What recent research tells us about evacuation decisions during Hurricane Matthew

By: Josh Eachus

A common question in meteorology circles after Hurricane Matthew: why didn’t they evacuate? We can’t just say storm surge will be 7-11 feet. We can’t just say know your evacuation zone. Even combining the two, while a step in the right direction, isn’t enough. Riad et. al (1999) said that evacuation can be understood as the result of three basic social psychological processes: (a) risk perception, (b) social influence, and (c) access to resources. But at some point, it all becomes too much! Say, what? It sounds like paralysis by analysis. But, in striving for the zero fatality outcomes, we continue to weave fundamental concepts of risk perception and information processing into weather messaging.

(Note, as simply a “blog post” this is in no way a complete look at literature on evacuation decisions, risk perception or information processing. Rather, a few newer works have been acknowledged with some widely known and practiced concepts)

I am an ardent supporter of threat and efficacy messaging, a strategy best summarized as “empowering the user.” Several years ago, meteorologists focused messages on threats—wind, rain and storm surge. Social science then blessed the field with a simple question. What if messages were more geared toward the user—them, you, us? Born was a new means of reaching the masses with warning messages to convey protective actions and not just terrifying weather jargon. That counterproductive attempt at instigating fear control processes has been largely quelled in the last half-decade or so, but there continues to be times where “awe for our own science” still gets in the way of inspiring danger response processes.

Let us consider the standard warning framework. It leads with the imposing threat followed by an efficacy message. However, recent research places some nuance in that procedure.

Work done by Morss et al (2016), found that while survey participants receiving extreme impact statements were likely to take action, they also viewed the messages as overblown. The implications from this assessment could be counterproductive. By some anomaly or shift in storm track, should the user find underwhelming impact, we’ve now created a message that to a particular population will be ineffective and potentially not linguistically outdone in the future. If an area isn’t truly “uninhabitable for weeks,” what verbiage is used the next time such a storm threatens with that same possible outcome? Herein lies the call for probabilistic elements to impact statements. But at the same token, if you give a gambling man a 10% chance everything will be OK, do they take that chance? We on the warning side, cannot take that chance.  

Even if the right balance is struck, as in the case of Hurricane Matthew, we saw several examples of people that did heed the efficacy messaging and still weren’t safe. Many maneuvering away from the North Carolina Coast were caught, to their surprise, from inland flooding. Even the most brilliantly crafted warning message can be geographically insular. Is a coastal evacuee going to read the section on heavy rain? Is an inland resident going to read the section on surge? Is either going to tie all of these extreme impact statements to one another? Weller notes, “Although a great deal of research has focused on identifying the characteristics of effective messages, there has been too little focus on the different types of threats of different types of hurricane scenarios.” A broad-brush forecast cannot be issued for the same storm. Our next step (somewhere down the road) is microscale, neighborhood specific forecasts that better personalize the threat on an individual level. Unfortunately, for some, what’s going to happen in County A or Town A isn’t enough. To fully grasp the situation, some need to know what is (or could) happen in Neighborhood A or on Street A.

We’re damned if we do, and damned if we don’t. Indicate significant threat and risk and get accused of hyping. Provide proven protected action only to have a misunderstanding of spatial scale render it useless. The land of weather warnings is a minefield of potentially erroneous messaging.

Now consider this wonderfully detailed hurricane local statement example from NWS Jacksonville. Every possible hurricane threat is considered, detailed and given a recommended action. What more can we ask for? Perhaps the question is what less can we ask for? Is this possibly too detailed? That statement read 1,693 words. For perspective, this long-winded blog post is only 1,478 words.

The Limited Capacity Model (LCM) states that “people have a limited amount of mental resources to process information.” Under the assumption that people are information consumers with a limited capacity to do so, people must make compromises as the amount of data continues overwhelms available attention. So now you’re telling me that we need to say all of these things, but in saying them it is too much for people to understand? Can we say all of that in fewer words? Can we communicate the interwoven impacts these hazards have on one another? Can we begin to assess and mitigate socioeconomic impacts that may deter some from evacuating?

