Prof. Steve Levitt on Big Data's Value

By: Sonal Somaiya, Class of 2019

Last week, renowned economist and University of Chicago Economics Professor Steve Levitt, co-author of Freakonomics, its many sequels and co-host of the Freakonomics Radio podcast, hosted a brown bag luncheon on “Learning From Data” at the Harper Center. Per the title of his talk, Levitt’s centered his comments around how data is informing some of the real time economic outcomes we observe and act upon in the global marketplace.

Levitt focused on a few key examples of first, how data collection is increasing the robustness of our study of economics and second, how that data is showing unique trends that are also related to the work of another University of Chicago bigwig, Richard Thaler, in behavioral science and economics.

steve 1.jpg

As we all know from our economics courses, it’s easy to assume that things like demand curves are perfectly linear and smooth but this is perhaps an assumption we won’t need to make in the data-driven future. To prove this point, Levitt highlighted the example of Uber. Uber, he noted, is one of the first companies whose data could truly get us to an accurate, real time estimation of a demand curve for ridesharing options. Uber’s application effectively tracks customer sensitivity to a whole range of different prices across geographies which, when plotted, does in fact look like the demand curves we draw in an economics class. Levitt noted that this is a “legendary” development in economics and pricing strategy. Instead of having to estimate and triangulate on concepts like consumer surplus and willingness to pay with tools like a conjoint analysis, Uber’s data gave management unbelievably accurate insight into exactly what their consumers were willing to pay for rides and allowed them to price accordingly. Knowing things like GPS coordinates of users also allows Uber to cut their data by geography, projected income level and gender to tease out other informative trends in demand.  

But the data don’t always show us trends we might expect. Levitt highlighted another example where the data didn’t necessarily show us a trend that economic theory may have projected. On their website, FreakonomicsExperiments.com, Levitt and his co-author, journalist Stephen Dubner, ran an experiment where users could seek help making a tough decision… with a coin flip. Their experiment was designed to measure whether the coin indicated whether to make a change or remain with the status quo decision, whether or not the person took the action and their subsequent level of happiness a few months after making the decision. Though a totally unrelated coin toss should hypothetically have no impact on any of these measurable factors around a decision, Levitt and Dubner found quite the opposite. “Individuals who are told by the coin toss to make a change are much more likely to [do so] and are happier six months later than those who were told by the coin to maintain the status quo.”

Now as cool as that might be, it’s a terrifying thought. Can our satisfaction with a life decision really hinge on a random visit to a website where you click to flip a coin? Why does the data show this trend? Should we all start flipping coins to potentially increase our satisfaction with tough life decisions? 

A final point Levitt highlighted was that data must be taken with a grain of salt. Put another way, big data and ideas are complements, they go hand in hand, but data does not substitute ideas and analysis. How we actually leverage the fact that data shows that people’s happiness is influenced by the mere addition of a coin flip exercise is where the real power of data lies. Levitt’s newest endeavors focus on this. How can better, more specific knowledge of our economic presence through demand, supply and surplus actually influence some of the decisions we make, how we frame questions or how we structure processes. Think back to a great example from Richard Thaler’s work in behavioral economics. Merely flipping a prompt from opt-in to opt-out can have humungous impact on how much people actually use a product. But it is just as important to gather that data as it is to leverage it to inform how you structure a 401K program, for example.

As Levitt summarized, “everything is data, it’s just a matter of [whether or not] I am clever enough to figure out how to use it”. As we move into an era where data about every action we take will be captured, the combined forces of economics and computer science will allow us to use that knowledge to create impact and change.