Mortgages are the largest component of the fixed-income index which implicitly short options (prepayments). Therefore, unlike the non-callable sector of the fixed-income index, the actual performance of mortgages strongly depends on what happens with volatility. In CSFB research we have been emphasizing that taking a view on volatility is often an inevitable consequence of investing in mortgages. It is our view that we are in a period of low actual volatility and declining option demand.

Relevance Of Actual vs. Market-Implied Volatility

Recently, we have been emphasizing the relatively high level of market-implied volatility versus the actual mortgage volatility. We will briefly outline the relevance of each. Mortgage OAS are calculated using a term-structure of swaption volatilities (market-implieds). However, risk managers typically quantify the risk in mortgages using the empirically observed P/L volatilities. These are, obviously, influenced by the level of realized volatility. Therefore, VAR (value-at-risk) measures for mortgages are generally driven by the level of realized volatility. Similarly, the cost of dynamic delta hedging will be related to the realized volatility. Realized (or short-dated) volatility impacts the actual "convexity" cost of mortgages. On the other hand, the volatility exposure, or vega, of a mortgage security is determined by medium-term swaptions. From a total return perspective, actual volatility will affect convexity hedging costs (due to re-balancing) while medium-dated swaption volatility will primarily impact mortgage price performance.

Mortgage OAS Widen When Short-Dated Volatility Rises vs. Long-Dated

From an OAS perspective, medium-term swaptions, such as 5x10s, have the highest impact on mortgages. Changes in short-dated volatility (options with expires of less than one year) have a negligible impact on current coupon pass-through OAS, as mortgage options tend to be more long-dated. Consequently, we would anticipate an OAS widening when realized volatility is significantly higher than medium-dated swaption volatility (e.g., 5x10s). For the reasons discussed above, we expect the market to, in general, penalize mortgages more severely for short-dated volatility increases than would be implied by an OAS analysis.

Empirically, one can compare the level of the current coupon OAS against the volatility differential between 1mx10 and 5x10 swaptions. Instead of using actual volatility, it is more practical to use 1mx10 swaptions as a proxy, as they are very highly correlated with actuals and provide a smoother series. The OAS in the CSFB model are calculated using a daily calibration to both the 1mx10 and 5x10 swaptions (as well as the rest of the term-structure of volatility). A comparison reveals that OAS widenings predominate when the volatility differential between medium- and short-dated options is approximately greater than 3%. (When the volatility differential is smaller, widenings and tightenings occur with similar frequency and are more equally distributed around the expected relationship.) Instances when the differential is greater than 3% also correspond to periods of high actual volatility.

Actual And Market-Implied Volatilities Converge

We frequently compare the actual daily mortgage rate volatility against the implied levels from 1mx10yr and 5yrx10yr swaptions. The actual mortgage volatility has recently been about 30% below market-implied. That would correspond to an expected excess return of 20 bps for the current coupon mortgage over that derived from market-implied OAS. It is our view that this divergence will not persist at the current level. We anticipate a decline in option demand due to four factors: low actual volatility, an increase in puttable/callable corporate bond issuance in first quarter 2000, in the current rate environment servicer pipelines should be stable, and, finally, FASB 133 which makes accounting for option hedges more complicated, and therefore makes options less attractive for servicers. Some of the decline in volatility has already been observed over the last week, where 5x10 swaption dropped by half a vega.

To summarize, we anticipate that the low volatility environment will be a major driver of mortgage performance in the first quarter of 2000. Late in the final quarter of 1999, we already observe increased trading volume and liquidity in pass-through. In particular, dealers seem far more comfortable with mortgage risk and appear to be increasing exposure.