Othmar Winkler

Interpreting Economic and Social Data: A Foundation of Descriptive Statistics (2009)

Hardcover: 265 pages Publisher: Springer-Verlag Berlin Heidelberg (2009)

Back Cover

“Interpreting Economic and Social Data aims at rehabilitating the descriptive function of socio-economic statistics, bridging the gap between today's statistical theory on one hand, and econometric and mathematical models of society on the other. It does this by offering a deeper understanding of data and methods with surprising insights, the result of the author's six decades of teaching, consulting and involvement in statistical surveys. The author challenges many preconceptions about aggregation, time series, index numbers, frequency distributions, regression analysis and probability, nudging statistical theory in a different direction.

Interpreting Economic and Social Data also links statistics with other quantitative fields like accounting and geography. it is aimed at students and professors in business, economics and social science courses, and in general at users of socio-economic data, requiring only an acquaintance with elementary statistical theory.”

First section:

1.1 Stating the Problem

“Statisticians accept as a self evident principle that there is one general theory of statistics that applies equally to all fields,' biology, economics, engineering, demography, environmental sciences, sociology, etc. (Fig. 1.1). Yet, important applications in economics and the social sciences in general are not covered by what today is considered 'the theory of statistics.' This calls for a review of the situation, of methods that do not apply, and important aspects of socio-economic applications that are not supported by statistical theory. The peculiar nature of the data in socio-economic statistics requires a different basis than is available at present and makes it unlikely that a general `Theory of Statistics' can satisfy the needs of this scientific field. Historically, the turn toward inference came from the discovery of random sampling, from experimentation in agriculture and other applications in the natural sciences. We proceed as if socioeconomic statistical data are like those in the sciences, ignoring that they differ in important ways. Because of this, the applications of social, business and economic statistics are not adequately supported by today's statistical theory.”

Preface: Introduction

“On a snowy winter morning I boarded a crowded city bus, unable to use my bicycle for the usual 10 km commute to Georgetown University. In preparation for teaching that morning, I began perusing the textbook on business and economic statistics that I had adopted for this course. At the next stop, a young woman took the seat next to me, the only one remaining in the full bus. Shortly after settling in, she turned to me:: `Excuse me sir, is this statistics?' she motioned to the textbook. `Yes', I responded, surprised, `Business and Economic Statistics.' At this, a look of revulsion overcame her “Ugh... Statistics was the only subject I could never handle in college...” She trembled at a memory that still haunted and upset her. At this reaction a series of similar, though less dramatic, occurrences came to mind. Few other academic subject seem to evoke the distaste that the mention of statistics seems to elicit. Does it have to be that way? I grappled with possible explanations for a long time.

This book is my response that evolved gradually over decades of teaching a variety of business, economic and general statistics courses, using the newest textbooks available, and being involved in survey work and statistical consulting. I wondered why these textbooks on business and economic statistics presented the subject matter as a watered-down version of mathematical statistics, which itself evolved from problems of measurement and observation in the natural sciences. These textbooks treat socio-economic data like the measurements in the natural sciences and present the subject as an application of probability, grouped around the Gaussian, Poisson, F, C² and other statistical distributions, sampling and statistical inference. There was no interest or concern about how to interpret the messages about society contained in the wealth of published economic and social data. They fail to see that this is the raison d'etre of the entire statistical enterprise which also should be the main purpose of statistics courses for social scientists. These courses fail to present statistics as the instrument for scanning the economic and social environment and to monitor important aspects of social reality.

It is the aim of this book to re-orient statistics towards making sense of economic and social data. It is an attempt to rehabilitate 'descriptive statistics' as a respectable part of statistics, re-orienting it toward the description of society which in fact was its original purpose and still is the ultimate goal of all statistical endeavors. This book is addressed to the literate and numerate public, trying to open their eyes to various basic facts that are commonly overlooked, in short, to lead them to a fuller awareness of simple basics, to encourage asking questions and to look for answers in the fine print that accompanies tabulations of socio-economic data.

It is also the aim of this book to draw attention to the neglected twilight zone, the no-man's land between the partisan efforts of statisticians who, inspired by applications in the natural sciences, turned to probability, controlled experiments, model building, etc. on one hand, and the applied fields of social sciences on the other. Statistical theorists feel that they are the guardians of a true science, concerned with the purity of its theoretical core with little regard for interpreting economic and social data. On the other side are social scientists and economists concerned about discovering timeless laws of economics, intent on condensing them into mathematical models. They too are less concerned about using statistical data to monitor and also influence events in society. And last, but not least, there are the dedicated statistical foot soldiers who take censuses and surveys, and prepare tabulations. They too have no time for making sense of their data about society.

