Hurricane Mortgage Losses Are Surprisingly Low

We selected three high-default quarters following Ian, split between path (red) and non-path (blue). These are compared to statewide benchmarks: hurricane-free/crisis-free periods (green) and financial crisis (orange). While Hurricane Irma (pink) and COVID (purple) appear in the time series, both are excluded from the our risk-adjusted analysis in following sections.

Comparing to other Florida events adds perspective: despite being the costliest hurricane in Florida’s history, Irma’s default rates (pink) were 4x higher than Ian’s (blue). Ian’s path rates were modest at 1.3x hurricane-free/crisis-free averages and roughly one-fourth of crisis-period rates

Ian exhibits the expected pattern: the red spike corresponds to higher path default rates with blue non-path rates tracking green hurricane-free/crisis-free levels. However, this comparison is incomplete because it does not control for risk.

While risk can be measured across multiple dimensions such as FICO, LTV, and DTI, updated values aren’t readily available for Fannie loans. We developed and used the “Spero Portfolio Risk Measure” —a proprietary metric that is robust, intuitive, and theoretically sound. Constructed on a loan-by-loan basis and updated each quarter, we then aggregated similar observations to make our comparisons on a risk-adjusted basis. Our measure provides comprehensive portfolio or sub-portfolio assessments. Our methodology is transparent and explainable—never a black box solution.

We’ve applied similar analytical frameworks to develop effective and intuitive loss forecasting (CECL and stress testing) models using both internal bank data and external data sources, delivering modeling solutions that are both powerful and interpretable.

When we control for risk using the Spero Portfolio Risk Measure, the differences in default rates become more pronounced.

LGDs show almost the reverse pattern, but the differences aren’t as great as with default rates.

LGDs during Ian are similar whether they occur inside or outside the path, but both are uniformly lower than hurricane-free LGDs. Thus, while Ian-related default rates appear to be higher than the other groups, LGDs during normal times appear to be approximately twice the Ian-related LGDs.

We would argue that the difference in hurricane-related LGDs could be attributed to borrowers receiving insurance claims or FEMA assistance, or these could represent strategic defaults or payment delays, since such delays were permitted during the recovery period.

So, when higher default rates and very low (almost no) LGDs are multiplied together, what happens to charge-off rates? The resulting net charge-off rate becomes statistically indistinguishable from zero. In fact, Ians NCO’s are 8x smaller than the hurricane-free/crisis free average NCO (we do recognize both are very small, avg. nco rate for the path and hurricane-free/crisis-free being 0.08 bps and 0.75 bps respectively.)

The analysis reveals a pattern we’ve found characteristic of post-crisis storms: a sharp, but temporary, spike in delinquencies for loans directly in the storm’s path, with little effect on LGD’s. We suspect this is due to a mature regulatory and support framework, including federally-backed forbearance programs that carry customers through until the National Flood Insurance Program and FEMA assistance is available.

We identified three quarters with abnormally high default rates and compared four categories: hurricane path (red), non-path areas during those quarters (blue), Texas-wide hurricane-free/crisis-free periods (green), and Texas-wide financial crisis period (orange). Red and blue correspond to the previous graph.

Consistent with most storms analyzed, post-hurricane (in path) default rates far exceeded financial crisis levels—in this case, 7x higher. Additionally, non-path default rates closely match hurricane-free/crisis-free periods, confirming the localized nature of the hurricane impact.

While risk can be measured across dimensions such as FICO, LTV, and DTI, updated values aren’t readily available for Fannie loans. We developed and used the “Spero Portfolio Risk Measure.” It is a proprietary metric that is robust, intuitive, and theoretically sound. Constructed on a loan-by-loan basis and updated each quarter, we then aggregated similar observations to make our comparisons on a risk-adjusted basis. Our measure provides comprehensive portfolio or sub-portfolio assessments. Our methodology is transparent and explainable—never a black box solution.

For clients, our methodology is transparent and explainable—never a black box solution. We’ve applied similar analytical frameworks to develop effective and intuitive loss forecasting (CECL and stress testing) models using both internal bank data and external data sources, delivering modeling solutions that are both powerful and interpretable.

Notice that quarterly default rates within the storm’s path are substantially higher than those outside the path, while those outside of the path are similar to default rates in hurricane-free times. In fact, average quarterly default rate within Harvey’s path (~97 bps) experienced 7x higher hurricane-free/crisis-free times (roughly 13.6 bps) and 4x over crisis rates.

Comparing this to the storm that followed less than a month after, Irma exhibited the same 7x larger versus hurricane-free/crisis-free times but diverged in its crisis comparison (1.2x), a difference explained by baseline conditions: Florida’s crisis was substantially more severe (5x hurricane-free/crisis-free rates) than the crisis context surrounding Harvey (1.7x hurricane-free/crisis-free rates).

Geez, that makes the storm look very bad, but what about LGDs?

LGDs show almost the inverse pattern, but the differences aren’t as great as with default rates.

LGDs during Harvey are similar whether they occur inside or outside the path, but both are uniformly lower than hurricane-free LGDs. Thus, while Harvey-related default rates appear to be three-to-five times higher than the other groups, LGDs during normal times appear to be approximately twice the Harvey-related LGDs.

