Insurance and Progress
To understand computer-age progressivism, we must begin with insurance.
Insurance?
In 1936, Liberty magazine published an eight-part series by James M. Cain called Double Indemnity.[1] Based on the 1927 murder of Albert Snyder by his wife and her lover, the mystery revolves around an unlikely detective.
Although the death is ruled a suicide, a sharp-eyed insurance adjuster determines it must be a murder.
"You were raised in private schools, Groton, and Harvard. While you were learning how to pull bow oars there, I was studying these tables. Take a look at them. Here's suicide by race, by color, by occupation, by sex, by locality, by seasons of the year, by time of day when committed. Here's suicide by method of accomplishment. Here's method of accomplishment subdivided by poisons, by firearms, by gas, by drowning, by leaps. Here's suicide by poisons subdivided by sex, by race, by age, by time of day. Here's suicide by poisons subdivided by cyanide, by mercury, by strychnine, by thirty-eight other poisons, sixteen of them no longer procurable at prescription pharmacies. And here—here, Mr. Norton—are leaps subdivided by leaps from high places, under wheels of moving trains, under wheels of trucks, under the feet of horses, from steamboats. But there's not one case here out of all these millions of cases of a leap from the rear end of a moving train. That's just one way they don't do it."[2]
Insurance as a Data Problem
In the 1870s, insurance companies paid local politicians, doctors, and others to tell them about the people buying life insurance. Were they good people? Were they trustworthy?
By the 1930s, insurance companies relied on actuary tables to determine when someone might die. They relied on information about a person's habits—including what they ate, whether they used drugs, if they smoked, etc.—occupation and family history to determine how long a person was likely to live. Insurance companies sent people throughout the United States to examine gravestones for birth and death dates to assign average longevity to families.
In an era when engineering and science meant certainty, these tables took on an air of certainty never encountered in human history. Certainty about what families might produce children who would live longer and which families would (certainly) experience a high level of childhood death. Certainty about which personal habits correlated with a longer life and which correlated with a shorter life.
In the progressive 1930s, all this knowledge seemed ripe for the project to improve humanity. Suppose an insurance adjuster could tell that smoking, for instance, or having black hair would be tied to a shorter life. Couldn't society use that knowledge to improve society by extending life?
For instance, if people with a yearly checkup tend to live longer, shouldn't everyone have one? Soon enough, insurance companies were building networks of doctors and requiring yearly checkups to keep an insurance policy current.
Lessons Learned
The experience of insurance companies in the 1930s through to the 1960s proved four things.
First, you can use statistics to predict human health and longevity. From here, it is easy to infer you can use statistics to predict many other things about humans.
Second, you could change human behavior for the person's good if you attributed your knowledge to science or engineering. If you place before people the carrot of having insurance and the stick of "if you don't, you're more likely to die early," you can change people's behavior—en masse.
Third, you can create certainty out of seemingly uncertain things if you approach the problem at a large enough scale. While they couldn't predict much about individual people, they could predict a lot about a person by examining the group he fits into.
The more information you know about a person, the better you can fit them into a group. The smaller the group you can divide humans into, the more precisely you can predict an individual's longevity.
The more data you can gather about a person or a group, the more precise your predictions will be. The yearly doctor's visit prescribed to improve the insured person's health was another way to gather data to improve the insurance company's understanding of people. Insurance companies could refine their models to a high degree of reliability through this information-gathering process.
Following this logic, it seems possible to predict everything about every person if you have enough data.
Fourth, Fear sells.
Two slides from a presentation about home safety and fire alarms from the 1970s
Everyone is afraid of dying—and even more of dying at an early age. If you tell a person you know how to let them live a longer, healthier life, they will listen—and buy what you are selling, no matter the cost.
Ratcheting up the fear sells more. In the 1930s, through the triumph of the women's rights movement, wives were afraid of their husbands dying young, leaving them (and their children) without support. Fear grew epic proportions in a society becoming increasingly mobile and hence less attached to long-standing communities and family groups.
Insurance salesmen learned to catch the wife while her husband was away, selling her on the insurance, letting her do the hard work of convincing her husband to buy the insurance, and then coming back to have all the paperwork signed.
Insurance and Progressivism
By the 1960s, innovation at insurance companies was stalling. There was too much information, and it was too hard to process effectively. However, a new force became common at that moment—the networked digital computer.
However, before we get to the networked digital computer, we need to look at another part of the world that relies on the same four lessons—behavioral prediction, molding human behavior, statistical certainty, and fear.
Marketing and manufacturing.
[1] This series was later published as one of a series of “long short stories” in Three of a Kind and made into a movie in 1944. The story is classic Noir.
[2] James M. Cain, Double Indemnity, Reprint edition (Vintage Crime/Black Lizard, 2011), 69.