Keeping Calm with Demographics

Michael Stone, April 19, 2013, , (src)

Finding myself unable to get back to sleep after a volley of phone calls and texts this morning, I began to wonder:

“What does it cost to shut down transportation in Boston for a day?”

To get started anwering this question, I asked Google for some links to some relevant transportation, economic, and healthcare statistics. Here are some of the interesting tidbits that their software found for me:

Conclusion: despite the fact that estimating daily system usage from annual figures is going to blow up the width of our credible intervals, perhaps Schneier has a point?

Follow-up 1:

A note that I just sent to Gov. Patrick’s office:

Dear Governor Patrick,

As a result of some calculations I did this morning [1], I am more
firmly convinced than ever that we are much better served as a city,
state, and nation by following Bruce Schneier's wise advice to "keep
calm and carry on" [2] rather than by giving in to our
(understandable!) desire to change our behavior to "stay safe".

As a result, will you please rescind the "stay indoors" request as
soon as possible so that we can get on with our lives *in parallel*
with the FBI and other Boston-area law enforcement staff's efforts to
apprehend the remaining known bombing suspect?

Thanks very much,

Michael Stone

[1]: My calculations are available here:

The conclusion of the calculations is that we should expect your
request to stay indoors for just one day to cost us tens of millions
of dollars, which we can ill afford to lose in today's economy, and
perhaps a few lives (e.g., as a result of missed doctor's

[2]: Bruce Schneier's call to carry on with business as usual is here:

Follow-up 2:

Another thought that occurred to me earlier this morning: during WWII, several sides conducted strategic bombing campaigns, dropping kilotons of high explosives and incendiary weapons, just to try to get enemy cities to shut down their transportation and industrial networks.

Thus: why are we willing to shut down one of our most profitable cities for a pair of guys with some guns, a car, and a couple of home-made bombs, one of whom is now dead and the other of whom we haven’t heard much from recently?

Follow-up 3:

My good friend Shauna has pointed out one potential silver lining to the shelter-in-place request; which I’ll paraphrase as:

“Well, at least we may save some lives by keeping people off the roads for a day!”

though, sadly, I’m not sure that the data will bear her out:

(Still, perhaps some non-fatal injuries will be prevented?)

Follow-up 4:

A friend, Damien, and a favorite former colleague, Chris, have both offered some great criticism (thanks!), each piece of which makes the puzzle all the more intriguing. Paraphrasing (and with responses included inline):

  1. Damien asks: “How much of shutdown cost will be totally lost?”

    • Me: I think my 90% credible interval, calibrated with the “Equivalent Bet Test”, covers something like \((\$0, \$2B)\), with most of the mass on \((\$1M, \$200M)\) but this is just a guess. How could we get a better estimate?

      • Damien offers the following Fermi decomposition:

        \[\mbox{Lost Wages} > \alpha \cdot \mbox{# workers} \cdot \mbox{mean hourly wage}\]

        where \(\alpha\) describes the fraction of workers who we think were affected. As for the other variables: according to this May 2012 BLS report, there are about 2M people employed in the Boston-Cambridge-Quincy MA-NH metropolitan statistical area and their mean wages are about $30/hour.

        Next, just to make up some vaguely plausible numbers for \(\alpha\): let’s imagine that, as is the case nationally, 60% of workers are paid hourly and let’s also imagine that 25% of the people employed in the Boston-Cambridge-Quincy MSA lost hours. In that case, we’d have:

        \[ \begin{align*} \mbox{Lost Wages} & > 0.25 \cdot 0.6 \cdot \mbox{2,000,000 people} \cdot \$\mbox{30 / hour / person} \\ & > \$\mbox{9,000,000 / hour} \\ \end{align*} \]

  2. Chris and Damien both ask: “What about the ‘benefit’ side of your cost/benefit analysis?”

    (Then, to sharpen the question, Chris introduces the idea of the value of a statistical life which, in the US, is commonly estimated to be around $7M. As a result, he argues (still paraphrasing) that “It’s worth spending quite a lot in order to reduce the expected number of fatalities / lost QALYs!”)

    • Me: I agree 100% that there’s a large potential benefit, quantified by the VSL argument, to the “shelter-in-place” request. However, I also think that the expected value of that argument is much lower as a result of the following two uncertainties:

      1. Is the attacker still near Boston?

      2. Is the attacker still dangerous?

      the first of which argues that the “shelter-in-place” request doesn’t help and the second of which argues that the expected number of lives remaining to be lost here is low.

  3. Chris asks: “Aren’t you misreading Schneier? I think he objects to uses of security that are unrelated to directly catching suspects, but this is not such a use!”

    • Me: First: to anyone who is assisting or has assisted with the effort to apprehend the remaining suspect: thank you – I am grateful for your work and for your willingness to put yourself in harm’s way on behalf of everyone else.

      Next, regarding the “shelter-in-place” request: I’m happy to accept that the request is tactically helpful for apprehending the remaining suspect and that it makes that task easier and safer for the law enforcement personnel and immediate bystanders who are assisting/involved.

      Sadly though, this in no way counters Schneier’s point that drastically changing our behavior to “stay safe” in way that hurts ourselves morally, economically, or resilience-wise is a strategically unsound response to asymmetric attacks on civilian populations.

  4. Chris asks: “Finally, why do you think that anyone will die as a result of missed doctors appointments?”

    • I would be surprised if many people died, but here’s why I think that at least some might:

      1. I believe that some patterns of interaction with doctors increase the population-wide life expectancy.

      2. As a result, I believe that if enough people miss enough doctor-visits, hospital admissions, etc., without some compensating heathcare replacement, then the population life-expectancy will decrease – i.e., then some people will die (or lose significant quality-of-life) sooner than if they received the care that we expected them to receive yesterday.

      3. As a result, the question becomes: “which patterns of missed doctor-patient interactions result in QALY losses?”

      4. To be clear, I am very uncertain about which patterns of missed doctor-patient interactions will result in QALY losses or death.

        However, as a result of my uncertainty, I put some mass on the possibility that there are some patients whose healthcare outcomes are sensitive to the timing of the care that they receive and that some of these patients’ access to care will be disrupted by the “shelter-in-place” request, thereby resulting in QALYs being lost.

      5. Finally, to demonstrate that there is both a large enough population of patients and that there is enough variation in outcomes within that population of patients, note that according to the CDC MMWR, we’ve seen about 121-212 all-causes deaths per week in Boston in 2013 so far. That is:

        1. there are plenty of people just in Boston who are sick enough, unlucky enough, or exposed to sufficiently dangerous environments to be put at risk of dying and

        2. from uncited background knowledge I claim that, among people who become seriously sick, there is considerable variation in outcomes even given undisrupted access to care.