The name sounds invented.
Simulmatics.
It could easily belong to a fictional contractor in a novel about manufactured reality.
It did not.
The Simulmatics Corporation was founded in 1959. It assembled political scientists, psychologists, computer specialists and advertising people around an audacious proposition: large amounts of human behavior could be converted into data, modeled by computers, and used to predict what people would do next.
The company did not invent political polling. It tried to do something more ambitious with the information polling had already accumulated.
It wanted to simulate the electorate.
480 kinds of voter
Researchers began with years of archived public-opinion polling. Different surveys had asked different questions in different ways, so the information first had to be normalized.
Then came the abstraction.
The electorate was divided into 480 voter types, defined through combinations of characteristics such as geography, urbanization, income, race, religion, sex and political affiliation. Political questions were reduced to 52 issue clusters.
The resulting structure was essentially a large matrix describing how particular categories of people had responded to particular categories of questions. Simulmatics then constructed a simulated population that researchers could use to explore how changes in issues or strategy might affect voter behavior.
It was crude by modern standards. It was also recognizably modern.
- The individual person disappeared into variables.
- The variables became a model.
- The model became a tool for deciding what might work.
That is the important lineage.
Kennedy, 1960
Simulmatics conducted research for the Democratic Party during the 1960 election. Its first major political work included analysis of Black voters in Northern states. After Kennedy received the nomination, the Democratic National Committee commissioned further work, including analysis touching on his Catholicism.
What the campaign actually did with any of it remains disputed, and this page does not resolve the dispute. The claim worth making is narrower and harder to argue with: by 1960 it was possible to propose, seriously and to serious people, that an electorate could be simulated before it was addressed.
What survived the company
Simulmatics did not last. The idea did.
Strip away the machines and the vocabulary of the period and the proposition is the one every targeting system since has rested on: that a population can be decomposed into categories, that the categories predict response, and that a message can therefore be chosen rather than argued.
The company's later work in Vietnam matters for the same reason. It was the same conviction pointed at a different population, under conditions where being wrong cost more.