Many can’t forget the foreboding satellite images of Hurricane Katrina in the hours before it made landfall on America’s southeastern coastline. The devastation that followed left 80 percent of New Orleans inundated, forced tens of thousands of residents to take refuge on rooftops and in sports stadiums, and lead to $135 billion in total damages.
In the aftermath of Katrina, amid feelings of sympathy were those of bewilderment—why did government fail to prepare for a disaster that was fairly predictable? After all, hurricanes are frequent along the Gulf Coast. Storm planners participated in a mock hurricane training exercise just the year prior, an exercise that highlighted, for example, the number of residents needing to be evacuated in the event of a levee breach in the New Orleans area. As one former emergency official remarked at the time, “What troubles me the most is the fact that they knew the potential impact, the potential loss of life, knew how many people would be stranded … and they did not use every resource humanly possible to get people out of the city.”
Even though media outlets reported on the possible impact days before the hurricane hit, a substantial number of residents decided not to evacuate, and officials made little effort to help those who lacked the resources to do so.
Destructive storm events continue to pummel the Gulf Coast, and estimates from the Census Bureau show that between 2000 and 2016, population levels along this stretch of coastline have risen by 24.5 percent, compared to 14.8 percent for the United States as a whole. Why, after such devastation, would people continue to migrate to an area that puts them at risk?
Disaster risk assessment models usually overlook the difficulties people have in processing low probabilities.
Behavioral economics can help explain why we misperceive the threat posed by low-probability/high-impact natural disaster risks, and why individuals and governments underprepare for their impacts. In a series of experiments that led to the Nobel Prize–winning prospect theory, Daniel Kahneman and Amos Tversky noted that people struggle to evaluate low-probability events. The result: people tend to either greatly overweight these probabilities or neglect them altogether.
Despite these insights, disaster risk assessment models, which are used to estimate monetary damages from disasters, usually overlook the difficulties people have in processing low probabilities. A recent article in Nature Climate Change highlighted that these models rarely integrate realistic societal behavior, like people’s proneness to heuristic processes, which may influence the weight people place on low-probability events.
For instance, the threshold model heuristic implies that people ignore the probability of disaster risks if the probability is judged to fall below a threshold level of concern. In addition, the availability heuristic states that people assess the likelihood of an event by the ease with which similar examples come to mind. Apart from the use of heuristics, homeowners sometimes expect to receive post-disaster government compensation, which may be used as a substitute for private insurance and protection measures, a problem which has been termed the charity hazard.
Among the most destructive hurricane events in recent years, such as Hurricane Sandy in 2012 and Hurricane Harvey in 2017, estimates show that only around 20 percent of homeowners who were most directly hit were insured against flooding, presumably because many judged the flood probability to be lower than their threshold level of concern. And while experiencing a flood and media coverage after the event inspires people to purchase flood insurance, coverage levels drop after a few years without homeowners making a claim. It seems that experience with a flood triggers people to believe that another is likely to occur in the near future, consistent with the availability heuristic. But memories fade over time, and risk perceptions fall back below threshold levels of concern.
Furthermore, while people may anticipate being compensated by the government for damages suffered after a flood, in practice, relief payments are very low and depend on political motives. That is, disaster expenditures are usually influenced by election years and states considered important to the outcome of elections, such as swing states.
It is remarkable that most disaster risk assessments consider human behavior as constant over time, completely unaffected by individual natural disaster experience and/or the uncertain incomplete receipt of government compensation. The result of neglecting realistic human behavior is that flood risks are misrepresented, and policymakers do not have accurate information on which to base their adaptation strategies.
It is remarkable that most disaster risk assessments consider human behavior as constant over time.
One way of incorporating human behavior into disaster risk assessment is the so-called agent-based-modeling approach. Agent-based models can be used to estimate more realistic flood risks by simulating individual and government behavior over time, under alternative assumptions about the way in which we as a society adapt to disasters. That is, agent-based models can accommodate boundedly rational behavior and interactions between agents (e.g., social learning).
In an agent-based-modeling framework, assuming that individuals do not adapt to risk over time, even in a boundedly rational manner with heuristic processing, leads to the overestimation of future monetary damage costs relative to a situation in which measures are implemented. On the other hand, modeling adaptive behavior without accounting for the impact of potential government compensation and charity hazard would likely underestimate these costs.
Agent-based-modeling is a step in the right direction for generating more realistic future disaster damage cost estimates. A more accurate understanding of human behavior and its interaction with flood risk would facilitate government investment in flood protection, the development of building codes, and the setting of risk-based flood insurance premiums. Climate change is projected to increase the frequency and intensity of natural disaster risks in the future, and these developments promise to enhance our ability to cope with these immense challenges.