We would argue that the difference in hurricane-related LGDs could be attributed to borrowers receiving insurance claims or FEMA assistance, or these could represent strategic defaults or payment delays, since such delays were permitted during the recovery period.

So, when default rates and LGDs are multiplied together, what happens to charge-off rates? The hurricane path experienced net losses of approximately 0.69 basis points—modestly higher than Irma (0.2 bps) and Ian (0.08 bps), where flooding was contained to traditional flood zones. However, context matters: path charge-offs were actually 0.08 bps lower than hurricane-free/crisis-free periods and dramatically lower than the crisis-period rate of 2.28 bps.

It is worth noting that by our measure of risk, the riskiest observations arose in relatively normal weather conditions—the right side of the graph and do correspond to the highest observed default rates on the vertical axis.

What do we make of this? Depending on your portfolio and loan concentrations, that small difference could justify a tiny CECL overlay for a few quarters after a storm.

Like we did with Texas, we plot default rates by time and category, this time for Florida. However, unlike Harvey, we observe higher default rates outside of the defined path. Perhaps we should have selected more counties as being within the path, but we have proceeded with our original classification (from a previous version of this article) for consistency in the analysis.

Those in our path comprise 40% of all observations in Florida during Irma, while observations in our Harvey path only contained only 30% of Texas observations in the Fannie Mae Resi dataset. This provides an indication of the challenge of differentiating path vs. non-path in FL compared to TX. Moreover, as you’ll see, it doesn’t really matter in this case.

We found it interesting that default rates for Irmas path are the same magnitiude larger when compared to hurricane-free/crisis-free times as Hurricane Harvey (7x larger), but the same can not be said for a crisis comparision. Irmas path default rates are only ~1.2x crisis default rates in Florida (meaning that hurricane-free/crisis free times were about 1/5 the default rate of average crisis default rates), versus Harveys 4x crisis default rates in Texas (where the crisis default rates were ~1.75x hurricane-free/crisis-free rates.)

Consistent with the above time series, we see that path default rates uniformly exceed non-path rates, which uniformly exceed hurricane-free rates, when adjusted for risk.

Notice the nice, increasing, convex function formed by the green dots; that’s why we like our Spero Portfolio Risk Measure (see next paragraph). Also notice that if observations in the red and blue dots were grouped with the green dots, the relationship would be weaker, and that’s why we control for them. Average default rates for the hurricane path were 7x’s hurricane-free/crisis-free defaults, larger than crisis period default rates, which were 5.7x’s hurricane-free/crisis-free defaults.

While risk can be measured across multiple dimensions such as FICO, LTV, and DTI, here, so as not to give away our secrets, we’re using what we’ll call the “Spero Portfolio Risk Measure.” Our methodology is transparent with clients – we do not provide “black box solutions” Click on “methodology” to learn more.

As with Harvey in TX, with Irma in Florida we see that when we control for risk, LGDs in hurricane-free times exceed path and non-path LGDs. In fact, LGDs in Florida in the quarters immediately after Irma are basically zero.

Well, regardless of how high default rates are—even if they are 100%—when you multiply them by LGDs of zero, it should be little surprise that overall loss rates are… basically zero. Here, when risk is controlled, path, non-path, and non-hurricane observations are indistinguishable from each other: almost nothing is lost until risk increases. Hurricane NCO’s are 0.19x hurricane-free/crisis-free times, and 0.01x crisis times (whereas crisis NCOs are nearly 25x hurricane-free/crisis-free periods.)

Sandy’s effect on default rates shows patterns consistent with a post-crisis hurricane event. The analysis focuses on Long Island, where the storm’s effects were most concentrated. Default rates from the storm were about 2x default rates in hurricane-free/crisis-free times.

As with other analyses, we use the Spero Portfolio Risk Measure to control for risk differences across portfolios and time periods.

When adjusted for risk, quarterly default rates within Sandy’s path show elevation compared to non-path areas. The green dots representing hurricane-free times demonstrate the expected convex relationship with our risk measure.

Sandy occurred in the end-of-crisis era, so we expect patterns more consistent with Harvey and Irma than with Katrina. However, we know that NY was still recovering from the crisis – so we may not see as large of a difference in LGD’s hurricane event and hurricane free times. Hurricane path default rates were ~1.8x higher than hurricane-free/crisis-free periods.

Consistent with other post-crisis hurricanes, and despite the timing and generally lower number of observations when compared with other hurricane analyses, LGDs during Sandy show lower levels in the storm path compared to hurricane-free periods. This likely reflects insurance payouts, FEMA assistance, and forbearance programs that helped affected borrowers.

The pattern of reduced LGDs during hurricane periods appears to be a consistent feature of post-crisis storm responses.

When default rates and LGDs are combined, Sandy’s overall charge-off impact appears minimal when properly adjusted for risk. The net effect shows patterns consistent with other post-crisis hurricanes, with overall losses remaining very low – only 1.2x larger than hurricane-free/crisis-free periods. Because Sandy happened towards the end of the crisis, we do not include crisis comparisons like we do for other weather events.

