11/20 JAMA livestream: Great Barrington and John Snow

Stanford's Jay Bhattacharya helps explain the Great Barrington, MA, "Declaration," in a November JAMA livestream with Harvard's Marc Lipsitch. There are a lot of nuances in Jay's arguments that are omitted or only vaguely implicit in the GB declaration. 

But the biggest challenge I see in taking GB seriously is that its calls to protect (only) the vulnerable almost seem practically disingenuous. Lipsitch doesn't make this point exactly, but reducing community spread is clearly the second-best but practical solution. I wish Jay had offered a clearer assessment of the various marginal costs of these mitigation strategies.

Given the findings in JAMA Network Open by Ogedegbe et al., it has further become more questionable whether there really are vulnerable people outside of nursing homes, or whether instead racial and ethnic minority groups have been hit so hard by COVID19 because they must work and must live in closer proximity to one another. The costs associated with protecting them alone would surely approach the costs of "lockdowns" that reduce all activity, although it is also conceivable that the benefits of doing so would also be larger.

Survey completion during COVID19

Alternate title: "Whoomp! (There They Went!)"

Phase 1 of the Census Bureau's Pulse Survey, which ended in July, offered a valuable look into the unfolding status of Americans during the COVID19 pandemic. Phase 2 began on August 19, and it turns out that the survey instrument was shuffled and expanded.

As revealed in a tweet by Northwestern's Diane Whitmore Schanzenbach, unfortunately this change appears to have increased incompletions and nonresponse a fair amount:

Here's what we've learned so far about the spike in nonresponse in the Census Household Pulse Survey.

Of course, I'm focused on the Q on food sufficiency, but it would be straightforward to extend this to other sections. https://t.co/xU4E3kJsvs

— Diane Whitmore Schanzenbach (@dwschanz) September 10, 2020

I've been interested in the questions in the health module on insurance coverage. And unfortunately the redesign has created a very large discontinuity in the uninsured. Here is a graph of the Pulse data in 2020 showing the percent aged 18+ reporting no health insurance, which dove to 7.7%.

The other outlier is the Week 4 (5/21-5/26) measure, at 8.2%. 

In my earlier post I compared these statistics to readings from the American Community Survey in 2018 and earlier, and they looked pretty good. A next step would be to examine nonresponse and try to get a handle on just how big the bias is for this measure.

Bicycle accidents and demography

Demographers truly earn their wings when they are involved in bicycle accidents. Personally, I've been "doored" in San Francisco around 2003, resulting in a hip pointer; and a motorist turned right into me on Broadway in Oakland around 2015, breaking the car's side mirror and bruising me pretty good.

But that's nothing compared to the bicycling high jinks of the great Ansley J. Coale, briefly noted toward the end of his biographical memoirs at the National Academy:

There are so many other memories: the annual office picnic softball games in which we would each pitch for opposing teams; the trips we took together to the Caribbean; his accidents on his bike and especially the accident diving into the shallow pool in a Manila hotel (that resulted in his appearance with a huge Band-Aid on his face on the stage at the opening ceremony of an international population conference with President Ferdinand Marcos); and the many lovely dinners at the Coale’s home.

In chapter 13 of his 2000 autobiography, Coale describes an event while visiting UC Berkeley in 1987, right after a funny quote of Gene Hammell regarding "the plural of 'pedantic'":

Our friends had found us a wonderful house in Berkeley to rent, perhaps a mile and a half from the office. It was on the side of the steep hill that rises above the Bay in Berkeley, and had a view out through the Golden Gate as well as across to San Francisco, from the study in the basement and the porch on the first floor to the picture window in the main bedroom on the second floor. I had bought a bicycle and rode to work each day up and down the steep inclines along the way. Traveling down hill one morning at probably 25 miles per hour, I caught the front tire in a tarry slot, yanked on the handle bar, and soared over it to land on my head. Fortunately I had bought a sturdy plastic helmet; it was cracked; in its absence my skull would have surely been crushed. I was wearing a bag over my shoulder in which I carried a husky steel cable to lock the bike. Carried over by momentum I was severely bruised in my back by this cable. Picking myself up and waving off the motorist who had stopped to help if needed, I decided to walk back to the house and ask Sue to drive me to the office. When I first stood up I could not remember where the house was, so I simply started back uphill as the right direction, and by the end of the block I remembered how to go. Sue took me to a hospital on the campus and a very nice female doctor took x-rays in the area of the bruises on my back.

Coale had probably been taken to Cowell Memorial Hospital, which was demolished in 1993 to make way for the Haas Business School, located a few doors north of the Demography Department at 2232 Piedmont Avenue. Presumably other than not getting into a bike accident in the first place, the morals of this story are to wear a helmet, and to secure your bike lock so that it doesn't hurt you.

In Princeton's Town Topics on January 10, 1990, the following article appeared. The first sentence says it all:

A 72-year-old bicyclist suffered partial amputation of his right ring finger Thursday when his hand was struck by the side view mirror of a passing car.

Coale was apparently biking on the wrong side of Washington Road, probably heading from the old Office of Population Research building (Notestein Hall or "Cannon Club"), to get to the path past McCosh Hall.

I recalled hearing about the incident in Berkeley from members of the Demography Department when I was a graduate student in economics and spent much of my time in Demography in the late 1990s and early 2000s. While on sabbatical there in 2012-13, I wrote an article about bicycle accidents with Carl Mason, "Spinning the wheels and rolling the dice," where we argued that despite the possible benefits to cardiovascular health, bicycle commuting was probably best left to the middle-aged and older, at least in the U.S.

Health insurance coverage hasn't cratered yet

We should certainly be concerned about health insurance coverage in the U.S. during the COVID19 pandemic. But so far anyway, Census survey data suggest that either (A) people have successfully kept their insurance coverage despite the surge in unemployment, or (B) they don't yet know they may have lost their employment-based coverage.

Here's what spurred me to look more closely at this:

It’s worth noting that a quick glance suggests this study *models* what likely has happened https://t.co/OOKkdgurSt

— Ryan Edwards (@RyanDEdwards) July 13, 2020

The Times article cites a study by Families USA that leverages work by Urban Institute researchers from March 31, combined with unemployment rates. In other words, these are estimates of what should happen to health insurance when folks lose coverage through their employers.

But we also have some direct measures of health insurance coverage, albeit imperfect ones, through the innovative Census Pulse survey that debuted in late April. Technical notes show low response rates, around 2-4%, which were anticipated but still remain concerning. The mode of interview was online, with contact via email and text message.

The Pulse survey asks exactly the same question about health insurance coverage as appears in the annual American Community Survey:

Respondents who check "No" for all of these boxes are uninsured. Below is a time series of the share of U.S. residents aged 18 and older who indicated they were uninsured either in the ACS (2008-2018) or in the Pulse:

Finally, here's a zoomed portion of the graph during the Pulse period only:

The quality of the Pulse data is the big open question here. Published standard errors from Census imply that there is only about a ± 0.6 percentage point confidence interval (i.e., 1.96 SE's) around these point estimates, due to sampling error alone. The big question we don't know the answer to is whether people who are more likely to answer the Pulse at all are more or less likely to have lost their health insurance than others. You would think that folks who are unavailable are probably those who are busy working, which would bias this estimate of uninsurance upward.

Needless to say, permanent losses of jobs and insurance could be right around the corner. But for now, it does not appear to have happened.