Thought Leaders in Big Data: Interview with Sandy Steier, CEO of 1010data (Part 3)

Sramana Mitra: More than they needed to.

Sandy Steier: One of the things that now everyone knows is that they are very complicated securities. So complicated in fact, that in some cases they are too complicated to be safely traded. In those days they weren’t as bad as today; nevertheless, they required a lot of analysis, and they continue to require a lot of analysis. The kinds of analysis you do on them, besides various kinds of mathematical modeling, incorporates all sorts of market information.

A mortgage-backed security is a security, like a bond, that is collateralized by a pool of mortgages. When the homeowners pay interest on their mortgages, that gets passed through to the investors. So, the investors are lending to the homeowners. In order to understand how a security is going to perform in the future, it is important to know if the mortgage will pay off early. That can happen if someone moves early and they pay off their loan, or they default on their loan – unfortunately that has happened quite a bit over last few years. There are all sorts of reasons why a loan may suddenly go away. It is important, if you are trading securities, that if you depend on cash-flows from those mortgages you know what is going to happen to those mortgages, or at least have a good guess at what is going to happen.

What you typically do is analyze the historical performance of mortgages. There are tens of millions of outstanding mortgages in the U.S. Over the course of the last few decades, there have been closer to 150 million mortgages that are gone, but they did exist. There is historical information about all those mortgages that tell you what happened to them when interest rates go up or down. Do they pay off? Are there issues? Looking at the mortgages helps you figure out what mortgages you own might be credit information about the home owners. If you happen to know that this house is owned by someone who also has a bunch of credit cards and those cards are getting maxed out, that is not a good sign. There is a good chance that this mortgage is in danger and that it might default in the near future. This can be done on an anonymous basis – you don’t need to know who the person is, you just know there is a loan in a certain area and you happen to know that whoever owes that loan also has the following credit score and has the following credit cards and auto loans. That is important information.

Additional information would be housing price appreciation or depreciation. How are the value of houses appreciated or depreciated in the area where that house is? This also gives us metrics about interest rates and various other things. This is just the tip of the iceberg. There is a lot of other information – dozens of data sets available to do that analysis. Some of those data sets are relatively small, but some of them are big data. Some of them have billions of records.

We are very fortunate in that most of the players in this market, especially in the sub-prime market, which was still a tremendous hit five years ago, use 1010data to do that kind of analysis. The way that works is that there are all these data sets available to various vendors who collect that information – sometimes it is free and provided by the government, in other cases companies collect this information. Those owners of the information put the information onto 1010data, where we take it from them, and then all the users of the data use it on 1010data. It is a sandbox where all the suppliers of the data put the information onto it and most users of the data use it in 1010Data. There are two advantages to that. One is that the suppliers of the data don’t actually have to send the data to all the different players – they put it in one place. The other advantage is that these various users of the data don’t have to build their own databases and do their own thing with the data. The data is already there.

SM: In this case, how do you get paid?

SS: If some investor wants to analyze the data they buy the data – they buy the rights to the data from the data vendors. Then they also pay us, 1010data, for the rights to use the system and for us to maintain that data on our system. Basically the customer ultimately pays two parties. They pay us for the use of our system and for the fact that we are managing the data on our system, and they pay the data vendor for the data itself.

This segment is part 3 in the series : Thought Leaders in Big Data: Interview with Sandy Steier, CEO of 1010data
1 2 3 4 5 6