Understanding Travel Risk Advice: Part Two
How to spot misleading and dangerous destination risk ratings
Be it COVID-19, crime, or social unrest, pre-departure planning and accurate awareness of destination risks are critical to safe and intelligent leisure or business travel, both domestically and internationally.
In part one, we introduced the argument for a healthy skepticism of open-source and commercial country risk ratings, focusing on the problem with nominal and ordinal scales. The ubiquitous use of ‘Medium Risk’ or ‘Level 2/Yellow’ travel risk ratings is mainly responsible for our comfort and acceptance of a practice that we know today to be measurably uninformative and potentially dangerous.
The problem with ‘color me in’ travel risk world maps is not limited to scales but is compounded by an additional, related, and possibly even more severe fallacy: the flaw of averages.
Geographic and demographic averaging
Nassim Nicholas Taleb, the famed statistician and author of Black Swan and other writings on risk, brilliantly summed up this widespread error with the quote: “Never cross a river if it is four feet deep on average.” You don’t need advanced statistical knowledge to recognize this problem if you ever tried to inform a foreign visitor about the dangers of visiting your own country. If your instincts suggested: “…Well, it depends which city you plan to visit, or even which neighborhood”, you’d be right.
Let’s see how this applies to someone visiting the USA. For expedience, we shall focus on homicide risk — which, incidentally, is the most reliable and commonly used metric when rating and comparing violent crimes internationally.
The United States’ homicide rate currently hovers around six (6) per 100K. So let’s assume that any rate between three (3) and nine (9) per 100K would fall in the same rank or class — i.e., ‘Yellow’ or ‘Level 2’ — which is more or less how these nominal scales are built.
How informative is this? Well, your chances of visiting a U.S. city with a population of one hundred thousand or more and a murder rate of six per 100k are less than one in ten. And the odds of ending up within the ‘Yellow’ range (3 to 6) would still be smaller than those offered by the flip of a coin (see image above.)
A travel decision-support tool with the potential of landing you in a city with 300% (Dallas, 18), 500% (Philadelphia, 30), or even 900% (Baltimore, 54), the suggested risk level is not only worthless but an outright disservice. Yet, this is precisely what happens any time you see one of those attractive, colorful travel risk maps.
So, next time someone suggests to you a country’s Medium-High or Level 5 risk rating is indeed very generic but helpful nonetheless (“…it conveys a general sense of the risk” and “it helps your employer attend to its duty of care”) you can confidently ignore their advice.
Geographic averaging is further aggravated by demographic averaging, which is just as severe. In Missouri, for example, a black 30-year-old male is nearly 50 times more likely to be murdered than a white female coetaneous. These enormous differences in victimization risk between demographic groups appear in most cities and countries, even if inconsistently. Moreover, the distinction between a visitor and resident status can significantly affect the weight of these ‘deviations from average’ and their direction (where being a tourist can translate into a risk discount or contributor.)
There are, in fact, plenty of metrics and even risks where such national boundaries maps and indices can be helpful and legitimately applied. Political stability or legislative protections for LGBTQ travelers are good examples. Unfortunately, criminal victimization and most accident risks are not..
You may now ask yourself how specific or narrow a risk estimate (rating) should be in order to be accurate and useful. Of course, the answer depends on the risk domain (severe weather, for example, can be much more hyperlocal and dynamic than kidnapping). Still, in the context of travel risk advice, you generally want to look for city-level data. While neighborhood-level ratings can be beneficial and are becoming increasingly available, it’s safe to assume that as a visitor, you’ll hit various parts of a city — even if only in transfer. If the risk data is appropriately normalized, crime averages (which by now should raise your internal alarms) are pretty effective in comparing one city to another.
Remember that the local ‘experts’ will inevitably object to data-driven, independent crime ratings, arguing that the data is unfairly skewed by a few bad neighborhoods and population groups (gangs, ethnic minorities, other criminal organizations.) But this is a common and easily debunked argument, given that crime and accidents in every city fall closer to a Pareto distribution (20% generating 80% of incidents) rather than being evenly distributed.
