NYUSPS Schack Institute of Real Estate faculty member Tim Savage became a full-time professor in 2019 after years of working in data science and econometrics. Professor Savage, who holds a doctorate in economics from the University of North Carolina-Chapel Hill, has worked in areas such as machine learning and its use in business contexts, and has spoken about the intersection of commercial real estate, macroeconomics, and monetary policy. Professor Savage teaches students about the value of applying data science and economic tools to their careers in real estate. He spoke with Schack about his interests and the current and future landscape of the field.
February 3, 2020
Q & A with Clinical Professor Timothy Savage, Ph.D.
Schack: Can you start by telling me about your background in data science and economics and how you wound up in that area?
Savage: I earned a Ph.D. in economics with a specialization in econometrics, which is just the application of the tools of data science and machine learning to questions we have about how economies function. I studied mathematics as an undergraduate.
I went into consulting immediately after grad school largely focused on using these tools to evaluate the competitive implications of mergers. You can actually deploy these tools quite extensively doing merger analysis.
[After the financial crisis] I retooled myself as a financial econometrician using the same tools to answer a different question in the realm of finance. I did a fair amount of work in residential real estate. In 2016 I took over a research group called CBRE Econometric Advisors, based in Boston. After a while I missed teaching and opportunity arose to teach at Schack.
Schack: I wanted to ask what brought you into real estate—are there specific aspects of the real estate industry you find interesting from a financial or economic perspective?
Savage: Yes. As an asset class it’s highly differentiated, highly heterogeneous, which differentiates it from standard equities or bonds or even cash. What I try to teach in my real estate finance class is: the capital for commercial real estate is global, but the assets are local. It’s a different risk profile [than what] faces somebody trying to do equity or bond trading.
As an economic question, I think real estate now has emerged into being one of the most interesting asset classes out there. And the demand in the current economic climate of very low interest rates, negative interest rates in Europe, has increased the demand for physical or hard assets like real estate. There’s a wall of capital coming in to acquire these assets, and I find all that fascinating to think about.
Schack: Could describe how big data, machine learning, and other major technological trends have impacted economics in specifically real estate?
Savage: It’s starting to. In commercial real estate we are moving into a world where the data are much more granular geographically and we know much more about the individual asset. We know very specific features.
In commercial real estate big data will be wide and not long. Picture an Excel spreadsheet—you have many many columns and relatively few rows. There may be 10 million buildings in the US that large institutional investors might acquire, but they observe many things about those buildings.
The types of algorithms to study that question are very different from [those] that economists normally use to study labor markets. We have to deploy algorithms that are much more like [what] Netflix uses. Netflix observes many features about me so the algorithm is a feature reduction algorithm that basically says "what are the three or four most important features about Tim that can guide our suggestion regarding a movie?"
Schack: How does one assess the most useful categories of data when you have so many?
Savage: It’s a great question. In fact, we have to reach back to an algorithm that was developed 80 or 90 years ago, called the Principal Components of Analysis. It is able to reduce the variables or features that one would evaluate based on how much they drive the variability in the outcome we’re interested in—in this case, a film recommendation.
We can layer on top of it reinforcement learning where, for example, if the Netflix algorithm makes a recommendation to me and I don’t like it, I essentially can punish the algorithm. That improves the algorithm to me.
Schack: Could tell us something you’ve found in your research that you think is most surprising to people in your field or broadly?
Savage: I think in this field there is an overconfidence in the ability to forecast where in the near term where the economy is going. For example, many people were saying this time last year that interest rates were going up and made those statements with confidence. Yet we’ve seen a substantial decline in long-term interest rates even over the last year. We know it’s very difficult to predict the paths of interest rates; if it were easy, people would make a lot of money doing it.
Schack: Could you take us through what a project or a day might look like for someone with an economics or data-focused skill set? Thinking of a student studying under you today.
Savage: It’s helpful to think about a narrow problem, then think about an experimental design or data source that might be used to explore that singular use case. Hello Alfred is not a solution to the entire real estate industry. It is a solution to the large number of people who live in multi-family units who want one source for dry cleaning and takeout food.
Schack: What skills and ideas are you most concerned that students are walking away with?
Savage: What I really want them to do is to approach data science and machine learning with a critical view. I can develop 20 correlations between here and the door and none of them could be informative in a meaningful way. What I want my students to come away with is: just because it’s easy doesn’t mean that’s the right thing to do.
Schack: As you’re looking at housing trends in cities nationwide is there any advice you have?
Savage: Yes. As we move forward we need to think about more innovative ways to build out commercial real estate and to build out residential real estate. Looking at ideas around modular housing, where you can reduce costs by essentially creating a homogenous product that a resident themselves can adapt. That has the potential of going a long way to dealing with what is a real problem, which is the lack of affordability for lower-income people. The urbanization phenomenon extends across the entire income spectrum, and we do need to come up with creative ways.