# 5 Levels of Risk Analysis: We’re All Going to Die

Ever watch those YouTube videos where somebody will take a topic or piece of music and go through X levels of complexity, from simplest to most sophisticated? Malaguena, as played by a beginner, intermediate, advanced, pro? Or donuts as made by a newbie, an experienced amateur, and a chef? How about we do that for risk of death?

We’ll leave off the Level 0 analysis, by far the most common – basically, I don’t understand risk and therefore will dismiss risk analysis as unimportant nit-picking, and simply take whatever steps everybody else in my social group is taking alleged to reduce whatever risk my social group decides needs reducing. (1) We will instead start with rational risk analysis.

Note: Even though the approaches proposed are, except for the first and last levels, seriously flawed when considering an individual’s risk, it is sobering that, even as flawed as they are, any attempt to understand risk of death dispassionately in our current state is an improvement. Onward:

## How To Understand Your Risk of Death – 5 Levels:

Level I – Everybody dies. You are at 100% risk of death. No getting around it. Your death, my death, anybody’s death, is as certain as taxes. When, not if.

Level II – Risk of dying this year, as a member of the human race: is somewhat less than 1%, or 1 in a hundred. This means that, across the population of the planet as a whole, something less than 1 out of 100 people walking the earth or born during a given year is going to be dead before the year is out. You can see this in, for example, UN death rate data, or just applying common sense: by age 100, just about everybody is dead; therefore, the annual death rate for everybody taken together could plausibly be a smidge under 1% (It’s more complicated than that, to be sure, but in a very simple, homogenous population, getting neither bigger nor smaller, nor older nor younger year over year, it would roughly be right.)

Pro Tip: when you see very accurate numbers – not a reasonable-ish claim like ‘about 0.9%’ but a remarkably accurate-sounding numbers like ‘0.8762%’ for estimates based on huge, difficult-to-count numbers, like world population, the klaxons of your BS detectors, if working properly, should be firing on 11. You get a death rate by dividing raw numbers of deaths by a population estimate. It would be surprising, given the people, methods, and motivations involved, if world population estimates were accurate within, say, a billion people. It’s not easy counting up a population – look at how involved the US Census is. Death counts seem much more simple – but are they? In, say, South Sudan or Mongolia or the Amazon rain forests? People don’t just die off-grid, as it were? Most other nations cannot or do not put the sort of effort into it the US does, and have reasons, often, for fudging. Three or 4 decimal places of accuracy might be plausible in the US, maybe, but the world? No. All those decimal places are pretty good indicator that somebody is trying to snow you, to keep you from thinking about how much uncertainty surrounds their numbers.

Level III – Risk of dying this year, as an American: your (and my) chances of dying this year simply by dint of being American is, according to the UN, about .8977%. Just about 9 out of every thousand of us in the US are expected to die in 2021. Let’s blow those numbers out to see some totals: using the last census count of something like 332 million Americans, looks like right about 3 million Americans are expected to die in 2021 in the normal course of things (See: Level I). That does not include any effects of COVID, by the way.

Level IV – Risk of dying by age: The astute observer will have noticed that death does not come at random, usually, but rather that some people are a lot more likely to drop dead than others over any short period of time, a year, say. Age is a huge part of this, such an astute observer will observe: young people tend to think they are immortal, which is a fairly reasonable or at least understandable position, given how few of their contemporaries die in a given year. They may know that one kid who died in a car wreck, or some poor kid who killed herself, but these seem pretty special cases – *I* drive good! I don’t want to kill myself! Old people, on the other hand, see their contemporaries increasingly dropping as they age, until it finally gets them.

And the data, as collected by insurance companies for a couple centuries now, bears this out. If you make it out of infancy and don’t have a crippling disease, you are all but guaranteed to make it to 50 – over 95% of American women and 92% of American men survive to age 50. It’s only about age 56 for men and 63 for women that the chance of death per year approaches the UN overall estimated US death rate of 0.8977%.

The obvious starting point, then, for any understanding of your risk of death is: how old are you? If you’re under 50, in general, it’s very low, but then starts slowly climbing until, finally, everybody is dead – around age 110 or so. We all seem to know this and take it for granted, except when we don’t.

Level V – How healthy are you? Here’s the bottom line. Are you healthy? Then you have very little annual risk of death above the background ‘stuff happens’ level. That is not a very high level, but is never zero. Let’s look at an actuarial table, for example, this one based off the Social Security website (relevant numbers included at the bottom of this post). Remember, this is the chance of death lumping all people of the same age together regardless of health. In reality, your or my or anybody’s chance of dying in a particular year has a lot more to do with overall health first of all, and risky personal behaviors second, than age directly.

