Natural Hazards Index: Dimensioning Physical Risks from Climate Change
How do you define risk when it comes to natural disasters?
Jeffrey Schlegelmilch, Director—National Center for Disaster Preparedness, Columbia University’s Earth Institute: If we really want to understand risk, first we have to understand the hazard. Are you in a floodplain? Are you surrounded by forest that has had longer and drier summers? Are you in an area prone to hurricanes? That’s the hazard. And your exposure is how closely you are situated to that hazard.
Then there’s vulnerability. Just because there is a hazard, and just because it’s near you, it doesn’t necessarily mean that you are incredibly vulnerable to it. It doesn't necessarily equate to high risk. If we’re building with wind-resistant materials, if we’re building in ways that allow water to flow without accumulating in ways that promote flooding, these are things that can reduce the risk. We can control and reduce vulnerability.
And capacity has to do with capacity to cope. Do you have the ability to evacuate? Do you have a car? Do you have a place you can go? To have a higher wind rating, can you afford to replace all the windows in your house? So, capacity starts to get into more of the sociological factors. And risk, of course, is ultimately the aggregation of all this.
What kind of data inform the Natural Hazards Index tool?
Jonathan Sury, Senior Staff Associate—National Center for Disaster Preparedness, Columbia University’s Earth Institute: Looking at all the possible hazards, we landed on 14 different hazards where we could find reasonably good data, as well as available nonproprietary data. These included coastal flood, damaging wind, drought, earthquake, extreme heat, floods, hail, hurricanes, landslides, tornado, tsunami, volcano, wildfire and winter storms. So, the end result is a nonranked, summative index with individual hazard scores.
There are four main buckets in which the data are classified: historical data, based on prior event data; probabilistic or predictive data, providing some percent likelihood; deterministic data—given some set of conditions, what the expected outcome might be; and a model data set, which uses multiple explanatory variables to approximate some outcome of interest.
Version 2.0 of the Natural Hazards Index was released in 2023 as a [US] census-track version, and the data are also kept by AB at the county level and the census-track level. The Natural Hazards Index is not a risk index. It tells us the presence of a hazard and how severe that hazard might be on a scale of one to five for each individual hazard and in summative form. But the tool does not tell us damage or loss estimates, or potential population impact. That’s the next part of the risk equation that AB is going to help fill in.
How can investors use the Natural Hazards Index as a tool?
Patrick O’Connell, Director—Fixed Income Responsible Investing Research, AllianceBernstein: When we think of measuring exposure, vulnerability and capacity, we’ve looked back over historical time frames and asked, What is the damage from these different perils over time? This is data that we’re mining from the National Oceanic and Atmospheric Administration, from universities that have data on all these major perils and the costs to clean them up, to mitigate them, to ameliorate them over time.
A few perils stand out as being higher risk than the rest: tropical cyclones, severe storms, droughts, floods. We wanted to create a framework that homes in on these most costly perils, because we think that they could do more damage to municipalities—cities, states, governments—or properties like hotels, warehouses, whatnot. Because there are different levels of materiality, we wanted to calculate that in a quantitative form looking backward, and then apply that to our logic to create a risk score.
What we’ve built is an approach to measure US counties by what we call adjusted investment risk. This is a combination of the hazards, using Columbia’s Natural Hazards Index, times exposure, vulnerability and capaciy.