How Our Days Became Numbered: Risk and the Rise of the Statistical Individual by Dan Bouk (review)

This book is about the uses of “big data” in the rapidly expanding insurance industry a century ago in the United States. Appropriately, the author is a member of the working group Historicizing Big Data at the Max Planck Institute. The big data from a century ago is an insurance company’s collection of information about individual applicants that was used to decide who should receive a policy and at what price. The detailed information, which often included standardized medical and credit reports, was stored in summary form on individual file cards. This process, according to Bouk, gave rise to the “statistical individual”—a construct much in evidence today.¹

Bouk based his study on the archive records of the Equitable Life Assurance Society of America and the Mutual Life Insurance Company of New York, as well as the papers of individuals who were central to the big-data enterprise in the insurance industry—Louis Dublin, Irving Fisher, Frederick Hoffman, and Alfred Lotka. Bouk complements this material with an extensive list of secondary sources.

The effects of the analysis that insurance companies performed on the data were many and varied, and Bouk covers them well. On the negative side, the data collected contributed to discrimination against blacks since the statistics based on them showed blacks to be poor insurance risks. Also included in the class of poor insurance risks were large numbers of people in the southern states; decidedly fewer insurance policies were offered in that area of the country. On the positive side, these data benefited the movement to promote better health, which troubled those insurance companies that saw their purpose as solely selling insurance. Others felt that an activist approach would change the mortality patterns in the population, leading to longer lives and thus to increased premium income for, and deferred payouts by, the companies. Insurance-data analysis also stimulated new medical insights. Insurance underwriters traditionally rated thin people as bad risks, under the impression that thinness was a strong indicator of tuberculosis, until the new data showed otherwise; being overweight was a bigger risk factor with regard to mortality.

The collection and analysis of yesterday’s big data (which pales in comparison with today’s) by the insurance companies relied on such new technologies as the Hollerith card sorting machine, index cards, the typewriter, and even carbon paper that enabled paper copies. Not unlike the case today, the big data of the past did not necessarily drive the technology; the Hollerith machine was developed originally to process the 1890 census of the United States. Bouk, however, allots only a few sentences to the importance of technology; the focus of the book is on the effect of the data collected, not the technology behind it.

Readers versed in statistical methods will find Bouk’s treatment of the Armstrong Investigation of 1905 into insurance practices interesting. The actuary Emory McClintock came under fire for his use of arcane smoothing methods, especially in determining income related to the distribution of dividends to policyholders. The smoothing methods pioneered by actuaries like McClintock and heavily criticized in the investigation have been refined and developed over the past 100 years to become a key tool in a statistician’s toolbox. Such discussions make How Our Days Became Numbered well worth the read.