A somewhat different take on the coronavirus pandemic. From Charles Murray at aei.org:
The pandemic is complicated. Deciding on good policy is complicated. But a basic aspect of the American experience is not complicated and ought to be decisively affecting policy: the relationship of population density to the spread of the coronavirus. The relationship means that a great deal of the discussion about why some cities are doing worse than others is beside the point. These analyses may satisfy a natural urge to assign blame, and some of them surely have merit, but the role of one underlying demographic variable, population density, is immune to manipulation by the smartest policy.
Population density refers to the number of people per square mile. In this discussion, I’m going to use the county as the unit of analysis, based on the most recent census data. The data on reported coronavirus rates come from the database assembled by The New York Times and available online.
A few examples will illustrate what various population densities signify. Rural counties with a county seat of 15,000 or 20,000 typically have population densities of fewer than 100 per square mile. Counties with substantial agriculture but a small city of, say, 50,000 people, typically have population densities of several hundred per square mile but fewer than 1,000. To have an entire county show a population density of 1,000 or more requires considerable urbanization. When a county reaches a population density of 2,000 or more, it is almost always a clearly urban county. Counties that pass the 3,000 mark are almost always part of one of the densest metropolitan areas in the nation.
Using the county as the unit for expressing population density introduces noise into the conclusions we can draw regarding coronavirus rates. For example, King County in Washington, where Seattle is located, has a density of only 1,023 per square mile because the county itself is so huge. King County’s coronavirus case rate is much higher in densely populated Seattle than in the rest of King County, and using the county as the unit of aggregation obscures the role of population density. But this means that the relationship of population density to coronavirus cases that I’m about to describe understates rather than exaggerate its real role.
The figure below shows the mean county-level coronavirus rates per 100,000 population for various levels of population density.
The rate doubles from counties with a density under 1,000 people per square mile to counties with a density of 1,000–2,000, remains about the same for counties with 2,000–3,000, doubles again for counties with densities of 3,000–10,000, and doubles yet again for counties with densities exceeding 10,000 per square mile.
It might be asked whether the simple size of the population is contributing to the relationship. The answer is no. When the size of the population and the density are both entered as explanatory variables, the size of a county’s population has no independent role. (Technically, when the reported coronavirus rate per 100,000 people was regressed on the logged values of population density and population for all US counties, the regression coefficients were positive for population density, with p<.001, and negative for population, with p nonsignificant.)
This relationship of population density to the spread of the coronavirus leaves many unanswered questions. New York City has reported coronavirus rates by zip code, and the zip codes with the most affluent and best-educated populations have had substantially lower coronavirus rates than the rest of the city. Is that coincidental or causal? Does race play a role? County-level data cannot inform these issues. I have assembled coronavirus data at the zip-code level for 13 of the largest cities in the county and will discuss them in a subsequent essay. For now, I want to underline a few policy considerations that flow from the relationship of county-level population density to the spread of the coronavirus.
The relationship of population density to the spread of the coronavirus makes sense. Big, dense cities throw people into close contact with many strangers far more often than life in small, dispersed cities, let alone life in small towns and rural areas. People in big, dense cities are more likely to use public transportation every day, which is likely to be jam-packed during rush hour. They spend more time in crowded elevators and in offices with dozens or hundreds of people. They eat out more often. Parks and other recreation areas are more crowded. The proportion of people working in customer-service occupations is higher. Compare the density of pedestrian street traffic in Manhattan or downtown Chicago with pedestrian traffic in any small city. Low density reduces contact even in the largest cities. Los Angeles County has a larger population than all of New York City, but it sprawls over more than 4,700 square miles and has a population density of just 2,488 — and it has a reported coronavirus rate of 0.3 percent, a fraction of New York’s.
The relationship of population density to the spread of the coronavirus has immediate relevance to protecting many of the elderly. The elderly, especially those who are past 70, have much higher fatality rates from the coronavirus than the young and are at center stage in discussing policy options. Once again, we must separate the complicated from the simple. Underlying conditions are complicating. Living in group quarters is complicating. Poverty is complicating, especially in big cities. But there is one large group of the elderly for whom the issue is simpler: retirees who live on their own in rural, small-town, and small-city America. It is easy for most of them to take care of themselves, and they needn’t be rich to do it. We (for I fall into that category) don’t need to go to a workplace every day. We don’t need to use public transportation. Nothing requires us to eat in restaurants or, for that matter, requiresus to have close interaction with anyone. Does quarantining the entire population give us some additional measure of protection? Perhaps at the margin, though I would like to see some hard data proving that point. But I submit that we elderly who live on our own can make ourselves “safe enough” unilaterally, through the precautions within our control. What proportion of the elderly am I talking about? Calculating that number would take some digging, but wouldn’t it be nice to know what it is if we want to make sensible policy? And that brings me to my main point:
The relationship of population density to the spread of the coronavirus creates sets of policy options that are radically different in high-density and low-density areas. Take another look at that chart showing rates of reported coronavirus cases by population density. Sixty percent of the American population live in counties represented in that column at the far left. Ninety-one percent live in counties represented in the first three columns. Conjoin those percentages with this simple truth: The sensible thing for government to do about the pandemic in a small town or small city is different from the sensible thing for government to do in a big, crowded city. I’m going to leave that statement where it is, without trying to spell out a policy agenda. I don’t possess the epidemiological expertise for that. Rather, I am saying that too many people in high places, in government and the media, have been acting as if there is a right and moral policy toward the pandemic that applies throughout America. That’s wrong. Disaggregating policy choices to reflect local conditions is essential.