Michael Brandt on Nowcasting

Pre-Reading Video

Live Session Recording

Michael Brandt

In our June Fuqua Faculty Conversation, Michael Brandt, Kalman J. Cohen Professor of Finance, presented:

Nowcasting: How to go from occasional economic news to a real-time read on economic activity and sentiment

What does economics have to do with meteorology (besides the obvious, that both are notoriously bad at forecasting)? Surprisingly much. Both are quantitative sciences, they deal with highly complex dynamic systems, and they rely on large quantities of incomplete, delayed, and often heavily contaminated data. Much like the Doppler Radar system on the local news tells us the current weather conditions (a practice meteorologists call “nowcasting”), Professor Brandt and corroborators have developed a technique that gives us a clear picture of current economic conditions. They take the information contained in thousands of economic data releases each month, such as CPI or the unemployment report, and distill it into easy-to-interpret measures for the four key dimensions of economic activity: output, employment, inflation, and sentiment.

View Professor Brandt’s Bio

One Response to “Michael Brandt on Nowcasting”
  1. john argenti says:

    Interesting topic, I look forward to your presentation. I am an alum and provide research to institutional investors. One comment I take issue with is the statement that economist forecasts are only available in monthly snapshots. For professional investors Bloomberg updates economists’ average forecasts daily, charts how the forecast has evolved etc… So to be really interesting your output has to be differentiated some other way. It should become cheaper to generage over time than the aggregate cost of “boots on the ground” economists but they get paid for many other things beyond the actual forecasts so arent going away, and the incremental cost of their consensus output is zero anyway to institutional investors who get it for free bundled with other research or from in their subscriotion to Bloomberg, Factset etc… What would be interesting if your forecasts are better than consensus not just matching. Does your model consistently predict certain individual numbers like employment more accurately than the consensus figures for example? Thanks,

Leave A Comment