askblog

1. Tina Lu and Ben Y. Reis write,

We found that Internet search patterns reveal a robust temporal pattern of disease progression for COVID-19: Initial symptoms of fever, dry cough, sore throat and chills are followed by shortness of breath an average of 5.22 days [95% CI 3.30-7.14] after symptom onset, matching the precise clinical course reported in the medical literature. Furthermore, we found that increases in COVID-19-symptom-related searches predict increases in reported COVID-19 cases and deaths 18.53 days [95% CI 15.98-21.08] and 22.16 days [95% CI 20.33-23.99] in advance, respectively. This is the first study to show that Internet search patterns can be used to reveal the detailed clinical course of a disease. These data can be used to track and predict the local spread of COVID-19 before widespread laboratory testing becomes available in each country, helping to guide the current public health response.

Pointer from John Alcorn.

2. In another Alcorn pointer, Elizabeth Williamson and others conducted a large cohort analysis of British patients. My sense of these studies is that we always see a very strong relationship between age and death rates and higher mortality for men than women. Which co-morbidities show up as significant varies, and I suspect this is due to the way that these tend to correlate with age. If you specify the age variable in a sophisticated, non-linear way, its explanatory power will be such that co-morbidities may not show much independent effect.

I would like to see some of these explanatory models run with no age variable at all, to see how much variation you can explain with co-morbidities alone. My guess is that such equations will explain a smaller share of the variance, but they may offer better insight into the relative importance of different co-morbidities.

Here is a press release for the study.

3. A software engineer criticizes the Imperial model.

It’s clear from reading the code that in 2014 Imperial tried to make the code use multiple CPUs to speed it up, but never made it work reliably. This sort of programming is known to be difficult and usually requires senior, experienced engineers to get good results. Results that randomly change from run to run are a common consequence of thread-safety bugs.

Pointer from Tyler Cowen.

4. From a correspondent:

I am a small business owner, and I see a big problem with the PPP Loans that is not discussed in the media.

PPP loan forgiveness is planned as a percentage of payroll costs over the eight weeks following loan distribution. Therefore, businesses that continue to operate, with normal payroll costs, expect to receive large loan forgiveness. However, those aren’t the businesses that need forgiveness; they have continued to operate. The businesses who need the money are shut down, don’t have any payroll costs, and will therefore receive little forgiveness.

My business has continued to operate with a modest decline in sales. On my application, I could honestly certify, “current economic uncertainty makes [the] loan request necessary to support the ongoing operations of the Applicant”; I am certainly facing plenty of uncertainty. However, I am unlikely to actually need the money.

I accepted the loan and will ask for forgiveness, because if the government is handing out free money, I am going to accept; the government takes plenty of my money without asking. While there might be a few angels who don’t apply for the PPP loan, most self-interested owners will accept forgiveness. As one of my industry colleagues said to another, “Time to PPParty!”

The solution? Make the PPP loans non-forgivable LOANS, with a long (5 years?) repayment period at 0-1% interest.

PPP was rushed out – rightly in my opinion — because of the immediate nature of the problem, but Congress has time to change the program in the six months before forgiveness and repayment start.

Nobody who already accepted loans will be hurt. They can simply repay the loan.

If PPP is changed from grants to loans, owners facing huge losses from closure can at least amortize that loss over a period of years to enable them to remain in business. Furthermore, there remains the possibility to forgive some loans in the future, after more careful analysis of actual losses incurred.

5. Tomaz Cajner and others write,

Using weekly, anonymized administrative payroll data from the largest U.S. payroll processing company, we measure the deterioration of the U.S. labor market during the first two months of the global COVID-19 pandemic. We find that U.S. private-sector employment contracted by about 22 percent between mid-February and mid-April. Businesses suspending operations—perhaps temporarily—account for a significant share of employment losses, particularly among smaller businesses. Hours worked for continuing workers fell by 4.5 percent. We highlight large differences in employment declines by industry, business size, state of residence, and demographic group. Workers in the bottom quintile of the wage distribution experienced a 35 percent employment decline while those in the top quintile experienced only a 9 percent decline.

1. A reader writes, “epidemiology as GDP-factory-ism”

So many epidemiology models seem to use THE value for R, or THE case fatality rate (CFR) or THE infection fatality rate (IFR). But these rates differ for different people and different circumstances. It seems the same kind of simplification that treats output as a single something. The simplification can be useful some times but there is such a temptation to use it without asking that question.

If you want to know why I am so adamant and ornery about the models, that is it. They remind me of macroeconometric models, which I am confident are misguided.

The modelers are still at it. The NYT reports,

The daily death toll will reach about 3,000 on June 1, according to an internal document obtained by The New York Times, nearly double the current number of about 1,750.

I would like to make the following bet with these modelers. I bet that the daily death toll in the last two weeks of May averages less than 2500. Whoever loses the bet has to shut up. If I lose, I stop blogging about the virus. If the modelers lose, then they have to stop reporting results from their models.

Note that Daron Acemoglu and others have disaggregated the typical model into more than one risk bucket. Tyler Cowen enthuses, “I would say we are finally making progress.” I say it’s just more social-engineering drunks searching under the lamppost.

2. Another reader points to an essay by Sean Trende. Difficult to excerpt, the essay seems quite rational to me.

No states are on anything resembling an exponential growth trajectory, almost all states are past a peak, and most states are substantially so. This would suggest that in many states, the question really should be how to reopen while keeping hospitals from being overwhelmed again.

As Tyler Cowen once predicted, we went from insufficient fear to excess fear. With excess fear, it will be difficult to re-start the economy. Even if restrictions are lifted, people will not be confident as consumers or entrepreneurs.

3. Alberto M. Borobia and others look at a cohort of patients at a major teaching hospital in Madrid. It is worth poring over the tables at the end. As I read table 3, out of 665 patients under age 50, only 5 died. That is a mortality rate of less than 1 percent among those hospitalized. To compute the overall infection fatality rate for those in that age group, one would have to multiply by the probability that an infected person becomes hospitalized. If the latter is 0.1, for example, then the IFR would be less than 1 in a thousand. Pointer from John Alcorn.

He also points to a study by Zichen Wang and others of patients in three New York hospitals. As I read the tables, obesity does not seem to be associated with a greater likelihood of death, but hypertension does.

And he points to yet another study, in the LA area. They find that a big difference of male vs. female.

One thing I would like to see from these cohort studies is a really careful analysis of the relationship between the risk from age and the risk from comorbidities, given that the high correlation between the two.

4. Robin Hanson writes,

We are starting to open, and will continue to open, as long as opening is the main well-supported alternative to the closed status quo, which we can all see isn’t working as fast as expected, and plausibly not fast enough to be a net gain. Hearing elites debate a dozen other alternatives, each supported by different theories and groups, will not be enough to resist that pressure to open.

Winning at politics requires more than just prestige, good ideas, and passion. It also requires compromise, to produce sufficient unity. At this game, elites are now failing, while the public is not.

I am not rooting for the elites to win. I don’t think any top-down solution is going to work well. Letting individuals decide which risks they are willing to take is probably the best approach. As someone who will be making risk-averse choices, I do not think others’ riskier choices pose a significant threat to me.

5. A commenter writes,

We shouldn’t be trying to conquer fear so we can go back to the old economy. We should be building the new economy that has an order of magnitude fewer casual human interactions.

Maybe this is overstating it. But I do think that we will see new patterns of specialization and trade, and we need a lot of capitalism to get there.