And this is an excellent introduction to our third and final source of travel risk misinformation:
Judgment Bias, Noise, and Unhelpful Heuristics
Our personal risk judgment — the evaluation of what is likely to harm or kill us — is frequently distorted by cognitive fallacies ranging from base-rate neglect (our tendency to ignore statistics in favor of anecdotes and personal experience) to substitution (adopting answers from simpler but different questions) and availability heuristics (which makes us overestimate the likelihood of salient and recent events.)
Many of these thinking patterns and shortcuts are byproducts of evolutionary demands where they may have offered advantages for the survival of the species in a hunting and gathering context — but are frequently very ill-suited and counterproductive for our modern environment.
In the spring of 2020, Safe-esteem conducted an extensive Personal Risk Judgment survey of executive protection and travel/risk intelligence professionals, as well as general public participants. The results reinforced the notion that most people are spectacularly bad at estimating the relative weight of risk domains (crime, accidents, health) and comparing violent crime risks between two or more large, well-known cities. Protection and travel professionals only edged the general public by tiny margins but were considerably more confident in their assessments.
The bottom line is that most of us are terrible at making quantitative assessments of threats and hazards. Today, we know that our gut instincts are not as wise as we hoped, despite the popularity of literature and lectures on the power of intuition.
The majority of open and commercial risk information sources only exploit and or aggravate these fallacies by emphasizing recent events (a crime, terrorist attack, or large demonstrations), individual opinions, and anecdotes (“I have been there at least ten times and I never experienced any problem — it’s very safe!”)
The increasing availability of global, significant events news — what security practitioners colloquially refer to as ‘recent intel’ about the destination — translates more often than not in judgment noise rather than accuracy.
Being aware of current circumstances and significant security events can, of course, be extremely useful, but only to the extent that these events inform our behavior without becoming a substitute for risk measurement or rating. What does it mean?
Think of recent and significant events like car accident alerts without a road map and general traffic information (which is what an accurate, data-driven risk rating system represents.) The former without the latter may lead you to work from home to offset the risk of missing a presentation, leaving you unaware that the accident had no impact on traffic in your direction or that there are plenty of alternate routes.
Similarly, if you lived in Washington, D.C., and were planning to visit Paris, France, seeing multiple news alerts of Yellow Vests demonstrations and riots may push you to overestimate the risks. As a result, you may reconsider the trip and ultimately ignore the fact that while in Paris, your chance of being murdered will be nearly 90% lower than at home.
Remember, large-scale, violent demonstrations are relatively easy to avoid compared to other violent crimes, which means the trip to Paris can still result in a high discount on your life risks — even during a period of frequent public demonstrations.
Let’s Wrap It Up
In our information age, it has become increasingly apparent that being misinformed is more likely than having no information at all. And there are few domains where this can arguably harm or even kill you more directly than personal and travel risk judgment.
Yet now, you are equipped with a new awareness of some of the common fallacies and perils found throughout open and commercial travel risk information and rating sources. In this two-part article, you learned to:
- Understand the limitations and built-in errors of nominal and ordinal scales when applied to personal risks like accidents or violent crime.
- Beware of the flaw of averages and how colorful world maps could be offering a deeply illusory — and dangerously misinforming — answer to the question: “How safe is…”
- Recognize the limitations and flaws of human judgment and gut feelings, particularly when applied to quantitative risk assessments.
Finally, there are also encouraging upsides to the information deluge and other positive developments in this domain. Thanks to Big Data, Machine Learning (ML) / Artificial Intelligence (A.I.), a growing body of research in cognitive-behavioral science and risk communication, new organizations like my firm are bringing disruptive, multi-generational advances to help individuals, families, and organizations improve measurably the quality of their risk judgment and decision-making.
