Conclusion

Can risk assessment be made fair? The conclusion of Calculating Race returns to actuarial science’s foundations in probability. The roots of probability rest in a pair of problems posed to Blaise Pascal and Pierre de Fermat in the summer of 1654: “the Dice Problem” and “the Division Problem.” From their very foundation, the mathematics of probability offered the potential not only to be used to gain an advantage (as in the case of the Dice Problem), but also to divide material fairly (as in the case of the Division Problem). As the United States and the world enter an age driven by Big Data, algorithms, artificial intelligence, and machine learning and characterized by an actuarialization of everything, we must remember that risk assessment need not be put to use for individual, corporate, or government advantage but, rather, that it has always been capable of guiding how to distribute risk equitably instead.

Life

Chapter 1 focuses on the early history of race-based insurance. When the Newark-based Prudential Insurance Company of America incorporated in 1875, it revolutionized the American insurance industry by offering policies to the working class for an affordable three cents per week. What made the Prudential doubly unique was that the company insured not simply industrial laborers, but also African American laborers. The company was not in the progressive vanguard, though. Rather, the Northern upstart, in contrast to its Southern competitors, simply had not thought to craft a company policy to explicitly ban African Americans from purchasing life insurance. Just five years after becoming the first insurer to cover black lives, the Prudential began to charge differential, race-based premiums and commenced a public relations effort to defend its discriminatory practices. This foundational chapter traces how the theoretical work of scientific racism became embedded in the business practices of American insurers.

Proxies

Calculating Race’s fourth chapter demonstrates that race has become so highly correlated with other social statistics that actuarial science in general has developed a baked-in racial bias. Racial discrimination by proxy (e.g., zip code standing in for race) can be glimpsed in the disparate impact of data-driven decision-making in housing, healthcare, policing, sentencing, and more. Simply leaving out racial data in statistically aided decision-making distances institutions from claims of intentional discrimination, but a disparate, discriminatory impact lingers when other factors correlated with race power actuarial analyses. Chapter 4 considers how insurance law in the United States has defined the limits of acceptable discrimination. By surveying the progression of state-by-state regulations that prohibit or accept the use of race, gender, sex, sexuality, ability, age, and genetics in an industry that revolves around the ability to discriminate risk, it uncovers who the United States has historically chosen to protect.

Crime

From the mid-to-late nineteenth century, in the period after the use of branding and before the use of fingerprinting, penal institutions faced the problem of how to identify repeat offenders. In this interim, Alphonse Bertillon, a clerk in the Paris Prefecture of Police, developed an anthropometric system that measured the bodies of criminals at their intake and catalogued these measurements in order to identify them should they offend again. Calculating Race’s second chapter traces the importation of the Bertillon System of Classification to the United States, where its data collection practices were racialized. It then investigates University of Chicago sociologist Ernest Burgess’s 1920s work on this data set to build a formula for sentencing and parole decisions. The resulting algorithm from Burgess’s work relied heavily on race-based Bertillon data and factored race into its recommendations for length of sentence and supervised release, installing racial statistics as a key variable in matters of criminal justice.

Introduction

The introduction of Calculating Race positions the process of racial formation in the context of the risk society. The relationship between race and risk stretches back to the transatlantic slave trade, an era in which enslaved Africans were insured during their voyage across the Middle Passage. Calculating Race begins with a striking account of an uprising on the ship Thames, preserved as a historical record of how the nascent insurance trade of the early modern period viewed enslaved persons as both at risk and a risk. By insuring enslaved persons against the risk of mortality and insuring the ships that carried them against insurrection, maritime slave insurance established a complex relationship between enslaved Africans and risk even before they came to the Americas. This introduction sets the stage for this relationship and previews the intersection of risk and race in the United States over the course of the nineteenth and twentieth centuries.

Home

The economic collapse that set into motion the Great Depression of the 1930s was portended by mass mortgage defaults in the mid-1920s. To address this unprecedented housing crisis, New Deal legislation created the Federal Housing Administration (FHA) to insure mortgage loans. Without predecessors or peers and faced with a national emergency, the FHA turned to risk-rating experts in real estate valuation to craft underwriting policies that would shape the geography of the country and cement racial segregation in the United States for generations to come. Chapter 3 details how FHA officials utilized risk-rating standards that disqualified people of color from obtaining federally subsidized mortgage insurance. This institutional discrimination had the deleterious effect of essentially precluding people of color from obtaining middle-class America’s most important wealth-generating asset: the single-family home. Though others have written about the agency’s policies before, my analysis is notably the first to locate each version of the FHA’s underwriting manual, to take stock of each facet of race-based risk rating until the conclusion of the practice in 1947, and to analyze the agency’s effect on the lending industry thereafter.