The Risk Communicator: January 2009

by Dr. Robin Dillon-Merrill

Following catastrophic natural disasters, people often ask why individuals did not evacuate and why government and first-responder organizations were so appallingly underprepared. A consistent conclusion of hurricane evacuation studies is that "evacuation behavior is a consequence of the perceptions which people form about risk prior to taking protective action" (Fitzpatrick and Mileti, 1991; emphasis ours). Our research focuses on the linkage between prior experience and perceptions of risk, whereby the degree of success people feel in surviving prior disasters heavily influences their perceptions of the risk involved in these decisions and thus how they respond to warnings of subsequent impending disasters.

When people, by chance, escape prior disasters they have experienced a "near-miss." We define different types of near-misses: 1) a didn't near-miss, wherein a prior near-miss is interpreted as "a failure that did not happen"; versus 2) an almost near-miss, wherein a prior event that did not happen also has some characteristic that leads it to be interpreted as "almost happened." We believe that didn't near-misses will bias individuals towards risky future decisions while almost near-misses should attenuate this bias.

We explore the robustness of this near-miss phenomenon by testing how the two different types of near-misses influence both the general population -- in this case, people who currently live in New Orleans (67 percent of our sample evacuated for Hurricane Katrina) -- and a sample of professional emergency managers, including those solicited in the December 2008 issue of The Risk Communicator.

In the studies, participants read a short statement explaining that they lived in an area subject to hurricanes, and that the National Weather Service was tracking a hurricane that had a 30-percent chance of hitting their community with moderate force within 36 hours. They were further told that they lived alone and had no pets, and that to evacuate they would incur a sure loss of $2,000. If they stayed, however, the collateral damage of a moderate hurricane (above and beyond damage to house, such as car, self, etc.) would add up to $10,000.

Although all participants faced the same dilemma, they varied in some of the life experiences they were assigned to consider.

Participants assigned the didn't near-miss condition read:

You have lived in this house through three prior storms similar to that forecasted and you and your neighbors have never had any property damage.

Participants assigned the weak almost near-miss condition read:

You have lived in your house through three prior storms similar to that forecasted and have never had any property damage. In the last storm, however, a tree fell on your neighbor's car and completely destroyed it.

Participants assigned the strong almost near-miss condition read:

Participants assigned the no near-miss information (control) condition read:

You have no specific data regarding past hurricane impacts to your property.

Additionally, after the evacuation task decision was completed, we asked the participants a series of gamble questions, one of which was isomorphic (identical in form) to the hurricane evacuation decision:

You are stuck with a losing gamble. If you take the gamble, you have a 30-percent chance of losing $10,000 and a 70-percent chance of losing nothing ($0). You can avoid the gamble by paying $2,000 up front. Do you accept the gamble or pay the $2,000?

This gamble was hidden among three other unrelated gambles to disguise the similarity of the tasks. We discarded the results from those unrelated gambles.

The graph below shows the percentage of participants in each condition who choose to evacuate, and -- as can be seen from the graph -- participants with didn't near-miss information evacuated significantly less than those in the other conditions:

The results affirm our contention that near-misses can prime a particular mindset. Results show that even a weak almost near-miss can make participants choose the risk-averse option and decide to evacuate despite the sure loss. Moreover, the stronger the almost near-miss, the more likely participants are to choose the risk-averse option and evacuate. Although the expected value for all of these decisions is the same, the context surrounding them influences the choices people make. Specifically, the type of near-miss highlights a particular aspect of the situation that is either positive or negative, which then appears to either prime either complacency or mitigation action.

In addition, our results show that people with didn't near-miss information make riskier choices than those without this information, because thenear-miss events lead them to perceive a lower level of risk regarding the decision situation. The interpretation of near-misses as evidence of a system's resiliency means that too many potential failures go under-diagnosed, and few will invest in mitigation or protection actions the next time.

Our results also show that almost near-miss information (i.e., reminding people of a system's vulnerability) can push people back toward decisions with lower risk. This is an important finding for organizations that must communicate risks to the public and invest in protective actions.