In the summer of 1969, I was in Kenya. I had left MIT in 1965, and these were economies, in Africa, that were clearly not described by the standard models. There were some developing-country problems like sharecropping, for instance, that were not standard problems that economists had worried about for a long time. But the most dramatic was the unemployment problem. The standard theory is that demand for labor equals supply of labor and there’s no such thing as unemployment. And obviously that’s wrong, because we see all the time periods of unemployment. Central bankers and conservatives, people like that, they say the problem is unions have raised wages too high. Or that government has put on too high of a minimum wage. And what was so interesting about Kenya was you couldn’t use those theories to explain what was going on, and yet they had very high unemployment. Wages were, in the urban area, clearly much higher than in the rural area, and that was attracting migration from the rural to the urban to get jobs. The question then was why didn’t the competition drive down the wages in the way that standard theory predicts? That totally undermined both the standard theory and the standard explanations for what happens when the standard theory doesn’t seem to be working.

Now, for hundreds of years economists had used models of some imperfect information but had recognized that the world wasn’t well described, perfectly described, by perfect information. But they didn’t know how to analyze, formalize, rigorously think about information imperfections. I came back from Kenya all excited about the new insights I had, and I was on the phone with a guy, George Akerlof, who had been my closest friend when I was at MIT, about the insights that he had. He had just come back from India. And some of his basic insights came from his thinking about India � I often think it helps people to stimulate their mind by being in a very different situation. It required a real paradigm shift. It was a big shift because this was a model that had dominated � the perfect-information model had dominated for centuries.