The Data Revolution and Economic Analysis

Innovation and Economic Growth

Article Snapshot


Liran Einav and Jonathan Levin


NBER Working Paper No. 19305, May 2013


Big data will change the landscape of economic policy and theory. Data will complement sound theory and common sense. Consideration of many complex factors will replace simple models.

Policy Relevance

Economists will be able to make more effective predictions.

Main Points

  • In the past, a good retail data set would include records of product categories; today, data includes detailed product information, time of sale, shelf locations, and more.
    • Data is available in real time.
    • Data sets are much larger, including millions of transactions.
  • Businesses use predictive modelling and machine learning to target marketing or predict risk; economists are skeptical of this, as consumers can change behavior to defeat new policies. Big data could change this attitude.
  • Big data inspires novel research; One study found that replacing a teacher in the bottom 5% (measured by teachers’ effect on test scores) with an average teacher raises the lifetime earnings of students.
  • Economists construct simple models with a few changing variables to explore different outcomes; big data enables the use of methods from statistics and computer science, including machine learning, to consider many complex variables.
  • Economists modelling large-scale effects depict the average consumer, firm, or patient; with big data, researchers could include many individual variations from the average in their analysis.
  • Large-scale data sets available for research include the US Census and the Panel Study of Income Dynamics; some data from the IRS, Medicare, or Social Security Administration is available.
    • Privacy and confidentiality concerns must be addressed.
    • Some access protocols are cumbersome.
    • Norway, Sweden, and other nations show that much more access can be provided.
  • Private firms’ may ask researchers to sign nondisclosure agreements; researchers might need to combine multiple data sets to make interesting connections.
  • When working with large data sets, a part of the problem is to determine what is in the data, how to manage it, and what questions to ask.


Get The Article

Find the full article online

Search for Full Article