Even beyond all of this, we haven’t even scratched the surface of considerations needed in extreme weather warnings. Morss points out the infinite cultural worldviews to which a message must appeal. One may reject a message if it interferes with their beliefs. As Weller et. al found (2016), many who stayed in Galveston during Hurricane Ike despite evacuation orders, were concerned with “protection of property and protection from looters.” (Looting, by the way, has been widely proven a disaster myth, yet still a real concern for residents) Others cited having sturdy homes, no bad experience in previous storms or simply, didn’t want to sit in traffic after nightmare scenarios from Hurricane Rita. “Insufficient resources” continuously shows up as reason people do not evacuate. Meteorologists, emergency managers and communicators need to be ever-aware of these people and bring that assistance to them when possible.


Reasons to Evacuate, Weller 2016


Reasons to Stay, Weller 2016

Did anyone see the couple interviewed on the Weather Channel prior to Matthew? They stayed behind because they had been through Wisconsin blizzards. We instantly assume that’s just plain stupidity—but not everyone has the background knowledge that comes with a meteorology degree. Surely, there are meteorologists who smoke with doctors rolling their eyes at that risk. Maybe that same doctor decides to ride out a hurricane when a meteorologist told them not to! I literally had a local doctor call me last week and ask if he should cancel his trip to Turks and Caicos—a smart man that just didn’t have expertise to process risk in that arena. The point is, in risk perception, decisions are based on believed success of action over its inconvenience and not everyone thinks of weather risk in the same terms that we do. Morss said, “A more effective approach may be to frame the information in ways that decrease message rejection among certain populations, for example, by presenting information in a way that affirms rather than threatens cultural values and narrative templates.” Can we turn the message into “positive action to avert a risk” rather than “possible threat, protective action?” Just a simple switch in subject and predicate could go a long way. Heck, Morss’ assessment even means race is potentially a determining factor in evacuation decisions. Via 700+ residents affected by Hurricanes Andrew and Hugo, Riad (1999) found that African Americans were much less likely to evacuate to due believing the storm threat was not high enough or homes were strong enough. Now we even have to consider different messages for different segments of the population? Is that even logically feasible?

Let’s summarize some characteristics of an effective warning message based on this discussion:

  1. Clear threat and mitigating action
  2. Strongly worded with minimal hyperbole
  3. Concise
  4. Personalized for the user
  5. Culturally ambiguous (that’s a tough one!)

The considerations are many. Can they all be achieved at once? Of course not. But if you look at warning messages from even ten years ago, the improvements have been vast. We’re slowly but surely factoring each tidbit of risk perception and cognitive processing into our meteorological repertoire. We won’t get there overnight, or even by the next storm, but it gets better each time.

I think thewxsocial.com colleague, Dr. Gina Eosco summed up this issue best by saying via Twitter:

With that in mind, I am not saying the weather warning system is broken. In fact, I’d say the system is growing, improving and even thriving. Let’s keep our heads up and keep learning. What I am really trying to offer is an answer to that common question in our minds: why didn’t they evacuate? The answer is out there. We’ll find it. We just have to know where to look.


Morss, R. E., Demuth, J. L., Lazo, J. K., Dickinson, K., Lazrus, H., & Morrow, B. H. (2016). Understanding Public Hurricane Evacuation Decisions and Responses to Forecast and Warning Messages*. Weather and Forecasting, 31(2), 395–417.

Riad, J. K., Norris, F. H., & Ruback, R. B. (1999). Predicting Evacuation in Two Major Disasters: Risk Perception, Social Influence, and Access to Resources. Journal of Applied Social Psychology, 25(5), 918–934.

Weller, S. et. al. (2016). Should I Stay or Should I Go? Response to the Hurricane Ike Evacuation Order on the Texas Gulf Coast. Nautral Hazards Review, 17(3).

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.