Different decisions depending on whether you are tactical or strategic

This article, Evacuation and the Danger of Mindless Obedience by Tom Mullen, is a little reflexively anti-government for my taste but it does touch on an important challenge which is not often articulated.

That challenge is the difficulty of making tactical vs. strategic trade-off decisions in complex dynamic systems. In this instance, Mullen is questioning the wisdom of mass evacuations at the approach of an hurricane. He starts with a framing anecdote:

I relocated to Tampa, FL in 2004 for a job opportunity. I moved into my apartment on Saturday, July 31 and started work the following Monday. One week later, the local news was dominated by Hurricane Charley, which was predicted to be the first hurricane to directly hit Tampa since 1921. Sound familiar?

I was told I was in a mandatory evacuation area, meaning 911 calls might not be answered for anyone who refused to leave. Nevertheless, most of my new, local friends told me to ignore the order. “They always say that,” I was told. “It always ends up being a lot of rain, nothing more.”

Something inside told me they were right. Still, I was new and although I never really feared for my life, I had recurring visions of local first responders pulling me out of the rubble as the media ridiculed the “Yankee who couldn’t follow simple instructions.” So, I played it safe and booked a hotel in Orlando.

On August 13, the hurricane turned right and made landfall in Port Charlotte, about 100 miles south of Tampa. It was a Category 4 when it hit, meaning winds of over 150 mph. It then made a beeline for – you guessed it – Orlando.

It hit the hotel I was staying in that evening as a Category 2, with winds about 106 mph. I watched out my window as the palm trees first bent all the way over in one direction, then straightened up for a while, then bent all the way over in the other direction as the southern end of the storm passed through.

Yes, my well-meaning local government sent me directly into the path of the storm. It was very impressive to watch, but, needless to say, I never evacuated again. After 2005, we had relatively mild hurricane seasons until I moved back to Western New York in 2014.

Mullen then introduces a little data. Referencing Florida's preparation for Hurricane Irma.

The state and local governments ordered 6.5 million people to evacuate their homes and, on September 8, a Washington Post journalist in Gainesville reported what anyone who has ever lived in Florida (including this writer) knew would happen:

But Florida has only two main roads: interstates 95 and 75. They are parking lots, and have been for days. People are sitting in their vehicles, completely stopped on four-lane highways, running out of gas. There are no exits on these roads for scores of miles at a time. Once you get on a Florida highway, you are not getting off.”

Imagine if everyone ordered to evacuate had cooperated and a dangerous hurricane hit one of those highways. The death toll could have been an order of magnitude larger than it was. We know this from experience. Just twelve years ago, 2.5 million people attempted to evacuate Houston before Hurricane Rita made landfall.

In total, some 130 people died in that evacuation, more than have ever perished in a hurricane in the state’s history, with the exception of the 1900 Galveston storm. Of those deaths, about half occurred before the storm hit Texas.

Mullen wants to make the point that people have to decide for themselves what are the appropriate cautions to take and not simply rely on what the government says. Fair enough, I don't disagree.

But as interesting is the nature of decision-making under uncertainty and whether there are well-founded heuristics for evacuation or not.

Mullen references one storm evacuation in which half of the total storm fatalities were associated with the evacuation. He also exempts the Galveston Hurricane of 1900 from his analysis. But of course one should always be skeptical of analyses which drop inconvenient data. Galveston killed some 10,000 people (see Isaac's Storm: A Man, a Time, and the Deadliest Hurricane in History by Erik Larson for an excellent history of the event.)

This begins to make the tactical/strategic trade-off a little clearer. What should your rule of thumb be in terms of whether or not to evacuate a region in the track of a hurricane, especially when hurricane forecasting is extremely imprecise and inaccurate? The example Mullen gives is pertinent where a population evacuated to one location given the forecasted track of the hurricane, only to have the hurricane veer and move over the evacuated population at their new site.

None of these numbers in the following thought experiment are accurate, they are space holders.

Let's grant that when you evacuate some number of people, there will be a corresponding mortality rate as a consequence of the evacuation. There will be car accidents, there will be deaths from moving otherwise incapacitated people, there will be accidental fires, etc. To put an arbitrary number on it, let's say 10 people will die from evacuation related events per million people in excess of their remaining in place without a hurricane striking.

So if your coastal region contains 6 million people on whom you impose a mandatory evacuation order, you know you are in effect imposing a death sentence on some 60 people, and likely the most vulnerable people. If the hurricane veers in one direction away from them, that will be 60 deaths that could have been avoided by staying in place. If the hurricane veers towards them in their new shelter, they will have the hurricane deaths PLUS the evacuation deaths. If there is no evacuation and it strikes the original location, then you have just the hurricane related deaths.

In Galveston in 1900 hurricane forecasting was rudimentary and there was no real evacuation and there was consequently a loss of some 10,000 people. Using that experience alone, it is always going to be worthwhile evacuating, even if hurricanes veer, because strategically, the rare time when a hurricane does hit with high mortality will always exceed the mortality cost of more frequent but far less lethal evacuations.

The complexity of this decision-making is further complicated by a wide range of other complexities, including the fact that over time, building codes and improvements in infrastructure resilience can decrease the mortality rates of hurricanes.

The problem is that there is no good way of determining the balance of implication of loss of life today based on the loss in the past when circumstances were dramatically different.

And nobody in advance can determine whether we can anticipate a Katrina (powerful storm over urban area that is ill-prepared) an Irma (powerful storm over low density areas that are well prepared) or something in between. A central planner with a very long time frame perspective, a strategic perspective, may heuristically choose all the time to evacuate and that will be the right strategy overall in a long time frame consideration. Similarly, for any individual in any one instance of a hurricane, they might rationally and correctly decide to shelter in place given the right circumstances.

A further complication is from an equity perspective. Natural disasters, on average, disproportionately affect lower income people and people in ill-health. This is true whether they evacuate or shelter in place. I don't know that it is the case, but I suspect that under circumstances of evacuation, the disproportion is even greater still. If true, then that means the decision to evacuate means, on average, 10 deaths per million, and should it also be a factor in decision-making that the poor and sick are 9 of those ten deaths in evacuations but only 6 in 10 if they shelter in place?

This is a question without resolution. But it is useful to remember that there are decision-making scenarios which are unavoidable and yet for which there is little useful information and for which the reward/cost consequences are different depending on whether you are a long term centralized strategic decision-maker or whether you are a short term tactical decision-maker.

The decisions are always going to be consequential but they will never be correct because the burden of the decision is variable and because the average (strategic) is never the same as the individual (tactical).