As you may notice from the ‘Outline,’ this book departs from the usual structure, but instead follows the steps of the statistical process in a rather abstract, theoretical manner, from the very start of conceptualizing the socio-economic phenomenon to be investigated to the final tabulation of the data. Standard topics, like the Gaussian curve, probability theory and symmetrical, well-behaved frequency distributions are treated at the end of the book, if at all. The initial chapters deal at length with topics that are usually missing in textbooks such as aggregation, statistical aggregates and ratios. They form the backbone for the interpretation of socio-economic data. Then follow three chapters on time series as the most frequent form in which data are published. These chapters are given priority over frequency distributions in one or more dimensions, treated toward the end.

This book, by the way, is not meant as an introduction to statistics, nor as a “how-to-do, hands-on” manual. Its concern is to make sense of socio-economic data. and to shine new light on various misconceptions the reader may have acquired in previous statistics courses. Only a minimum of mathematics will be required. Calculations are relegated to the five optional appendices. Although mathematical statisticians may find this book pedestrian and simplistic, some abstract thinking is involved and the reader is asked to be patient with unfamiliar ideas.

This book is intended for everyone who has to deal with data about society: students and teachers in business, economics and social science courses, economists, social scientists, financial analysts, market researchers, business and economic forecasters, sociologists, managers, demographers, even geographers. It is my hope that the chapters of this book will open up a new understanding of socio-economic data for their meaningful interpretation, allay bad feelings toward our field, and stimulate further developments in the indicated direction.

Description of Chapters:

Chapter 1 provides a short view of the developments that led to the present situation in socio-economic statistics. The powerful influence and the band wagon effect of the developments in statistics in biology, agriculture came to dominate all fields of statistical application. This chapter points out that socio-economic statistical data are quite different from the measurements in the sciences.

Chapter 2 traces the statistical process, from the conception and formulation of a socio-economic phenomenon, such as unemployment, poverty, productivity or crime; to the identification and recording of the relevant `real-life-objects' which portray that social or economic phenomenon: human beings, entities such as corporations, or events, such as births, work accidents or business mergers. The simplified records of these `real-life-objects' then become the 'statistical-counting units'.

In Chap. 3 the subsequent grouping of these `statistical-counting-units' into suitable aggregates is discussed. These new entities, the statistical aggregates, are defined by their three `dimensions': the subject matter, the time period, and the extent of the geographic area covered. As to the subject-matter “dimension,” the qualitative characteristics of the statistical counting units are important for the formation of a hierarchy of sub-aggregates. The magnitude of each of the three `dimensions' of an aggregate determines how to interpret the gains and losses from aggregating the `statistical-counting-units'. These statistical aggregates represent the bulk of the data in socio-economic statistics. They are quite distinct from the data in the natural sciences, an important matter that has not received due attention.

In Chap. 4 a variety of ratios is discussed as simple and effective analytical tools. These ratios allow us to perceive and make sense of the underlying economic and social reality conveyed by these aggregates. Despite their pervasiveness and importance, ratios have rarely been discussed.

Chapters 5, 6 and 7 study the development, over time, of economic and social phenomena through time-series of socio-economic data.

Chapter 5 presents a critical view of the customary decomposition of time series into trend, seasonal pattern, business cycle and randomness. Instead of the mathematical decomposition into the standard components, time-series should be understood as quantitative economic and social history that can be interpreted meaningfully through a hierarchy of simple ratios between aggregates. These figures are not to be understood as abstract algebraic numbers.

Chapter 6 explores the fact that statistical data lose their relevance over time and become obsolete and less relevant for anticipating the future of a situation in society. Good forecasting requires acquaintance with the historic development of the underlying economic or social forces. Much depends on the speed with which the data become obsolete. The level of aggregation also affects obsolescence. All this requires judicious decisions regarding the weight older data should be given in a forecasting model, and the point in the past from which on the data of a time series should be disregarded.

Chapter 7 has two parts. In the first part, Sect. 7.1, Price-Index-Numbers are discussed as an important type of time series. A simpler, ratio-based approach is presented that is more transparent and easier to interpret than the historic Price-Index-Number formulations currently in use, allowing for understanding and interpreting the actual changes in price levels. In the second part, Sect. 7.2, Index-Numbers of Production are critically reviewed. Different production concepts are discussed and simpler ways of measuring production and productivity are developed.

Chapter 8 deals with the interpretation of highly asymmetric frequency distributions that predominate in economic and social data. Simple measures are presented to deal appropriately with these highly asymmetrical data, to assess and interpret centrality, asymmetry and dispersion.

Chapter 9 discusses the puzzling case of one particular regression analysis that changed my views on cross-sectional data in general. Without going into the algebra of their calculation, specific problems in Regression and Correlation with aggregate data are discussed.

