Wednesday, September 2, 2020

Coronavirus -- NY Times biased statistic

Yesterday's The New York Times here included the following by David Leonhardt.

Here’s a jarring thought experiment: If the United States had done merely an average job of fighting the coronavirus — if the U.S. accounted for the same share of virus deaths as it did global population — how many fewer Americans would have died?

The answer: about 145,000.

That’s a large majority of the country’s 183,000 confirmed coronavirus-related deaths.

No other country looks as bad by this measure. The U.S. accounts for 4 percent of the world’s population, and for 22 percent of confirmed Covid-19 deaths. It is one of the many signs that the Trump administration has done a poorer job of controlling the virus than dozens of other governments around the world.



End quote. Is Leonhardt trying to say that the USA has the highest number of deaths per million population in the world? I knew this isn't true, so I suspected he either erred or something reeked.

I copied some numbers from the Worldometer late yesterday and did some calculations.






Deaths/ Col 3

Deaths % World Population % World 1M_Pop. – Col 5
San Marino 42 0.00% 33,943 0.00% 1237 0.00%
Peru 29068 3.38% 33,047,700 0.42% 880 2.96%
Belgium 9895 1.15% 11,598,176 0.15% 853 1.00%
Andorra 53 0.01% 77,286 0.00% 686 0.01%
Spain 29152 3.39% 46,757,881 0.60% 623 2.79%
UK 41504 4.83% 67,946,337 0.87% 611 3.96%
Chile 11321 1.32% 19,143,709 0.25% 591 1.07%
Italy 35491 4.13% 60,446,519 0.77% 587 3.35%
Brazil 122596 14.26% 212,813,742 2.73% 576 11.53%
Sweden 5813 0.68% 10,109,890 0.13% 575 0.55%
USA 188827 21.96% 331,330,464 4.24% 570 17.72%
Mexico 64414 7.49% 129,158,631 1.65% 499 5.84%
Iran 21672 2.52% 84,171,055 1.08% 257 1.44%







World 859917
7,808,984,000
110

The numbers in the 7th column -- rightmost and Leonhardt's statistic -- for the USA, Brazil, Mexico, UK and Iran are very close to Leonhardt's. So something reeked. How can the USA rank 11th in column 6 but way ahead of all the other countries in column 7? It would clearly be biased to compare only the numbers in column 2, which depend on population size. Leonhardt's statistic manages to magnify the role of a larger population. Brazil and the USA are pretty close in deaths per million population. Leonhardt's statistic magnifies their difference.

USA, Brazil, and Mexico are the top 3 using Leonhardt's statistic. Their populations and deaths are also the top 3. The ranks of columns 2 and 7 match almost perfectly. If Leonhardt had simply based his comparison on column 2, it would have been more honest, but his bias also very obvious.

A simple example will shed more light on how Leonhardt's statistic is biased. Imagine a world of 100 million people, 4 million deaths, and only 3 regions, with people and deaths as follows.

Region       Population       Deaths
#1              30 million        3 million
#2              10 million        1 million
#3              60 million        none

Regions #1 and #2 have the same death/population, 10%. Region #1 has 75% of deaths and 30% of the total population. 75% - 30% = 45%. Region #2 has 25% of deaths and 10% of the total population. 25% - 10% = 15%. Eureka! Leonhardt's statistic much depends on population size, even magnifying the role of a larger population. He built in a bias. Clearly his motivation was to blame President Trump and his administration as much as possible.

Next assume Region #1 and #2 have half of the above death counts, or double them. Leonhardt's statistic remains the same, meaning death rates really don't matter! His statistic is nonsense.

Leonhardt claims that only 183,000 - 145,000 = 37,000 Americans would have died if its death per million population was as low as the world's. That's a big "if", because the USA had a much higher infection rate than most other countries. So did several other countries such as Belgium, Spain, UK, and Italy, but Leonhardt had no blame for their politicians. The USA's cases/population is about 5.6 times the world's.

"There are three kinds of lies: lies, damned lies, and statistics." -- Mark Twain

I wrote another post about Leonhardt's bias here.

No comments:

Post a Comment