Archive for September, 2021

Your Personal Covid Risk

September 14th, 2021 No comments

I’ve spent like eighteen months trying to figure out how to think about and understand this question, in a way that lets me make what seem like sensible, everyday decisions. I think I’ve gotten there, or close. I’m sharing here in case it’s helpful to my gentle readers.

I’m fully vaxxed. Here’s a typical, day-to-day question: If I go out to dinner in Seattle with some random friends, indoors, unmasked, how much of a risk is that? This Dave Leonhardt article finally gave me the numbers I needed to figure that.

His (literal) headline takeway: In the U.S., if you’re vaxxed your daily odds of getting infected are about 1 in 5,000. In low-infection, hi-vax areas like Seattle, more like 1 in 10,000. (Per Leonhardt, only three places in the U.S. even collect that data for vaxed vs unvaxed: Utah, Virginia, and King County, WA. Yay us.)

But what in the hell do I do with that number? What does it mean? He tries to help: “It would take more than three months for the combined risk to reach just 1 percent.” That three-month multiplication is well-intentioned, but it’s an odd, arbitrary choice of period.

I realized long ago: when you ask “what are my odds/chances of getting infected?” (and then etc. from that), you have to ask, your odds over what period? Otherwise it’s meaningless.

So now jumping to the best thing I’ve seen, a personal Covid risk calculator that some SF folks built.

It starts with an arbitrarily-chosen personal annual risk “budget”: “I’m willing to accept a 1% annual risk of getting infected.” (This choice is baked into the site, right down to its name: “Microcovid.”) Divide that risk budget by 12 for monthly budget (0.08% risk), 52 for weekly, whatever. They use weekly, which I also find useful.

Now compare: a 1 in 10,000 daily risk, 0.01% (which sounds super low, right?) is 3.65% annual risk. (Just multiply by 365.) So I’m like, “1% annual is kind of a ridiculously low risk budget, given the low ensuing risk of hospitalization much less death.” Especially if you’re vaccinated. Those worst outcomes are very unlikely.

So I’m like, what’s a benchmark annual risk I could compare it to? Try this: An average person’s daily risk of a home accident/injury with a doctor/ER visit is 1 in 5,000. That includes kids and elderly, who are more accident-prone.

That’s 7% annual risk. Once every 14 years. Six times in an 85-year life. Seems a decent ballpark estimate to my anecdotal experience/observations, if you include childhood/old-age injuries. Maybe a bit high. Whatever.

So, say I change my annual covid-infection risk budget to 7%. Then the calculator sez: if I eat out with four friends, indoors, nobody’s masked, restaurant has a HEPA filter running, that only consumes 5% of my weekly risk budget. You can tweak those numbers as you wish; I might do so as well. But it’s not a bad starting ballpark for me.

This doesn’t touch on risk you pose to others, community risk, risk of exponential spread in the population. Or, say, the risk to your restaurant servers (notably including my daughter). Those are things I also definitely consider. But this is a baseline of what you’d need to start with, to consider those subjects.

Microfoundations: The Long Con

September 10th, 2021 No comments

I wrote this as a comment response to Ryan Avent’s great post on his Substack blog. (You should subscribe. I did.) I thought I’d share it with my gentle readers here. Lightly edited, including one additional paragraph at the end.


Hi Ryan. Great piece, thanks. A few responses:

1. “authors of published papers are not always required to make available the data underlying their work”

Not just the data! They need to provide the actual analytic mechanism, software, that they use for the calculations. A replicator cannot be expected to re-create it perfectly based on verbal explanations or even the algebraic formulas in their papers. Detailed implementation issues always arise, and replicators can look directly at how the originators dealt with them — the precise, coded derivations of different measures. Plus errors, of course, Reinhart/Rogoff being the obvious example. Gimme the spreadsheet. Or stata code *and* the spreadsheets, as in Piketty & Co.’s DINAs, whatever.

2. This all cuts to the demand for “microfoundations.” In most cynical terms, the synonym for that is “post-facto armchair psychological/behavioral justifications for model assumptions about human reaction functions.” Which generally derive their rhetorical weight from the degree to which they seem “obvious.” Making a bit of a leap here, in practice where confused notions of individual vs collective “saving” rule, this means that assumptions which seem obvious to minds steeped in puritanical Calvinism tend to dominate economic theories and models. (Even Marx had a very heavy dose; Minsky even more so. And etc.) Vs models focusing on the observed, emergent behavior of different groups, classes, etc., whatever their microcauses might be.

So (entering the Office of Self-Aggrandizement here), I’d like to bruit the following model as one that completely eschews and refuses to do that post-facto rationalization and justification veiled as microfoundations.

Even though what seems to be an ironclad “obvious” explanation is lying on the ground waiting to be picked up: “The bottom 20% turns over its wealth in annual spending six or seven times faster than the top 20% because duh, declining marginal utility.”

Just: that’s what the top/bottom 20% groups *do.*

It’s like modeling the fluid dynamics of water in a whirlpool, or passing through a venturi. Sure, understanding the H20 molecule interactions provides a deep and rich understanding of water’s viscosity. But for the fluid model you just measure the viscosity and Bob’s your uncle.

Fully cynical view: The whole microfoundations business was/is basically a very clever dodge to require puritanical calvinism in all macroeconomic analysis. Blowing smoke and emitting chaff to to distract from and discredit any models in which group norms, cooperation, emergent properties, etc. trump simplistic additive (and “obvious”) steely-eyed self interest.

All of which has resulted in a massive and dominant intellectual infrastructure justifying insanely concentrated wealth, based on the false, moralized labeling and rhetoric of “patient savers.” (I’m looking at you, Paul Krugman.) Not just on the right, either; significant aspects of this leak into left/heterodox economics as well.

Thanks for listening… /rant