In agency MBS, the specified pool market prices high-LTV loans at a pay-up over TBA for the prepayment protection they offer. This relationship has been well established for both high-LTV purchase loans as well as MHA (Making Home Affordable–i.e., modification and refi programs for troubled loans) production. But what about low-LTV loans? In this post, we uncover counterintuitive behavior showing that low-LTV loans also pay slower than their generic counterparts. This relationship holds even after accounting for other factors such as loan size, FICO, loan purpose, age, and occupancy.
To start, we examine agency loans with original LTV significantly below market average. Using RiskSpan’s Edge Platform, we queried results over the last five years, controlling for vintage (2013+), age (9-60 months), loan balance (225k+), and FICO (720+). We break the results into LTV cohorts, isolating loans 70-80 LTV (orange), 60-70 LTV (green), and LTV less than 50 (blue). In-the-money speeds on the low LTV sector are a full 3 CPR slower than 60-70 LTV and more than 5 CPR slower than 70-80 LTV.
Figure 1: CPR vs Refi Incentive for various LTV cohorts. We removed the 50-60 LTV cohort for ease of reading.
Perhaps this relationship is driven by HPI gains? To test this, we search for loans using current LTV (mark-to-market LTV) as the bucketing criteria. We still see speeds 3 CPR slower for loans with current LTV less than 50% (blue line) versus 50-60% LTV (black), and 5 CPR slower than 70-80% current LTV (orange).
Figure 2: CPR vs Refi Incentive for various current LTV cohorts. To compute current LTV, we use the FHFA all-transactions index at state level to adjust the home value.
The behavior persists when we control for occupancy status. Low-LTV loans pay slower for both owner-occupied (first graph, low LTV in purple) and investor loans (second graph, in blue). Investor loans show less of a speed differential when faced with 100-150bp of refinancing incentive, but there still is a difference of 1-2 CPR.
Figure 3: CPR versus refi incentive for owner-occupied loans. Each line represents a different current-LTV cohort.
Figure 4: CPR versus refi incentive for investor loans.
Finally, when we control for loan purpose and hold other characteristics constant, we see that very low LTV purchase loans pay similarly to their higher LTV brethren (first graph, low LTV blue line).
Figure 5: CPR versus refi incentive for purchase loans. In this graph, the LTV-to-refi relationship breaks down.
For refi loans, very-low-LTV loans (purple line) are once again prepay more slowly than loans with higher LTVs. Loans in the refi-only 0-50 LTV bucket were 6 CPR slower than 50-60% LTV loans when both were faced with 100bp of refi incentive.
Figure 6: CPR vs incentive for refi loans. Very low LTV refi loan show significantly less response to incentive than 70-80 LTV loans.
Counterintuitively, mortgage loans with very low LTV ratios generally prepay more slowly than comparable “normal” LTV loans (I.e., LTVs around 70%) when faced with similar refi incentive, holding other loan characteristics constant. Our analysis shows that loan purpose accounts for the largest differential in prepayment behavior among very-low-LTV loans. Refi loans with LTV < 50% pay significantly slower than “normal” LTV refi loans, whereas LTV does not appear to affect prepay speeds of purchase loans. For other characteristics, we see that very-low-LTV loans pay consistently slower than normal LTV loans. Mortgage data makes us scratch our heads sometimes.
Note: The analysis in this blog post was developed using RiskSpan’s Edge Platform. The RiskSpan Edge Platform is a module-based data management, modeling, and predictive analytics software platform for loans and fixed-income securities. Click here to learn more.