The articles discussed here reflect the print version of The Wall Street Journal, rather than the online version. Links to online versions are provided. The publication dates and some of the contents do not match.
10 October 2018 – Baseball is well known to be a game of statistics. Fans pay as much attention to statistical analysis of performance by both players and teams as they do to action on the field. They hope to use those statistics to indicate how teams and individual players are likely to perform in the future. It’s an intellectual exercise that is half the fun of following the sport.
While baseball statistics are quoted to three decimal places, or one part in a thousand, fans know to ignore the last decimal place, be skeptical of the second decimal place, and recognize that even the first decimal place has limited predictive power. It’s not that these statistics are inaccurate or in any sense meaningless, it’s that they describe a situation that seems predictable, yet is full of surprises.
With 18 players in a game at any given time, a complex set of rules, and at least three players and an umpire involved in the outcome of every pitch, a baseball game is a highly chaotic system. What makes it fun is seeing how this system evolves over time. Fans get involved by trying to predict what will happen next, then quickly seeing if their expectations materialize.
The essence of a chaotic system is conditional unpredictability. That is, the predictability of any result drops more-or-less drastically with time. For baseball, the probability of, say, a player maintaining their batting average is fairly high on a weekly basis, drops drastically on a month-to-month basis, and simply can’t be predicted from year to year.
Folks call that “streakiness,” and it’s one of the hallmarks of mathematical chaos.
Since the 1960s, mathematicians have recognized that weather is also chaotic. You can say with certainty what’s happening right here right now. If you make careful observations and take into account what’s happening at nearby locations, you can be fairly certain what’ll happen an hour from now. What will happen a week from now, however, is a crapshoot.
This drives insurance companies crazy. They want to live in a deterministic world where they can predict their losses far into the future so that they can plan to have cash on hand (loss reserves) to cover them. That works for, say, life insurance. It works poorly for losses do to extreme-weather events.
That’s because weather is chaotic. Predicting catastrophic weather events next year is like predicting Miami Marlins pitcher Drew Steckenrider‘s earned-run-average for the 2019 season.
Laugh out loud.
Notes from 3 October
My BS detector went off big-time when I read an article in this morning’s Wall Street Journal entitled “A Hotter Planet Reprices Risk Around the World.” That headline is BS for many reasons.
Digging into the article turned up the assertion that insurance providers were using deterministic computer models to predict risk of losses due to future catastrophic weather events. The article didn’t say that explicitly. We have to understand a bit about computer modeling to see what’s behind the words they used. Since I’ve been doing that stuff since the 1970s, pulling aside the curtain is fairly easy.
I’ve also covered risk assessment in industrial settings for many years. It’s not done with deterministic models. It’s not even done with traditional mathematics!
The traditional mathematics you learned in grade school uses real numbers. That is numbers with a definite value.
Pi = 3.1415926 ….
We know what Pi is because it’s measurable. It’s the ratio of a circle’s circumference to its diameter.
Measure the circumference. Measure the diameter. Then divide one by the other.
The ancient Egyptians performed the exercise a kazillion times and noticed that, no matter what circle you used, no matter how big it was, whether you drew it on papyrus or scratched it on a rock or laid it out in a crop circle, you always came out with the same number. That number eventually picked up the name “Pi.”
Risk assessment is NOT done with traditional arithmetic using deterministic (real) numbers. It’s done using what’s called “fuzzy logic.”
Fuzzy logic is not like the fuzzy thinking used by newspaper reporters writing about climate change. The “fuzzy” part simply means it uses fuzzy categories like “small,” “medium” and “large” that don’t have precisely defined values.
While computer programs are perfectly capable of dealing with fuzzy logic, they won’t give you the kind of answers cost accountants are prepared to deal with. They won’t tell you that you need a risk-reserve allocation of $5,937,652.37. They’ll tell you something like “lots!”
You can’t take “lots” to the bank.
The next problem is imagining that global climate models could have any possible relationship to catastrophic weather events. Catastrophic weather events are, by definition, catastrophic. To analyze them you need the kind of mathermatics called “catastrophe theory.”
Catastrophe theory is one of the underpinnings of chaos. In Steven Spielberg’s 1993 movie Jurassic Park, the character Ian Malcolm tries to illustrate catastrophe theory with the example of a drop of water rolling off the back of his hand. Whether it drips off to the right or left depends critically on how his hand is tipped. A small change creates an immense difference.
If a ball is balanced at the edge of a table, it can either stay there or drop off, and you can’t predict in advance which will happen.
That’s the thinking behind catastrophe theory.
The same analysis goes into predicting what will happen with a hurricane. As I recall, at the time Hurricane Florence (2018) made landfall, most models predicted it would move south along the Appalachian Ridge. Another group of models predicted it would stall out to the northwest.
When push came to shove, however, it moved northeast.
What actually happened depended critically on a large number of details that were too small to include in the models.
How much money was lost due to storm damage was a result of the result of unpredictable things. (That’s not an editing error. It was really the second order result of a result.) It is a fundamentally unpredictable thing. The best you can do is measure it after the fact.
That brings us to comparing climate-model predictions with observations. We’ve got enough data now to see how climate-model predictions compare with observations on a decades-long timescale. The graph above summarizes results compiled in 2015 by the Cato Institute.
