An excerpt of a larger article by Susan Kulakowski, a director in the residential mortgage group at Fitch. The entire article is available in the "Mortgage Principles and Interest" newsletter, a regular product from Fitch which can be accessed at www.fitchratings.com.
Mortgage lenders have traditionally relied on the "three Cs" of lending - Collateral, Credit, and Capacity - when approving or declining a mortgage application. In its rating approach, Fitch has described the credit and capacity issues of the origination process. This article describes the basic types of collateral valuation that are currently in use, focusing on automated valuation systems that complement the work of appraisers.
Purpose of Full Appraisals
The accurate valuation of collateral is a crucial component in the process of originating a mortgage. Traditionally, appraisers have provided this service of a "full" appraisal to mortgage originators, basing their valuation on an inspection of the property, a comparison of the subject property to similar properties in the vicinity that have recently sold, and an understanding of current local market economics. The purpose of this appraisal is to confirm the market value of the property. The lender then compares the market value to the contract sales price to verify that the balance is permitted under the lender's underwriting guidelines. The lender can then fund the mortgage with the expectation that the property is not overvalued and, should the lender foreclose on the property, can be sold without a loss.
For a purchase money mortgage, the buyer and the seller have already determined the value of a property by agreeing to a purchase price. Assuming that the sale of the property is an arm's length transaction between independent buyer and seller, the sales price is the most accurate estimate of the "true value" of a home. It is unusual for an appraisal to differ greatly from this sales price for "typical" properties (although somewhat more likely for properties that have atypical features and/or are very expensive).
However, for nonpurchase money mortgages, accurate appraisals become much more difficult. With a refinance mortgage, for example, where there is no confirmation of value by an independent buyer, the appraiser must rely on the sales prices of recently sold, similar or "comparable" properties, and recent sales trends in the local market. This lack of an objective sales price increases the variance of the appraisal and may result in a lower than anticipated sales price after foreclosure.
This risk is minimal for rate/term refinances, where the borrower is simply refinancing an existing mortgage to obtain a better mortgage rate or term. However, this risk is exacerbated with cash-out refinances, where the borrower monetizes estimated-appreciation-based equity from property, and the amount monetized drives the transaction's attractivenes to the borrower.
Finally, other origination guidelines may not require a full appraisal or any appraisal at all. Under "125 mortgage" or "high loan-to-value ratio (LTV) mortgage" programs, which require limited or no equity from the borrower, a lender may rely on only the borrower's credit and capacity when approving or declining a mortgage application. Because the lender is not relying on the property value in the event of foreclosure, the expense of an appraisal may be too great to justify for these lending programs.
During the past several years, originators and other market participants, such as mortgage insurers, have developed alternative means of valuation to supplement or replace full appraisals. Originators have done so because appraisals are not always appropriate and because automated alternatives can provide more consistent, objective results at a substantial cost savings over full appraisals. In addition, the best automated systems provide not only a valuation for the subject property but also an indication of the "confidence" of the valuation. This indication of the accuracy of the automated valuation provides the originator with the likely range of values for the subject property. It will be a narrow range when the model can provide a highly accurate estimate and a wide range when the model cannot.
Of the three common valuation methods, comparable sales analysis most closely resembles the traditional appraisal. An important component of the traditional appraisal is the documentation of the characteristics of the subject property, as well as the features of other recently sold properties that are physically close to the subject property and of similar age, characteristic, and condition. The sales prices of these "comparables" support the appraiser's opinion of the value of the subject property. In a similar fashion, comparable sales databases are culled to find properties similar to the subject property in age, characteristic, and condition. Based on analysis of actual sales, the value of each characteristic is determined and applied to the subject property, providing a current estimate of the value of the subject property.
Repeat Sales Indexing
An essential form of statistical estimation of property values relies on a repeat sales methodology to create an index of the housing price trend in local markets. The repeat sales methodology matches the recent sale of a property to the sale of the same property sometime in the past and calculates the appreciation or depreciation from the prior sale to the recent sale. With enough matched pairs, accurate indexes of market appreciation can be charted at various geographic levels. The indexes can then be used to "bring forward" the value of the subject property from the time of its last sale to the present.
The most important advantage that housing price indexes provide is that they reflect the performance of the entire market. Because the indexes are built using market data, they accurately track the broad trends of the entire market. These indexes are also easy to use when evaluating an overall market and valuations on individual properties are simple and inexpensive to obtain.
Hedonic models are the third commonly used valuation method. These models are built on property characteristic databases that contain detailed information about the features of each property, as well as each property's sales price. Regression analysis is performed to build a model that describes the monetary contribution of each property feature to the overall price of a property. Because homes with similar features generally bring similar prices, the relative value of features and, consequently the overall value of a property, can be accurately established.
The automated valuation methods described provide several advantages over full appraisals. Not only are they more convenient and less expensive to lender and mortgagor but, if properly developed, are also statistically unbiased. Several studies conducted by major lenders to validate estimates against the actual sales prices generally show that automated estimates tend to be unbiased while appraisals tend to be biased upward. That is, automated models are no more likely to overestimate a value than they are to underestimate a value, while appraisals are somewhat more likely to overestimate a property value.
The most powerful approach to property valuation combines the three automated approaches - comparable sales, housing price indexes, and hedonic models - to obtain accurate, unbiased estimates of value. None of these approaches is foolproof; all are limited by the quality and quantity of the available data, as well as the statistical skills of the model developers. Particularly in rapidly changing markets, the savvy appraiser brings an informed, current understanding of market developments to the appraisal process. An automated process cannot easily duplicate that understanding. However, property valuation is a process that lends itself well to a statistical approach. Given the appropriate data and models, automated valuation systems can offer reliable and timely estimates of value at a lower cost than do the traditional full appraisal process.
Although few originators have moved to an entirely automated appraisal process, many use automated valuations for less risky mortgage applications and as part of a quality control review of appraisal quality. For these purposes, automated valuation models provide less expensive, unbiased and potentially more accurate valuations, especially for nonpurchase money mortgages. In addition, because the automated systems provide confidence measures, the lender receives a more balanced picture of the realistic range of a property's true value. Automated valuations may be less appropriate in those instances where the data are limited or unavailable, the property to be valued is unusual in some aspect, or significant improvements have been made to the property since its last sale.
In addition, full appraisals will provide more accurate valuations in rapidly changing markets because of the delay in data collection and because appraisers, knowing their markets, can react to changes far more quickly.