Page 39

CRE Finance World, Winter 2012

Valuing Appraisals: Evidence from the CMBS Industry Extending the grenade metaphor to this reduced dataset, 60% Chart 3b brings more order to the data. The histogram shows of liquidations are in the “close enough” buckets between 0.75 buckets of AV/GP ratios. By way of example, the most frequent and 1.25:1. AV/GP ratio is between 1.0 and 1.1:1 — 311 liquidations out of 2,076 fall into this bucket. The fat right-handed tail reflects the If we strain to capture all 2,076 observations in a scatter plot, we 121 liquidations where AV/GP exceeded 2.0:1; that is, where the see in Chart 3a4 that over time more appraised values to liquidated appraisal exceeded 200% of the gross liquidation proceeds. If values exceed the ideal benchmark of 100%. Recall that this Chart you tally all appraised values and divide by total gross liquidation 3a (and 3b) consider only liquidations that occur within one year of proceeds, ($8.4 billion divided by $7.6 billion), you achieve a very the appraisal date. Not only does the accuracy improve considerably, benign 1.10:1 AV/GP ratio. As can be seen from both charts, this but there is a slightly downward slope to the relationship between aggregate accuracy misleads due to large, offsetting errors. While appraised values and gross liquidation proceeds over time. This the difference between total appraised values and gross liquidation supports our intuition: stale appraisals are less reliable than more proceeds is only $777 million ($8.4 billion minus $7.6 billion), the current appraisals. Where the larger dataset including appraisals absolute value of all errors approximates $2.1 billion. Of the 2,076 conducted more than one year from liquidation date produces an loans, 1,332 produced appraised values exceeding gross proceeds aggregate AV/GP ratio of 1.39:1, when the dataset is bounded by by $1.4 billion in aggregate. A further 737 loans produced appraised a one year gap, the ratio improves to 1.10:1. values $661 million less than gross proceeds.5 The line estimate on Chart 3a also supports the theory that the Chart 3b accuracy of appraisals has improved slightly over the four and one- All Property Types half years of data. Notwithstanding the Jackson Pollock jumble of Frequency of Liquidations by AV/GP; N=2,076 the chart, regressing liquidation proceeds against appraised values produces a 0.74 R2. For this dataset, a line model estimate of appraised values predicts 74% of changes in liquidation proceeds. But the standard deviation of errors (appraised value minus gross liquidation proceeds) is huge — exceeding $4 million. With an average appraised value of $4.0 million on these liquidated loans, users of appraisals should proceed cautiously. Chart 3a App. Value/Liq. Proceeds; N=2,076; X Axis = Loss Date Testing The Data Are there specific factors apart from the skills of the engaged appraiser that factor into the accuracy of the appraised value? Several different factors can be tested including the location of the collateral, the point in the market when the collateral or note was liquidated, differences in proceeds between note sales and property sales, the size of the loan and the property type, to name a few. Of these, this article investigates the latter two. Conventional wisdom theorizes that property types with longer lease structures would be easier to present value; rollover risk is less prominent and the costs and uncertainty of re-letting space is mitigated. If this is A publication of Winter issue 2012 sponsored by CRE Finance World Winter 2012 37


CRE Finance World, Winter 2012
To see the actual publication please follow the link above