5.1 Introduction
In the early 1970s, San Francisco was completing a huge new infrastructure project, the Bay Area Rapid Transit (BART) system. The project initially cost $1.6 billion and included tunneling under the San Francisco Bay. Policy makers were obviously interested in determining how many people would use the new system once it was built. But that is a problem. How do you predict the demand for a product that does not exist?
One solution is to ask people. A survey was conducted of people who were likely to use the new transport system. The survey asked detailed questions about their current mode of transport and asked them whether they would use the new system. The concern is that it is hard for people to predict how they would use something that does not exist. Berkeley econometrician, Dan McFadden, suggested an alternative approach. Instead of asking people to predict what they would do, McFadden suggested using information on what people actually do do, then use economic theory to predict what they would do.
McFadden argued that combining survey data with economic theory would produce more accurate estimates than the survey data alone (McFadden, 1974). In the case of the BART survey, McFadden was correct. According to the survey data, 15% of respondents said that they would use BART. McFadden estimated that 6% of respondents would use BART. In fact, 6% of respondents actually did use BART. Survey data is valuable, but people give more accurate answers to some questions than others.
The first part of the book discussed how exogenous variation is needed to use observed data to predict policy outcomes. Chapters 1 and 2 assume that observed variation in exposure to a policy is determined independently of unobserved characteristics. Chapters 3 and 4 relaxed this assumption but allowed economic theory to be used in estimating the impact of the policy. This part of the book extends the idea of using economic theory. This chapter introduces the idea of using revealed preference.
Today, the ideas that McFadden developed for analyzing the value of BART are used across economics, antitrust, marketing, statistics and machine learning. At the Federal Trade Commission and the Department of Justice, economists use these techniques to determine whether a merger between ice cream manufacturers, or cigarette manufacturers, or supermarkets, or hospitals, will lead to higher prices.
When Google changed the way it displayed search results, user traffic moved away from Google’s competitors. Such actions by a dominant firm like Google could lead to antitrust actions unless the changes also made users better off. By combining economic theory and data on the behavior of Google’s users, we can determine wheth...
