"Noise" and Why It Matters
Jun 9, 2021

My grandfather used to tell a story about old Sven, a carpenter who worked for him for years. Sven needed watching. His use of the carpenter's square and the plumb bob could be a bit irregular as the day wore on, but no one knew what Sven's problem was - until one morning he reported for work sober. A new book by Daniel Kahneman is always a reason to sit up and pay attention, but his latest, Noise, a Flaw in Human Judgement*, is especially important because it describes in detail a pervasive problem that we are all so used to that, like Sven's bottle issues, we often lose track of how dangerous and debilitating it is over time - at least until a Nobel prize winning guy like Kahneman rubs our noses in it and makes us accept the pain of hard, slow thinking.

All we really do in risk management is make decisions. We don't crochet pot holders or build bird houses. The only meaningful end product of our work day is making decisions about identifying, quantifying and managing risks. The quality of our work is the sum total of the quality of our decisions. "Noise" is Kahneman's term for all of the subtle, often hidden but powerful factors that muddle and subvert our decisions. Noise is the unwanted, unwarranted variability that makes our numbers and conclusions suspect.

Let's take an example of flawed judgment which Kahneman draws from the insurance industry (imagine that). Take a large group of expert P&C underwriters and task them with underwriting the same series of fictional risks. They all have the same data set. What do we think will be the range of difference in their responses? We know they will not all come to exactly the same pricing offer, but they should all be close, right? These are seasoned experts working from the same initial input. The carrier executives working with Kahneman's research group (KSS) estimated the median difference at 10%. The actual number? 55%. A similar test using claim reserving gave a spread of 43%. Not. Real. Close.

That difference is noise. The book is a 400 page dissection of how noise permeates our decision processes, usually while we are blissfully ignorant of its extent. Two different examples from the review in Business Insurance:


The rate of false negatives by radiologists screening mammograms for breast cancer ranges from zero to more than 50 percent. The variance in employee performance appraisals attributed to noise having nothing to do with actual performance is 70 to 80 percent. The probability that two interviewers will produce similar ratings after meeting the same job candidate is 62 to 65 percent.

The insurance business clearly has no corner on noisy decisions.

Every risk manager should read Noise because it explores the sources of noise in various types of decision processes and discusses ideas for "decision hygiene" which can minimize noise and the often expensive errors it can carry with it. It also distinguishes noise from bias, the other major component of poor decision making and discusses in detail why we make these errors over and over and how we can reduce noisy and biased thinking.

Poor decision making is expensive - and many times embarrassing to boot, like the risk consultants who convinced a major Northern California institution that it didn't really need the earthquake insurance it had carried since 1906. As a result of this expert advice, the risk management group cancelled the policy - four months before the Loma Prieta earthquake caused some $160,000,000 in now uninsured earthquake damage. Oops.

As it turned out, there were reasons, avoidable reasons, for Sven's occasional misadventures with the carpenter's square, but all he was doing was cutting and setting rafters. We deal with larger pieces of money and equipment, and with injury risks as well. How much noise can we wring out of our decisions if we pay attention to the decision tools Kahneman gives us? It's a long book - and worth every page.

*Find it on Amazon.



COVID and Robots

Michio Kaku, everyone's favorite physicist, recently gave us this little glimpse of days to come: "The job market of the future will consist of those jobs that robots cannot perform. Our blue-collar work is pattern recognition, making sense of what you see. Gardeners will still have jobs because every garden is different." New, post-COVID trends in robot acquisition seem to be validating his statement.

Robot orders (US) in the first quarter of 2021 rose 20% over the same period in 2020. After social distancing measures forced layoffs in labor-intensive factories, manufacturers turned to automation, despite the high cost. Now, they aren't going back, according to NBC News. Historically, carmakers have been the big users of robots, but that's changing rapidly. The biggest increases in robot orders for the first quarter of 2021 were in the metals industry - up a whopping 86% year over year - life sciences and pharmaceuticals – up 72% - and food and consumer goods - up 32%.

What about your industry? Obviously, you've been managing some degree of robot related risks for some time now, but are you fully up to speed on new process risks as robots begin to fill more and more links in your production lines, do ever more of your customer fulfillment tasks, trigger inventory actions, and so forth? Is the risk of one robot doing spot welds on a subcomponent the same as a series of robots working an assembly line, end to end? Does business interruption risk change in any way? What about cyber risks or risks associated with the ecosystem of vendors who keep your robots humming?

The headlong adoption of robotic solutions to a myriad of business task requirements is just one more way in which COVID is turbocharging what had been a more stately and deliberate process of wholesale change*. Are we keeping pace over here in our bailiwick in understanding and managing the associated risks? If you need help in risk management - hire a human while they're still available.

