Get a bunch of real estate people together, and inevitably the conversation will eventually turn to deals that went bad. The war story usually ends with a guiding principle the person took away from the experience, and if the story is a good one everyone who hears it adopts that principle too. However, way too often these principles are wrong. A few examples:
1) The manager of a group I used to work with (a very bright, experienced guy) was absolutely, positively convinced that Low Income Housing Tax Credit (LIHTC) projects which underwent only moderate rehabilitation were higher risk than gut rehab or new construction projects. The argument was that future capital needs were almost always underestimated for moderate rehab projects, while gut rehab and new construction projects started out in new condition. This story made sense, and when you looked at the problem properties in the portfolio it was true that 75% of them were moderate rehab projects. The only problem was, 75% of the entire portfolio were moderate rehab projects, which meant the moderate rehab projects were not underperforming after all.
2) The manager at another bank I worked at was convinced that mobile home parks were relatively high risk. That was contrary to my experience, so I did some digging into the past negative experiences and discovered it was true that there were previously underperforming mobile home park loans in the bank's portfolio. But, those loans underperformed during the most recent recession, during a period when many loans of all property types underperformed. In fact, although they were on the bank's watch list the mobile home park properties performed better than many of the other property types.
Why did these mistaken conclusions happen? Several biases are at work:
Availability Bias. People sometimes erroneously draw conclusions based on how easily an example comes to mind. The manager in the first example could easily call up many examples of underperforming moderate rehab projects, which led him to the erroneous conclusion that these type of projects underperformed. Recent events and vivid events also are overweighted. For example, after September 11, 2001, terrorism insurance on even moderate size real estate projects in moderate size metropolitan areas was seen as essential. This is no longer true, although there is no reason to think the risk has abated.
Attentional Bias . People sometimes focus on a limited number of possible explanations when examining correlations, and exclude the correct explanation. The manager in the second example focused on property type as the explanation for poor performance, and ignored the time element (recession conditions) which was a better explanation for the underperformance.
To understand why a property underperforms, you need to have a large, diverse sample and look at the correlation between a characteristic and underperformance. This is not something that is commonly or easily done. Most lenders do not have portfolios that are large and diverse enough to draw good conclusions from. I have worked for two organizations that did have such portfolios and did the work, but the conclusions were always viewed as proprietary. Fannie Mae, Freddie Mac, some of the large servicers, and CMBS data aggregators have the data and are probably using it internally, but it's unlikely to become public.
Even if the data were publicly available, I doubt it would be very useful because there is a huge omission in the data being collected. When I did my previous work, I initially was very disappointed to find only a few attributes that correlated with performance, and those correlations were weak. Property age, initial LTV and/or DSC, refinance vs. purchase, initial property condition, size of project - nothing I looked at correlated with property performance. It was only when I went back to original loan files and started looking at sponsor characteristics that I found some meaningful correlations. Specifically, sponsor net worth, liquidity, and experience in the property's market all had strong correlations with property performance. Why this is so is a topic for another post, but without controlling for these attributes any conclusions about other factors are going to be faulty.
Unfortunately, no one captures this information in a data tape form - it has to be dug out of the original loan files. I've asked people I know at rating agencies and Fannie Mae why they don't capture these attributes and the answers are (a) these are non-recourse loans so borrower financial attributes are not relevant, (b) the loans are generally assumable so the sponsor may change, and (c) the sponsor's financial situation will change over time, and there is generally no requirement for updated financial statements. These answers all miss the point; the attributes are highly correlated with performance, and should be captured.
In an ideal world the ratings agencies and regulators (FDIC, OFHEO) would require lenders to report standardized data at loan origination (including sponsor experience and financial information) and again at loan default (something similar to a death certificate). This data (redacted to remove confidential information) should be available for analysis to academia and the public. Until then, lending decisions will continue to be based on the limited experiences and biases of the lending officers making the decisions.
|