# Lying With Statistics

You can tell the truth and convey the wrong message. If you do it on purpose, that’s a lie.

### An example.

It is based on a 24 Aug 2021 CBC report that I doubt is a lie of any kind, but let’s see how the data can be used to create one.

“Of the 434 cases today with a known vaccination status, 279 were people who haven’t had a single shot. Another 41 cases were found in people who have had one vaccine, and a further 114 cases were found in fully vaccinated people..”

That is 64% of the new cases involved unvaccinated people, while 36% involved vaccinated people.

A possible misleading message is this. True but misleading. In 2020 not a single new case involved a vaccinated person. All the cases involved unvaccinated people. Clearly, it has become very risky too be vaccinated. The unvaccinated are showing up with a 36% reduction in their share of cases, while cases involving the vaccinated are exploding. From none to 36% of all cases.

Can you imagine the headline? “Vaccines implicated in 36% of all new cases. Unvaccinated cases falling”

### What to watch for.

Most statistical lies are based on just a few methods.

1. Inconsistent context. Like the example. There was no vaccine in 2020 and there was in 2021. Any comparison will be misleading.
2. Be sure the end points are representative of the whole. Carefully choosing the ends can be very misleading. Suppose I say my investment fund rose 68% in a year. 9 March 2009 to 9 March 2010. Nice, but what if you also knew the fund rose 16.5% in the one year from 1 July 2009 to 1 July 2010. Picking the beginning of the period carefully increased the apparent return more than four times. Never accept one number as conclusive.
3.  Comparison to small numbers. Suppose the odds of getting some rare disease are two chances in 100,000 careful research finds that if you eat peanut butter the odds become 4 in 100,000. The headline? Eating Peanut butter doubles your chances of getting the disease. There is still 99996 out of 100000 you won’t but 99996 compared to 99998 isn’t very newsworthy. Can you imagine this headline, “Eating peanut butter reduces your survival chances by two thousandths of one percent.” Both comparisons are true. The second will not affect your behaviour.
4. Watch for graphs using inverted scales. They convey the wrong message because intuitively  up means up.
5. Be sure the headline and the story convey the same message. Most people read only headlines.

### The takeaway

The news needs to be spectacular or people don’t pay attention. Think if reality means what the headline implies.

I help people have more retirement income and larger, more liquid estates.

Call in Canada 705-927-4770, or email don@moneyfyi.com

This site uses Akismet to reduce spam. Learn how your comment data is processed.