Most people get sick and die, usually, but not always, when they’re old. But it is not age, per se, that kills them. Old age eventually leads to the human body wearing out and shutting down, on the one hand, and an increase in diseases as the body weakens and cannot fight them off so well, on the other. It’s perfectly useful to say people die of old age when we are describing the outcome of this shutting down and weakening. Clearly, however, being old – being 80 or 70 or 65 – isn’t the cause of death, as plenty of people those ages are vigorous. Age is a proxy for health, useful as a generalization, not so useful in individual cases.

## Author: Joseph Moore

Enough with the smarty-pants Dante quote. Just some opinionated blogger dude.

## 3 thoughts on “5 Levels of Risk Analysis: We’re All Going to Die”

1. EA says:

Call me a rabbit; I’ll continue to wear it proudly. Your presentation of risk analysis spread over broad populations with Level V addressing only two potential factors: physical health and risky behaviors seems to me useless on a practical level (which is where I, anyway, live.) Take, for example, driving- a fairly risky activity – you seem to argue that the risk of driving is only expressed broadly across the population vs. statistics on death from other causes. In the practical world, the risk of driving varies based on all sorts of external factors. The risk of driving to the grocery store alone on a day with temperate weather is utterly different that the risk of driving 60 miles to the nearest metropolis, which is different than driving to the metropolis after 8 pm on St. Patrick’s Day, driving 200 miles while sleep deprived, with screaming children, etc. etc. et c.

1. The example you give for driving is a good one for the personal behaviors point. It boils down to assessing which personal behaviors affect outcomes. As in COVID, there are generally two subpopulations to consider for driving risk: there are those who drive carefully, keep their cars up, and stay out of dangerous driving situations – and those who don’t. (Of course, there are people in-between, but, as a former underwriter, I can tell you it’s surprising how small that group is.) The first group of careful drivers has a low background risk of accidents and death, a large part of which is getting clobbered by members of the second group. The second group – young males, drunk drivers, road ragers, cell-phone users – account for a very disproportionate percentage of ‘accidents’.

With COVID, however, the two main groups do not break down along lines of behavior. About 2/3rds of attributed deaths are among the 0,5% of the population in nursing homes. Not in a nursing home? Like a good driver, your risk is much smaller. But it has nothing to do with your behavior.

But good drivers wear seat belts! And smart laymen wear masks! Is that the proper analogy? Seatbelts do have a comparatively minor but real effect on reducing death and injury in car crashes, validated by years of accident data. Up until 15 months ago, no one anywhere had any data supporting the idea that masks *as used by the general public* reduces overall risk of airborne viral infection. No data demonstrating that masks work *as used by the general public* have yet to be produced. Yes, I looked at the “70 studies” – at least, I read each of the first five until I reached the point of incredulity: as a scientifically literate person, I know 1) studies in themselves are just claims; 2) those claims rest on evidence and logic; 3) most studies are ‘wrong’, meaning the evidence and logic do not sufficiently or at all support the conclusions. That’s why science demands criticism and replication of study results before claims can be taken seriously.

What am I, or any scientifically literate person, looking for, in a mask study? Let’s list it:

1.) Testing the actual claim. Saying masks work in clinical setting says nothing about if they work when put on by fumble-fingered non-professionals who have not scrubbed up and gloved up, and don’t change their masks (putting the old on into hazardous waste) every 20-30 minutes.
2) Responses to obvious criticisms. Study authors are often meticulous about the challenges they think they can answer, but silent on those they can’t. So, when a study ignores such things as logic of mask disposal – if they work, used masks are now loaded with deadly virions, and are biohazards themselves! Now what? – and focus on one, and only one, aspect of risk, we’re talking magic, not science.
3. Valid methodology. This one gets a little more technical, but, for example, meta-studies – studies of studies – need to account for a boatload of differences in each of the base studies they’re analyzing AND show that they are not cherry-picking studies they like. For starters.
4. Significance: studies all involve uncertainty – of method, of data, even in the basic math. Those uncertainties don’t go away just because I really, really want them to. To be meaningful, a study’s results must be significantly larger than the cumulative amount of uncertainty. One study of the legendary 70 showed something like a 2.5% reduction in spread rate across several very different population, where the counting had been done by different, or at least unexamined methodologies. That’s a very small difference claimed against a very large amount of uncertainty. And reeks of cherry-picking.

Conclusion: looking at the data coming out state by state, there is NO, as in NO, difference in outcome that follows mask use, lockdowns, compliance assumptions, or anything else at all. All that rigmarole simply didn’t matter. To see this, step one: stop listening to professional panic mongers, and look at the data yourself. If you don’t feel qualified to interpret the data yourself, I suggest William Briggs. http://wmbriggs.com/ Or me.

2. Then there’s the “You survive, but you won’t enjoy it option”. I am rather okay with the death bit aside from missing marriages and grandchildren et al.