How to Lie with Statistics

A Synopsis of How to Lie with Statistics by Darrell Huff When most people hear or read a statistic, they quickly have to decide if the numbers listed are valid or invalid. It is usually assumed that the author of the statistic is knowledgeable in the field to which the statistic pertains. However, on many occasions, the statistic is false, due to the author’s wording. Darrell Huff’s novel How to Lie with Statistics is a manual that can help individuals catch these lies. The novel allows readers to solve marketing ploys and dismiss certain statistics as faulty.
The first chapter focuses on bias. The book states that all statistics are based on samples, and these samples have bias. This means that no matter what the reader will have a biased opinion. This bias is spawned from the respondents replying dishonesty, the author choosing a sample that gives better results, and the availability of data. Huff uses a survey of readership of two magazines, which had refuting results. This is because, due to the readers’ personal biases, they answered the survey dishonestly.
This example closes the chapter, teaching readers to always assume that the sample has a bias. The second chapter focuses on averages. It states that there are actually three types of averages: mean, median, and mode. Mean is the arithmetic average. Median is the name given to the midpoint of the date. Finally, mode is the data point that occurs the most often in the data. Thus, the type of average used can alter the results of the statistics. The next chapter explains how sample data is chosen to prove certain results.

Many marketing campaigns use this technique. They choose sample sizes that give their wanted results. Huff’s solution is that one must determine if the information is a discrete quantity or if a range is involved. The following chapter discusses errors in measurement. It explains two measures for measuring error: Probable Error and Standard Error. The probable error uses the error in the measuring device used to measure the error in the measurement. The standard error is the standard deviation of the sampling distribution of a statistic.
Chapter five explains how to manipulate a graph in order to show the results one wants. For example, if one was using a line graph to show a 10% climb, they could remove the unused parts of the graph to make their results seem more extreme. The next chapter discusses how two-dimensional images can deceive readers. A picture may be increased in all dimensions, making it seem much larger than it really is, giving the impression of a greater increase. Chapter seven focuses on the semi-attached figure.
Simply put, if one cannot prove what they wish to prove, they can merely prove something else and then give the impression that these two things are the same. Huff uses cold medicine as an example. A pharmacist wants the reader to believe that their medicine cures the cold, but instead the label reads that the medicine kills 32, 132 cold germs. The pharmacist then hopes that the reader makes the assumption that because the medicine kills such a high number of germs, then it cures the cold. Huff is therefore teaching not to assume.
The following chapter focuses on cause and effect. Huff stresses that readers must ask for when certain data was collected and if the amount of data was adequate for the entirety of the experiment. Chapter nine teachers readers how to ‘statisticulate’, meaning how to manipulate readers by using statistics. This chapter is essentially a list of what to look for when determining the validity of a statistic. Huff explains various tricks, such as measuring profit on a cost price and how income calculations mislead by using children of a family as the average.
The final chapter instructs readers on how to talk back to a statistic. Huff emphasizes that readers must ask who the author his and how did he come to collect the knowledge listed in the statistic. Also, he encourages readers to question if someone changed the subject of the statistic. Finally, he explains that one must be able to understand the data presented, and if it does not make sense, then it is most likely untrue. Overall, Huff assists readers in how to determine if a statistic is valid or invalid. Though the book was published over fifty years ago, these methods are still in use.