*Robo-pets are even taking over the tasks of dogs and cats as companions for our elderly according to the New York Times. They seem to be doing a surprisingly good job.

So how do you feel about that, fellow risk managers?


Quick Take 1:
The More Things Seem to Change...

We've had this conversation before, only the numbers have changed:


The proportion of U.S. workers who tested positive for marijuana in urine climbed higher in 2020 while the overall share of positive drug tests plateaued last year, according to Quest Diagnostics Inc., one of the largest drug-testing laboratories in the U.S. About 2.7% of the approximately seven million drug tests Quest conducted on behalf of employers came back positive for marijuana - up from 2.5% in 2019 and 2% in 2016.

What does this mean at a time when more and more states have legalized medicinal and/or recreational use of marijuana? Should this be on the scope of your WorryRadar*?

Some industries have all but stopped testing for marijuana in recent years. Among hospitality and restaurant industry workers who were tested for drugs last year, 6.3% were positive for marijuana - one of the highest rates for any industry, according to the Quest data. Now that restaurants and hospitality venues are reopening or returning to full capacity and many appear to be scratching to find new employees, even less testing seems likely to be in the cards. But the Federal Department of Transportation spent part of last year making their new drug and alcohol testing database for the trucking industry operational. This new database has significantly contributed to a hiring crunch, said Avery Vise, a trucking analyst at FTR Transportation Intelligence. Since its launch, nearly 70,000 truck drivers have been added to the database for positive tests (drug testing is mandatory for truck drivers), complicating new driver hiring for many transport concerns.

Where does your industry sit on the continuum between hospitality and transport? Has COVID given you reasons to rethink drug usage policies, or to reiterate your pre-COVID restrictions? The specific issue of marijuana use by employees, given both the changes in state laws and the changes wrought by COVID itself may need reviewing sooner rather than later as you look to re-staff.

The Black Death changed the face of Europe in the Fourteenth Century in many fundamental ways. Some historians of economics credit the plague with leading directly to the development of modern capitalism by changing the nature of the employer-employee relationship. While COVID lacks the spectacular mortality of Yersinia pestis**, it attacks many fundamental ideas of how our economy works in 2021. Take nothing for granted.

*WorryRadar, pat pending, the app for every risk manager
**The plague bacillus

The Black Death, changing everything, one victim at a time.


Quick Take 2:
Does COVID Have a Backdoor?

When the COVID threat first began to take shape, most of the risk management community thought first of just two lines of coverage - workers' comp and business interruption. While COVID certainly has racked up some loss and litigation numbers with both, a different coverage is beginning to emerge as another potential lightening rod for COVID induced losses - D&O.

A new report by Advisen tells us: "Securities class action and shareholder derivative action lawsuits stemming from COVID-19 continue to be filed, indicating coronavirus-related Directors and Officers (D&O) losses may continue long into the upcoming year, according to Advisen loss data." Pharmaceutical and biochemical companies are prominent in the legal crosshairs of investor suits now. The suits follow several different angles, from fairly customary shareholder class actions to derivative shareholder and capital regulatory based actions. The Advisen report contains a neat wheel graphic which shows which industries are involved in what types of actions.

While there was a flurry of filings in 2020, new cases are still arising and given what Shakespeare called "the law's delay", we can expect the D&O suits to be a feature of court calendars from some time to come. It's too soon to tell what impact this may have on D&O renewals in 2021. Be vigilant. COVID looks set to help a new generation of lawyers pay off their student loans.

Not looking good for the big, bad wolf. Hope he has E&O.


Say It Isn't So...

A thought from my friend, the CEO of 50 Companie, in Holland (hence the bicycle reference):


If you see a group of young people cycling to school in the morning, know that 60% of them cycle towards a first career track that does not yet exist. Their first job will be a job that will be created in the coming years, while they are studying.

Combined with the knowledge that 25% of the professions that we now take for granted will be gone within 10 years, it makes it clear that we are dealing with a rapidly changing labor market. [my translation]

How many of the risks you or your successor will be managing ten or twenty years from now also do not yet exist or show up merely as tiny details today? What Big Deal risks may fade in the next decade? Tomorrow will not be today with a change in the weather.


Words to Remember

Another observation on Noise:

   "Why you shouldn't buy Bitcoin when you're hungry."

Title of Jason Zweig's review of Kahneman's new book in the Wall Street Journal. Just before a meal, our judgments become especially noisy. Obviously, Jason gets it.


Share This
* Required Fields