its infancy.

Even when investigators agree on the rules they should follow or the criteria they should meet, they often find it difficult or impossible to do so. For one thing, the criteria may conflict with each other. If one wants to study the relations between child-rearing practices and delinquency, is it better to ask many questions of a small sample or fewer questions of a larger sample? With a limited amount of money, one cannot have both a large sample and many questions.

Sometimes the criteria are so vague as to leave room for a wide range of judgments. In an early draft of Chapter 5 we criticized one investigator for not taking account of a certain variable. In an exchange of correspondence with him, it became clear that there were fundamental differences between him and us on whether or not this variable was really important.

Another reason why a competent investigator may not conform to some criteria of methodological adequacy is that it may be difficult or impossible to meet the conditions for applying the criteria. For example, one criterion of causality requires that the alleged cause occur before the effect. Thus it is possible to ask if inadequate supervision of the child is a cause of delinquency only if the supervision was inadequate before the delinquency began. In some cases, however, the causal order is unclear. Does delinquency cause poor work in school, or does poor work in school cause delinquency? Chapter 4 will treat questions of causal order in more detail. For now, however, it should be clear that methodological rules and criteria are guides to action rather than laws of nature and that it is difficult for even the most competent investigator to conform to all of the criteria he considers valid. All of these points should convey the same message to the student and the methodologist: Before condemning an investigator for violating some criterion or assumption, try to see if he could have avoided the violation and whether or not this might have led him into greater difficulties.

How to Read This Book

A methodological riddle: When it comes to reading the report of an empirical study, what is the difference between a layman, a researcher, and a methodologist? Answer: The layman reads the text and skips the tables; the researcher reads the tables and skips the text; and the methodologist does not care very much about either the tables or the text, as long as they agree with each other.

Labored as this riddle may be, it is all too accurate. The layman, however, is not always at fault for skipping the tables; as some of our examples show, it is often difficult to reconcile the tables and the text. Because most of the examples in this book are based on tables, and because part of our task is the methodological analysis of the relations between tables and text, it is essential to understand why tables are difficult to read and what can be done about it. Part of the answer lies in the large amount of information that each number stands for. One cannot skim a table as one skims passages of ordinary text. A single percentage usually stands for an entire sentence; to explain and discuss the figures in a complex table may require several pages. As with any concentrated food, it takes time to digest the figures.

Our first suggestion is to read the tables slowly. This is especially important for the reader who has not had some systematic experience in reading tables. The discussion of tables in Chapter 3 begins simply, so that even the reader who has skipped tables in the past will be able to read the tables in this book (and others as well) .⁴

Problems of understanding occur in this book at two levels. There is first the investigator’s account of his work. Here, of course, our discussions should be helpful, because we usually begin the analysis of a complicated example by restating the investigator’s arguments. Now the reader of this book is also a methodologist, and he should be as much concerned as we are with the extent of agreement between the text and the material analyzed. That is to say, our own discussions are also fair game for the kind of methodological critique that we present here. Because our analysis is necessarily more abstract than the material on which it is based, the reader will do well to proceed even more slowly in reading what we have to say about an example than in reading the example itself.

There is also the opposite danger: that some of the examples (and our analyses of them) will appear too simple. Some investigators are able to express their thoughts so lucidly that the reader finds himself going through the analysis almost as if it were a novel. Perhaps a few of our methodological critiques will read this well also. The danger here lies in misleading the reader about how easy it is to do empirical research or methodology. The difficulties he is sure to find when he tries to do these things himself may prove discouraging. For the record, then, we want to say that any such appearances are misleading. Only in textbooks does a research project move smoothly from hypothesis to conclusion. In real life, research is far less orderly; as in a work of art, what appears natural and simple is usually the product of hard labor.⁵ The same consideration applies to our methodological critiques. Some of the shortest and simplest took many hours of hard thought and, occasionally, argument between us. The reader should not be discouraged when he finds that research and methodological analysis are hard work for him too.

A useful technique in learning how to read and construct tables is to work back from the table to the raw (unpercentaged) figures of the original data, as we do several times in the text (at greatest length in Chapter 14), and then combine these figures to yield simpler tables. From the three-variable table of Chapter 14, showing the relations between delinquency, adequacy of supervision, and mother’s employment, one can make three two-variable tables—delinquency and supervision, delinquency and mother’s employment, and supervision and mother’s employment. Constructing, percentaging, and interpreting these three tables will give an insight into three-variable relations that can be had in no other way.

Our final suggestion is to take one’s hunches and intuitions seriously. In our own experience, we often began our analysis of an example with the vague feeling that something was wrong, but we were unable at the outset to say just what led to this feeling. As we studied the data from different perspectives, more often than not we found our hunches leading to a useful analysis. One cannot, of course, rely on hunches and such subjective feelings to demonstrate the validity of a methodological argument, but nothing in the canons of methodology prohibits one from taking advantage of hunches to suggest both problems and solutions.

The Methodologist and the Investigator

Literary critics, historians, and even some sociologists picture social research as a mechanical process of turning people into numbers.⁶ As the abstract study of empirical research procedures, methodology is presumably even more guilty of this dehumanization. This naïve view is fostered by the conception of science as a mechanical application of the scientific method. Our discussion of the principles and criteria for research decisions suggests, however, that there is no single scientific method; there are many scientific methods, and the consensus of investigators in each field determines which ones are acceptable at a particular time. In no field that we know of is empirical research a cut-and-dried routine. For social research we can even provide some evidence that there are artistic and subjective components.

In the spring of 1963 an experiment on the analysis of data was conducted at the Survey Research Center of the University of California, Berkeley. The participants were three experienced investigators: Charles Y. Glock, William L. Nicholls, II, and Martin A. Trow. As it happened, all three had done their graduate study at the same university (Columbia), all had worked in close proximity for several years, and all had taught the same course in methodology with one of us (Selvin). If...

Índice

• Cover Page
• Title Page
• Copyright
• Dedication
• Contents
• Introduction to the Transaction Edition
• Preface to the 1973 Edition
• Preface to the Original Edition
• Acknowledgments
• Part i The Nature of Methodological Criticism
• On Approaching Methodology
• 2 Critiques of Delinquency Research
• 3 Principles of Causal Analysis
• 4 Causal Order
• 5 Genuine and Spurious Relations
• 6 Links in the Causal Chain
• 7 Interaction of Variables
• 8 False Criteria of Causality
• Part iii Multivariate Analysis
• 9 Some Problems of Multivariate Statistical Analysis
• 10 The Shortcomings of Tabular Analysis
• 11 Concepts, Indicators, and Indices
• 12 Reliability and Scaling
• 13 Statistical Inference
• 14 Description and Prediction
• 15 Individual and Group Variables
• A Final Word
• Index