Tata-Dhan Academy: PDM


Programme in Development Management

A Pinch of Politics with A Lot of Statistics: A Review of “Data Analysis for Politics and Policy”

By Edward Tufte | © 1974, Prentice Hall College Division | ISBN: 0131975250 | 179 pages | Available online at http://bit.ly/cNgsx7

Forecasting is an important word in financial markets, disaster management, and many other areas that involve extensive decision-making. The book under review, titled Data Analysis for Politics and Policy and authored by political scientist and statistician Edward R. Tufte from Yale University, is unique in its kind. It drives a reader to think of politics from a statistical perspective. To write this book, Tufte has taken advantage of his experience as a political science teacher, his education in statistics, and his [then budding] expertise in the field of visual presentation.

The book has expressed statistical techniques with respect to politics and policy-making. The statistical procedures, as explained with diagrammatic explanations, are easy to understand. The explanation process starts with a simple theoretical explanation, and follows with the formation of statistical equations and graphical presentation; finally, the most important and remarkable, the explanation of the whole mechanism ends with a real life example. Throughout the book, the writer has targeted the politicians who know little about or do not know about statistical methods useful for politics and policy. To establish logic, the writer starts with a definition and then substantiates it with common examples, sometimes in the form of questions. To explain dependent and independent variables, for example, he asks “Does smoking cause lung cancer?” Clearly, lung cancer is the dependent or explained variable and smoking is independent or explanatory variable.

This book is ideally interesting, but operationally and practically challenged. It is interesting to read, easy to understand, and excellent to remember through real life examples. It is easy to internalise a stochastic variable with an expected mean of zero, which is graphically represented in the book as vertical drop from the population regression function. But can we determine the value of a stochastic variable for the next elected representative from “Z” constituency, for instance? It is difficult to determine by how much the average vote will differ for an old candidate. Ideally, we can come out with a figure with the help of historical data, but this is still not very practical as in a political perspective, getting a vote is not dependent on a single factor like economic condition of the voter, age of the voter, or the voter’s marital status alone. Rather, it depends on a mix of factors including communication, racial diversity, and geographical homogeneity. A few explanatory variables in a regression model showing the effect on a dependent variable are not sufficient to predict politics. A single vote in politics is a function of multiple factors. An appreciable attempt has been made in this book to explain this through multiple regressions; however, even a multiple regression model is subject to many disturbances within the model which dislocate the equation. These concerns are not addressed in the book in detail.

As a reader goes through the book, they may strongly feel that it reaches to politics but not to politicians and it reaches to policies but not to policy makers. One may like this book for its harmonised flow of thoughts to interpret politics and policies matters statistically; however, this book in itself seems insufficient to guide the one who is intended to use it in a practical manner. Going for multiple regression and dummy variable regression model is like walking on a razor’s edge: a little deviation will make a great difference. Plausible errors are always much more potential than the best possible outcomes in this respect. A little interdependence of any two or more variables will lead to a biased result. An omission of a single cross section of a data may lead to an invalidation of the whole process. Thus, the complexity defeats the purpose of the book because the politicians and policy makers (and perhaps their advisors) may not have the prerequisite expertise to comprehend these problem areas. To make the book qualified for its audience, the writer should have simplified it with more examples and stories. However, simplification and further explanation may also require more equations and more use of statistical techniques that will multiply the complexity. Unfortunately, I feel the ultimate objective of the book to provide some basic material for drawing political and policy inferences is not fulfilled.

Despite the shortcomings, the book establishes the role that data analysis and statistical approaches play in forecasting. The objective to improve the quality of statistical thought has been satisfactory throughout the book. However, at times, it seems to be a collection of data from a purposive sample to show success stories. This is not the best way to deal with issues of politics and policy changes that are much more sensitive in nature, issues where the immediate and direct impact falls upon the people. Still, it is an attempt to blend two different disciplines for a single purpose. Setting aside its shortcomings in operational feasibility, the graphical representations and explanations through live examples are appreciable. Briefly, it may be noted that Data Analysis for Politics and Policy is a unique book, very clear in demonstrating statistical techniques. It is most advisable for classroom theoretical study rather than for day-to-day politics and policy matters.

Dushasana Mahanta, PDM 10


  • Gujarati D.N. (2006). Basic Econometrics. New York: Tata McGraw-Hill.

Filed under: PDM 10, Spectrum, , ,

3 Responses

  1. shanti211087 says:

    Hello sir,
    you have beautifully presented the crux of the book. your article is a motivation tool for me. after reading your article I have decided to read this book. your elaboration of statistics is quite easy for me. thanks for this beautiful article.

    • Ananda Mahto says:

      I find that this is quite a good book introducing statistics and data analysis. I like how each problem is sequentially walked-through by the author.

      Mahanta might be right about some of the limitations, but I think that even though it is more than 30 years old, it is still definitely worth reading today.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: