Linking Economic Conditions to Financial Markets
Often times macro investors will look to map real economic data to financial market prices. This helps to come up with views on fair value and expected returns. But despite the clear theoretical underpinnings for this approach, most times economic data is only loosely connected. If the theory is right then this disconnect should partially be explained by measurement. Specifically, what are we measuring? should be a question asked by every investor that relies on aggregate economic statistics.
with the availability of property level detail, mortgage and real estate investors can overcome some of this data mismatch. By using dis-aggregated economic data in conjunction with loan level details one is able to get a sense of the economic conditions pertinent to a particular bond.
To illustrate, I map the FNMA multifamily loans database to local county and zip-code level labor market and housing conditions.
Although, I make it look easy, the hard part of this work is generally the mapping of names across platforms. Here I've used Zillow to map the zip codes to county names and city names. FRED uses a similar mnemonic for county names, but at times diverges from this mnemonic.
For economic data I've used the county level monthly unemployment rates. The idea is that I want to get as dis-aggregated as possible while maintaining a decent handle on data timeliness. Unfortunately there isn't much data which exists at this level of geographic detail and frequency.
We can use GeoFRED to collect the county level code names and download these into R using the quantmod package. As a side note this exercise taught me that I have 24 GB of RAM when I thought I had 36 GB! I also purchased an external 256 GB SSD hard-drive to help me deal with the large files that loan level data necessary require.
The figure below is the final product. FNMA DUS bonds have exposure to areas with lower unemployment rates than the national average. This might help partly explain their extremely low delinquency rates.
with the availability of property level detail, mortgage and real estate investors can overcome some of this data mismatch. By using dis-aggregated economic data in conjunction with loan level details one is able to get a sense of the economic conditions pertinent to a particular bond.
To illustrate, I map the FNMA multifamily loans database to local county and zip-code level labor market and housing conditions.
Although, I make it look easy, the hard part of this work is generally the mapping of names across platforms. Here I've used Zillow to map the zip codes to county names and city names. FRED uses a similar mnemonic for county names, but at times diverges from this mnemonic.
For economic data I've used the county level monthly unemployment rates. The idea is that I want to get as dis-aggregated as possible while maintaining a decent handle on data timeliness. Unfortunately there isn't much data which exists at this level of geographic detail and frequency.
We can use GeoFRED to collect the county level code names and download these into R using the quantmod package. As a side note this exercise taught me that I have 24 GB of RAM when I thought I had 36 GB! I also purchased an external 256 GB SSD hard-drive to help me deal with the large files that loan level data necessary require.
The figure below is the final product. FNMA DUS bonds have exposure to areas with lower unemployment rates than the national average. This might help partly explain their extremely low delinquency rates.