How Hurricanes Can Impact Population
Hawaii Dodged a Bullet in Lane, Florence was Bad for the Carolinas, and Now We’ll See What Happens With Michael
UPDATE: I wrote this post after Lane, before Florence. I am now updating it during Michael with an estimate for Florida, immediately below:
You can read the full post below. But for now, here’s my forecast of Florida’s population out to 2070, if Michael hadn’t hit, and now knowing that it did, and about where and how strong. The size of the “hurricane shock” is calibrated based on windspeeds at the county level, and the population share of Florida in the impacted counties.
As you can see, even with a very severe impact assumption, Hurricane Michael’s total effect on long run population is fairly modest. I will also take this moment to note that Florida’s population collapse will be insanely hard and fast when it happens. By the second half of the 21st century, Florida’s population is likely to be dropping like a rock.
Now back to the original post:
Natural disasters can have a big impact on population. This is widely known. Hurricane Katrina had a huge population impact on Louisiana, Hurricanes Irma and Maria have had huge impacts on Puerto Rico.
But not every hurricane has a large population impact. Hurricanes have been hitting American states for generations, and yet population growth has carried on apace. Indeed, the average disaster has fairly modest impacts on population growth. Economic research suggests that an additional disaster only lowers population growth by an insignificant amount, unless it is a very severe disaster.
This paper is interesting because it tells us several things about disaster migration:
- Hurricanes have larger effects than other disasters
- Rural areas tend to have bigger migration responses than urban ones
- A severe hurricane usually raises outmigration rates over the ensuing decade by 3 to 12% of the starting population.
The key thing to understand here is that hurricanes Lane and Florence both had/have the potential to be really optimally-structured disasters for causing maximum migration responses. Thankfully, Lane turned south of most of Hawaii, and did not cause catastrophic damage (although there was one fatality). Florence looks set to hit North and South Carolina smack-on with category 4 or 5 winds. The area to be hit is disproportionately rural; in Hawaii, most of the islands projected to be severely affected (and most places which were in reality damaged) were more rural islands. In the Carolinas, there are some cities like Myrtle Beach, Wilmington, and Jacksonville, but they are not extremely large, and most of the area projected to have the worst winds is rural.
I have built cohort-component models of the population of Hawaii and North Carolina. These are very similar to the models I have built for Puerto Rico though, of course, not precisely identical. We can use these models to get an idea of what kind of effect Hurricane Lane might have had if it had made landfall, and what kind of effect Hurricane Florence very likely could have when it does slam into the Carolinas.
Let’s start with Hawaii.
A Very Bad Year
Here’s a map of the Pacific hurricane season thus far this year:
That one that goes way out to the left and swoops close to Hawaii then turns away is Hurricane Lane. They got some serious rains and winds, but not landfall and the attendant destruction. All told, Lane probably did under $50 million in total damages to Hawaii.
Of course, Lane came in addition to another disaster earlier this year. The lower Puna eruptions have probably caused $800 million in damages, and resulted in thousands of people being evacuated and hundreds of homes destroyed. These eruptions impacted the same island that was hardest hit by Hurricane Lane later in the year.
And now, another storm is approaching Hawaii. Olivia topped out at category 4, but is now just a tropical storm. However, it will pass straight through the Hawaiian islands and undoubtedly drop a lot of rain, even as recovery efforts from Lane are incomplete.
In other words, Hawaii has had a very rough second half of 2018, particularly the south-eastern half of the archipelago.
So what effect might that have? Well, I want to be clear that these estimates are illustrative. It is extremely difficult to tell the exact scale of a disaster’s impact on migration in advance, because many factors impacting likelihood of migration are idiosyncratic. But here are my estimates of Hawaii’s likely population future, and what it might have been if Lane had made landfall around Honolulu.
The first thing to notice here is that Hawaii’s population is already in a decline period, as I have discussed at length before. However, Hawaii’s fertility and mortality situation is considerably less dire than Puerto Rico’s, as is its migration situation, and thus its decline is not likely to become permanent yet. If I continued this forecast out to 2100, the decline would become steady and you’d get a more Puerto-Rico-shaped population line.
Hurricane Lane would have had a severe impact; Olivia still could. However, Hawaii would eventually recover from a severe hurricane impact. Fertility is still high enough and mortality low enough that, even with a catastrophic disaster, population would still be hitting record highs sometime in the 2030s.
With that, let’s turn to North Carolina.
What the Bounce-Back Looks Like
We can turn to North Carolina next for a very different example of what a hurricane impact might look like.
Hurricane Florence is shaping up to be an extremely severe disaster. I don’t want to say otherwise. However, North Carolina is a big state with a lot of people. Most of its big cities are inland, far from the most severe effects of the hurricane. It is receiving robust net domestic migration and international migration. Furthermore, the impacted area, while large, is much smaller compared to North Carolina’s total population that Hurricane Lane vs. Hawaii or Irma/Maria vs. Puerto Rico. In other words, Florence’s effects on North Carolina’s population are almost certainly going to be far more muted than Lane’s on Hawaii or Maria’s on Puerto Rico. Plus, islands have more volatile population to begin with, as short-range relocations with low return costs are much more difficult.
SO with all those caveats… here’s North Carolina:
I made the lines skinny so you can see the difference. It is not big.
Now, I might be totally wrong. Maybe Florence will cause a much bigger disruption than I’m projecting. I do have it worsening net migration into North Carolina by about 45,000 people, which is actually quite a lot since the whole coastal North Carolina area has under 1.5 million people in it. Obviously there will be damage in non-coastal areas as well, but it is likely to be less demographically severe.
But honestly, even if you ramp up the effect massively, the core conclusion doesn’t change. Make the marginal effect a loss of 100,000 people. Make it 200,000. Make it 300,000!
North Carolina has a lot of people in it, and people are still going to move there! The reasons to move to North Carolina are still strong, these types of hurricanes are still uncommon, and North Carolina’s fastest-growing areas are quite far inland. From 2017 to 2027, North Carolina will probably add a million people in the baseline specification. Reducing that by a few hundred thousand is a big deal…. but also isn’t going to seriously throw off the growth trajectory.
In other words, a hurricane creates a radically different “revision of expectations” about a place based on the pre-existing trend, and the size of the place. Bigger areas are more resilient, and areas that are already growing quickly can take some hits without a person (or a business…) needing to make a serious structural re-calibration of their economic expectations.
That’s all for now, folks!