Market Compares Bloomberg Model to Andrew Davidson Model
Traders and investors must take a number of factors into consideration when they choose between mortgage prepayment models, and market players are currently comparing and contrasting the new default proprietary model recently launched by Bloomberg (the BPM model) with the model which it supplanted, the Andrew Davidson Prepayment Model (ADP model).
Although an article published by PaineWebber recently characterized the new Bloomberg option-adjusted spread (OAS) model as "a decided improvement," an open letter written Andrew Davidson & Co. in response to PaineWebber's comments pointed out deficiencies both in PW's review as well as the new Bloomberg model itself.
"Perhaps the most important element in understanding prepayment models is recognizing that all models have limitations in that they are an attempt to predict the future, while looking at the past," said the open letter, which was written by Andrew Davidson, the president of the company. "As we have described elsewhere, forecasting prepayments is like driving while only looking in the rearview mirror."
PaineWebber's analysis of the new Bloomberg model mentioned both advantages and caveats of the new quantitative tool, but stressed that "the model's OAS configuration is very much in sync with what common sense tells us about the market."
More specifically, the PaineWebber report stated that OASs on conventional 15-year collateral are higher, and more reasonable, on the BPM than the ADP model; it also mentioned that OASs on Ginnie Maes are lower on BPM than on ADP, "which fits intuitively; it said that Bloomberg's OASs on GNMA II's look much more consistent with GNMA I's; and "results from the new BPM OASs give results far more consistent with our market impressions than those provided by the ADP model."
The PW article did go on to mention many disadvantages of the Bloomberg model, the most salient one being that the BPM continues to use the Treasury yield curve to generate the prepayment speeds for Libor OAS, which "can bias OAS results against more prepayment-sensitive securities."
Though the PaineWebber report did note the downside to the new model, Andrew Davidson's response asserts that PW's comparison focused primarily on one aspect - relative OAS between products, which is a relevant consideration, "but not the only consideration." According to Davidson, the ADP model provides very reasonable relative OAS levels among conventional, government and jumbo mortgages.
Additionally, the ADP model features consistency of OASs across coupons. According to Davidson, for thirty-year conventional products, "our model produces OAS levels that vary by only three basis points across coupons from 6.0 to 8.0." In contrast, the BPM produces a range of nine basis points.
Further, Davidson says that the ADP model differs from the Street's models in that it has somewhat faster prepayments on discounts - which seems to be in line with a strong economy that has kept housing turnover at high levels.
"While Dealer Prepayment forecasts vary with fad and fashion reflecting short-term sentiment of the market, we strive to provide results that are more consistent," Davidson said.
Regardless of the differences between the models, it is clear that each investor looks for different qualities in a model: some look primarily at results - OAS numbers - and decide whether they like OAS for a particular instrument or on a particular date; others look at risk numbers or effective durations and determine what they imply in terms of hedging; some look at the long-term performance of a model, comparing the forecast with what has actually occurred (historical analysis); and finally, others look at prospective performance - deciding whether the model's forecasts fit with your own judgment.
"Investors need to consider all of those," Davidson said. "The broader the range of models and assumptions they're looking at, the better the understanding they'll have of their instruments."