This reinforces our hypothesis that post-crisis regulatory reforms and improved loss mitigation lowered hurricane-related mortgage losses.

Unlike many other events in our sample, Katrina occurred several years before the Financial Crisis and toward the beginning of our mortgage data history. Below we see peak default rates for Katrina to be immensely higher than the post crisis storms. We believe the relatively modest increase in Louisiana’s default rates during the crisis period, compared to the hurricane-free baseline, can be attributed to both storm damage and substantial population decline (New Orleans’s city population fell by an estimated 50% and metro fell by about 25%.) Default rates over the hurricane period were, on average, atleast 7x higher than the other events investigated (and, 37x higher than Hurricane Ian.)

Comparing average Lousina default rates shows a 3x increase from hurricane-free/crisis-free to crisis periods, which is tiny when compared to the 40x jump between hurricane-free/crisis free and the hurricane path.

Note the grey bars in 2016. We noted the spike in default rates while investigating Hurricane Katrina, and identified a flood that occured in 2016 in Louisiana.

We see that default rates for mortgages in the path greatly exceed non-path rates, which are barely higher than hurricane-free rates.

Like Harvey and Irma, we see a fairly strong (increasing) relationship for default rates as a function of risk for mortgages in the path. What’s striking here—compared to Texas or Florida—is that default rates for hurricane-free times don’t spike as dramatically as the risk measure increases. That’s partly a scaling effect, but it is also likely due to the fact that Katrina hit before the crisis, and Louisiana did not see the frothy, bubbly housing market that other Sunbelt states saw.

As with the other hurricanes, we see that both hurricane/crisis-free and crisis-related LGDs show a sharper increase as our risk measure peaks, and exceed both path and non-path LGDs.

Among non-Katrina events, Hurricane Sandy produced the largest NCO differential between hurricane and hurricane free/crisis free times at 1.18x, while other hurricanes exhibited no increase or lower NCOs. Katrina’s impact was substantially greater: hurricane NCOs exceeded hurricane-free/crisis-free NCOs by about 6.5x and crisis NCOs by about 2x.

The question arises: for a future hurricane, should one anticipate higher losses similar to Katrina, or much smaller differences like those observed for Harvey and Irma? (There was virtually no difference for Irma.) We would argue that following the financial crisis, both underwriting and servicing standards became stricter—both at Fannie Mae and at banks—so the 2017 storms are more representative of potential losses than the 2005 storm.

Below we see the a similar (slightly smaller) spike to Katrina in Louisiana. Like Irma and Katrina in LA, we see the blue non-path rates are a bit higher than the default rates in green, contiguous hurricane-free/crisis-free times; so, perhaps we should have widened the hurricane path – but we think the main affected areas are well covered.

Hurricane-impacted areas in Mississippi experienced default rates 23.5x higher than hurricane-free/crisis-free periods and a little over 12x the crisis period alone.

We come to the same conclusions we did for Katrina in Louisiana. Like Harvey and Irma, we see a fairly strong (increasing) relationship for default rates as a function of risk for mortgages in the path. Again, what is striking here compared to Texas or Florida—is that default rates for hurricane-free/crisis-free times don’t spike as dramatically as the risk measure increases. That’s partly a scaling effect, but it is also likely due to the fact that Katrina hit before the crisis, and Mississippi did not see the frothy, bubbly housing market that other Sunbelt states saw.

LGDs in hurricane path areas were lower than in hurricane-free/crisis-free, crisis, and non-path periods—consistent with patterns observed in other analyses. Katrina’s unique characteristic was its extraordinarily high default rate, meaning that even with lower LGDs, net charge-offs remained elevated.

We can ask the same question for Katrina in Mississippi that we did in Louisiana: for a future hurricane, should one anticipate higher losses, closer to Katrina, or much smaller differences, like those for Harvey and Irma? (There was absolutely no difference for Irma.) Katrina’s impact in MS was substantial: hurricane NCOs exceeded hurricane-free/crisis-free NCOs by about 7x and crisis NCOs by 1.5x.

As with the hurricanes, we see that default rates for mortgages in the path greatly exceed those not in the path, which are about the same as observation in hurricane-and-flood-free times.

Within the flood zone, default rates are non-monotonic with our generic risk measure because of the red point to the left. We still see the anticipated convex relationship between our risk measure and defaults during “normal times.”

Average default rates for path areas exceed non-path areas by 3-4x in basis points. Visual inspection of the scatter plot corroborates this finding, with flood zone default rates approximately 3x higher than the remaining categories on a risk-adjusted basis.

Once again, we see that LGDs during normal weather substantially exceed those in the flood zone—a cursory glance says by about a factor of three when adjusted for risk.

The 2016 flood caused a severe spike in defaults (3x normal), demonstrating it was a genuinely damaging event. However, the outcome could not have been more different from Katrina. The combination of higher-quality underlying mortgages and effective loss mitigation completely offset the rise in defaults. LGDs were suppressed so effectively that the final net loss impact was zero. Flood NCO’s were one-third of rates during the crisis in Louisiana, a stark comparsion to what Katrina was – alittle over double rates during the crisis.