Basically, it shows that, not only did the climate models overestimate the temperature rise from the late 1970s to 2015 by a factor of approximately three, but in the critical last decade, when the computer models predicted a rapid rise, the actual observations showed that it nearly stalled out.
Notice that the divergence between the models and the observations increased with time. As I’ve said, that’s the main hallmark of chaos.
It sure looks like the climate models are batting zero!
I’ve been watching these kinds of results show up since the 1980s. It’s why by the late 1990s I started discounting statements like the WSJ article’s: “A consensus of scientists puts blame substantially on emissios greenhouse gasses from cars, farms and factories.”
I don’t know who those “scientists” might be, but it sounds like they’re assigning blame for an effect that isn’t observed. Real scientists wouldn’t do that. Only politicians would.
Clearly, something is going on, but what it is, what its extent is, and what is causing it is anything but clear.
In the data depicted above, the results from global climate modeling do not look at all like the behavior of a chaotic system. The data from observations, however, do look like what we typically get from a chaotic system. Stuff moves constantly. On short time scales it shows apparent trends. On longer time scales, however, the trends tend to evaporate.
No wonder observers like Steven Pacala, who is Frederick D. Petrie Professor in Ecology and Evolutionary Biology at Princeton University and a board member at Hamilton Insurance Group, Ltd., are led to say (as quoted in the article): “Climate change makes the historical record of extreme weather an unreliable indicator of current risk.”
When you’re dealing with a chaotic system, the longer the record you’re using, the less predictive power it has.
Another point made in the WSJ article that I thought was hilarious involved prediction of hurricanes in the Persian Gulf.
According to the article, “Such cyclones … have never been observed in the Persian Gulf … with new conditions due to warming, some cyclones could enter the Gulf in the future and also form in the Gulf itself.”
This sounds a lot like a tongue-in-cheek comment I once heard from astronomer Carl Sagan about predictions of life on Venus. He pointed out that when astronomers observe Venus, they generally just see a featureless disk. Science fiction writers had developed a chain of inferences that led them from that observation of a featureless disk to imagining total cloud cover, then postulating underlying swamps teeming with life, and culminating with imagining the existence of Venusian dinosaurs.
Observation: “We can see nothing.”
Conclusion: “There are dinosaurs.”
Sagan was pointing out that, though it may make good science fiction, that is bad science.
The WSJ reporters, Bradley Hope and Nicole Friedman, went from “No hurricanes ever seen in the Persian Gulf” to “Future hurricanes in the Persian Gulf” by the same sort of logic.
The kind of specious misinformation represented by the WSJ article confuses folks who have genuine concerns about the environment. Before politicians like Al Gore hijacked the environmental narrative, deflecting it toward climate change, folks paid much more attention to the real environmental issue of pollution.
The one bit of information in the WSJ article that appears prima facie disturbing is contained in the graph at right.
The graph shows actual insurance losses due to catastrophic weather events increasing rapidly over time. The article draws the inference that this trend is caused by human-induced climate change.
That’s quite a stretch, considering that there are obvious alternative explanations for this trend. The most likely alternative is the possibility that folks have been building more stuff in hurricane-prone areas. With more expensive stuff there to get damaged, insurance losses will rise.
Invoking Occam’s Razor (choose the most believable of alternative explanations), we tend to favor the second explanation.
In summary, I conclude that the 3 October article is poor reporting that draws conclusions that are likely false.
Notes from 4 October
Don’t try to find the 3 October WSJ article online. I spent a couple of hours this morning searching for it, and came up empty. The closest I was able to get was a draft version that I found by searching on Bradley Hope’s name. It did not appear on WSJ‘s public site.
Apparently, WSJ‘s editors weren’t any more impressed with the article than I was.
The 4 October issue presents a corroboration of my alternative explanation of the trend in insurance-loss data: it’s due to a build up of expensive real estate in areas prone to catastrophic weather events.
In a half-page expose entitled “Hurricane Costs Grow as Population Shifts,” Kara Dapena reports that, “From 1980 to 2017, counties along the U.S. shoreline that endured hurricane-strength winds from Florence in September experienced a surge in population.”
In the end, this blog posting serves as an illustration of four points I tried to make last month. Specifically, on 19 September I published a post entitled: “Noble Whitefoot or Lying Blackfoot?” in which I laid out four principles to use when trying to determine if the news you’re reading is fake. I’ll list them in reverse of the order I used in last month’s posting, partly to illustrate that there is no set order for them:
- Nobody gets it completely right ̶ In the 3 October WSJ story, the reporters got snookered by the climate-change political lobby. That article, taken at face value, has to be stamped with the label “fake news.”
- Does it make sense to you? ̶ The 3 October fake news article set off my BS detector by making a number of statements that did not make sense to me.
- Comparison shopping for ideas ̶ Assertions in the suspect article contradicted numerous other sources.
- Consider your source ̶ The article’s source (The Wall Street Journal) is one that I normally trust. Otherwise, I likely never would have seen it, since I don’t bother listening to anyone I catch in a lie. My faith in the publication was restored when the next day they featured an article that corrected the misinformation.