Elena Manresa is fascinated by what she calls “the enigma of choice.” As an econometrician, she does statistical modeling based on the idea that economic data is different from any other kind of data (like medical or engineering data) because it shows us the decisions made by people, but not why. The why behind a data set is what Professor Manresa seeks to unveil.
Right now, Manresa says, “there is a significant interest in understanding the effects of living in a particular neighborhood or having certain associations.” Often, she says, we assume we know who is interacting with whom, but Manresa used panel data (a sample of individuals or firms over a long period of time) to provide a concrete method to show one’s network and sources of influence. In a recent project, “Estimating the Structure of Social Interactions Using Panel Data,” Manresa measures Research and Development (R&D) money that firms put into hiring scientists or developing new technology. Manresa observes that while the output of a firm varies according to its own R&D investment, we can still see that across a wide selection of firms, the output moves in similar patterns and varies according to the R&D of other firms. This suggests an influence or association between these firms. “The effect of networks can have significant implications for the design of economic policies such as tax breaks or subsidies,” Manresa says.
Manresa is also compelled by the potential of machine learning techniques to explain economic phenomena. In the hands of an econometrician, the algorithms computer science is producing can be used in the estimation of structural models. Manresa is currently studying algorithms that create pattern recognition (such as when iPhoto recognizes a face, or when online browsing behavior is mined to offer personalized ads). She believes that such algorithms, which sort through large amounts of data to find patterns, could be incredibly useful to economics by way of potentially ordering a huge mess of numbers.