There’s an old saying: “It’s not the things you don’t know that get you into trouble, it’s the things you know for sure that ain’t so.” I see this problem frequently in the mortgage valuation world. It’s especially common among recent business school graduates who are often captivated by elegant mathematical models that may or may not apply to the task at hand. Three examples come to mind:

1. Option-Adjusted Spread (OAS)

A highly valuable tool, just not for asset valuation. The OAS process involves randomly generating hundreds of projected interest rate paths, calculating the net present value (NPV) of cash flows for each path using a risk-free rate plus a spread, averaging these NPVs, and then adjusting the spread to align the simulated value with the actual market price.

Needless to say, this assumes you already know the market price. In practice, most users input the spread first and then calculate what they believe to be a market price. While this approach can shed light on the convexity characteristics of a portfolio, it does not necessarily produce a market-reflective price or accurately model borrower behavior.

2. Dealer “Consensus” Prepay Models

A complete misnomer. While these models highlight differences in prepayment speed assumptions among major broker-dealers, there is no true “consensus.” Measures of central tendency (typically mean or median) used to fit these models into valuation frameworks obscure the large standard deviations among contributors. As a result, these models are essentially useless as predictive tools.

3. The Forward Curve

This mathematical construct makes intuitive sense but, like the two tools above, it is not predictive. The underlying theory is that if the market believes one-year rates will be 3.0% and two-year rates 4.0%, it must assume one-year rates twelve months from now will be 5.0%. While mathematically sound, this logic has little to no relationship to what rates will actually be a year from now.

I examined one-year forward rates from 2020 through 2024 and compared them to actual one-year rates from 2021 through 2025. As shown to the right, when rates were rising, forward rates lagged actuals; when rates fell, actuals declined faster than forward rates. In today’s relatively stable environment (at least on the long end of the curve) the two have converged somewhat. Forward rates help illustrate the market’s implied expectations for future interest rates based on current conditions, but they do not enhance predictability of portfolio behavior.

The Takeaway

If the goal of modeling is to predict borrower behavior, these three tools offer little help. What’s needed instead is an empirically grounded understanding of how borrowers actually behave under various economic scenarios.

Level1Analytics’ monthly analysis of over 30 million loans provides precisely this insight. Why guess—when you can be right?

Explore Our Valuation Solutions.