You are here

Statistics in Society: The Arithmetic of Politics

Daniel Dorling and Stephen Simpson, editors
Oxford University Press
Publication Date: 
Number of Pages: 
[Reviewed by
Gudmund R. Iversen
, on

This book consists of forty-seven essays arranged in the following eight parts:

  1. Collecting Statistics
  2. Models and Theory
  3. Classifying People
  4. Counting Poverty
  5. Valuing Health
  6. Assessing Education
  7. Measuring Employment
  8. Economics and Politics

The authors all work at British institutions teaching and doing research, which means the specific contents of most chapters are of more interest to British than American readers.

The book grew out of activities supported by what is known as the Radical Statistics group, a twenty-five year old group unknown to me before reading this book. Radical Statistics is "a group of statisticians and others who share a common concern about the political assumptions implicit in the process of compiling and using statistics, and an awareness of the actual and potential misuses of statistics and its techniques." This means there is special slant to the chapters, and this is what makes the book of great interest for US readers as well.

Most of us like to think that the mathematical theory of statistics is politically neutral, and that misuses of statistics come down to simple questions such as computing percentages correctly and using the median when we should not use the mean. We also think that new statistical models arise out of the needs researchers have to analyze their data, not motivated by any political concerns. This book will change our views on these matters.

We do not have to look very far in our own country before we see that statistics is not free from political influence. Among other things, the year 2000 will bring yet one more census, and much political discussion has focused on whether statisticians in the Bureau of the Census should be permitted to use statistical methods beyond the pure counting of people to arrive at a count of people. The count has direct influence on the number of members of the House of Representatives each state will have, and our two major political parties split on this issue.

Francis Galton, Karl Pearson and Ronald Fisher are three of the central names in the development of least squares methods around the turn of the last century. We all know the story of how statistical regression analysis was developed through the study of heights of parents and offspring. Indeed, the very name regression comes from Galton's finding that there was a movement towards the middle height from one generation to the next. What is not as well known is that much of this work was inspired by the study of intelligence, and the field of eugenics motivated all three statisticians. They felt that the class structure in Britain was a hierarchy of innate ability, and the lower classes should be kept from having large numbers of children. We do not tell our students about this part of the history of regression and correlation.

A major difficulty with statistics is that it is a subject used by many people. It is therefore often difficult to see where the work of the statistician ends and the work of the subject matter person begins. Faults attributed to the statistician are more often due to errors and wrong interpretations by the user of statistical methods. One area where this occurs is the overreliance on the null hypothesis and the magic of the 0.05 p-value. Statisticians know that the p-value is only a guideline, and confidence interval estimation gives a much better sense of what the data are telling us.

In summary, the book encourages us to see the role of statistics in a wider perspective, recognizing that statistics exists within a social framework, and the more we become aware of that, the better and more useful our field will be. The book ends with the following questions we can use as a guide the next time we are told "another study shows...": Is that true? How do I know it is true? Where did it come from? Who wanted me to know that? What are the alternative explanations? Could I have done better?

Gudmund R. Iversen is professor of statistics at Swarthmore College. He has been long interested in the use of statistics in the social sciences. He is the author of Statistics: The Conceptual Approach, with Mary Gergen, and has written four of the little green books published by Sage on different statistical topics. Bayesian statistics has long been an interest of his. He can be reached at

The table of contents is not available.