Chapter 10 explores the relationship between statistics and the calculus of probability. Although socio-economic statistics is numeric, using mathematical symbols, algebra, geometry and graphs, it must not be considered as a branch of mathematics. Socio-economic statistical data have an important conceptual non-numeric component that defies a numbers-only approach. One must keep in mind that its purpose is the perception of very real economic and social happenings in historic time, and in geographic and subject-matter space. Misuses of probability, foremost the mis-interpretation and misuse of “Statistical Significance,” are critically reviewed.

Chapters 11 and 12, explore areas that social, business, and economic statistics has in common with subjects that do not readily come to mind as linked with statistics. While exploring these areas in these two final chapters the nature of socioeconomic statistics is further clarified.

Chapter 11 has more in common than is usually acknowledged. When statistics is not considered as a branch of mathematics, however, it is easier to see that macro economics, really National Accounting — which is essentially economic statistics - keeps track of the economy like financial accounting keeps track of a business corporation. The discussion reveals surprising affinities between socio-economic statistics and financial accounting.

Chapter 12 discusses the importance of geographic-spatial distributions, a matter that has been absent from the theory of statistics, though not from statistical field-work. Although specialized quantitative-statistical research abounds in geography, the geographic-spatial dimension has not been recognized as belonging to statistics and ought to be included in its theory.

Reviews and Comments:

Review by Prof. Thomas R. Dyckman: “Opening this book by Othmar Winkler is like splashing oneself with cold water at 5:30 in the morning. It’s a wakeup call! The author lays out his “call to arms” in the preface. In our quest to understand or “make sense of socio-economic data” (p. vi), we have come to rely too heavily on statistical inference (F, t, Chisquare) and on assumed symmetry and continuity. If we seek insights, we are enjoined to adopt instead the descriptive tools of statistics and apply them to aggregate observations and their categorizations. To understand socio-economic phenomena, it is essential to recognize that the contributing processes are purposeful and not random.” Reviewed in “The Accounting Review” by Thomas R. Dyckman: Prof. emeritus in Accounting at Cornell University.

Review by Prof. Dr. Peter Winker: “Othmar W. Winkler’s book is ... far from providing the usual type of content of a monograph in statistics, but rather challenges conventions by providing alternative views on the nature of data and how to analyze them.” “Instead of being just another textbook in statistics for economists, the book rather targets all experienced statisticians and econometricians who find it relevant to think carefully about the origin and properties of the data they use.” Posted in the journal “Jahrbuecher fuer Nationaloekonomie und Statistik.”

Review by Brady T. West: “This insightful text comes from a veteran scholar and targets individuals working with socio-economic (SE) data from business, economics, sociology, and other social sciences. The book provides a refreshing view on how data from these fields should be approached in a realistic manner. The author strongly and consistently advocates the use of straight-forward descriptive statistical methods to describe SE realities, rather than forcing SE data to conform to convenient probability models and inferential methods developed by mathematical statisticians for data from the natural sciences (which are much more likely to be governed by natural laws and “true” probability models). Interpreting Economic and Social Data will appeal to (and should be read carefully by) students and professional researchers in the social sciences who are responsible for analyzing cross-sectional or longitudinal SE data and generating written reports describing and interpreting the analysis findings.” “This book is an enjoyable read, and it does an excellent job of reinforcing the unique features of SE data and how statistical analyses should be tailored to these features to produce the most meaningful descriptions of SE phenomena.” Michigan Program in Survey Methodology Survey Research Center Institute for Social Research (ISR) University of Michigan. [Extract from a draft of his review to be published by the Journal of Official Statistics]

Review by Walter Krämer: “Most of what is covered in this book is either taken for granted or not discussed at all in standard textbooks on economic or social statistics, not to mention mathematical statistics.” Institut für Wirtschafts-und Sozialstatistik, Technische Universität Dortmund at Dortmund, Germany.

Review by Thomas Luke Spreen: “Othmar Winkler's Interpreting Economic and Social Data calls into question the tendency of social scientists to treat quantitative summary data as objective measurements as in the natural sciences. Winkler's observations on the subject are both thought-provoking and insightful.” Spreen is an economist, Division of Labor Force Statistics, Office of Employment and Unemployment, U.S. Bureau of Labor Statistics. Published in the Book Review Section of the August 2011 issue of the BLS Monthly Labor Review.

Comments by Dr. Keith Ord: “I am very impressed by the breadth of coverage and the deep discussion you have provided on a number of topics.

Review by McGee (Biostatistician, SMU) in the American Statistician.
Winkler's reply.

Review by Andrey Kostenko in the 2012 International Journal of Forecasting. “In conclusion, the book is written to share the author’s belief that ‘social and economic statistics, though numeric, is essentially quantified history of society, not a branch of mathematics’ (p. 232). Those who are close to this belief (or those who are yet to form their views on the subject) may find the book interesting.”
Winkler's reply.