Saturday, November 19, 2016

Be careful with numbers in the media and in politics

The blogster likes numbers. It* enjoys the relative comfort of measurements and predictions using numbers. Despite people's best efforts, we do get numbers wrong - just look at the polls that did not accurately predict the British EU referendum or the US presidential race.

While getting numbers wrong is simply a fact of life, the blogster still tends to get upset every now and then when numbers are used in a biased, misleading, or outright lying manner.

So, here is a short compilation, in no particular order and by no means complete, of the use of numbers that should make readers question the purpose and prompt more research.

We'll start with the absence of solid numbers.

No figures are collected
A time tested way to get around solid debate is to not collect figures at all. In Germany, this has worked for the number of homeless or the number of illegal aliens. Estimates is all they have.

It is too difficult to get good numbers
In Germany, this argument has been made for both the homeless and the illegal aliens figure. It holds true for illegal aliens - no country has really reliable numbers on those, but in the case of German homeless, the media have noted utter disinterest by successive governments to even try and get reliable figures.
In general, the more illegal something is, the harder it is to get good numbers.

Zombie numbers
There are numbers out there that we take for granted because they have been around for so long. Yet, there is no firm base for them. How can you spot them?

You cannot.

Unless you happen to know the provenance of, say, the "3000 deadheads in prison", you may never know that this was an estimate by someone not exactly qualified to make that estimate.

The real origin and meaning of some popular figures can be found with a little bit of research, for example, the claim that it needs two years before an employee is productive in a new job. Taught by such luminaries as French economics teacher and ex president Sarkozy, it is based on the claim that it takes 10 000 hours to learn a new skill. The TEDx presentation The first 20 hours -- how to learn anything tells you how the number 10 000 came about and how its meaning changed over time, basically from "become a chess champion" to "learn a new skill".

Complete omission of existing solid figures
In speeches and statements, one popular device is to provide no figures at all. This has been a staple in speeches on crime by "law and order" politicians and their media helpers.
It very likely contributes to the perception in the US as well as in Germany that crime is up despite statistics showing it has been down.

Partial omission
One current area of reporting in which we see this frequently is the Syria conflict. Numbers of civilians killed or injured by the Syrian government forces and Russian planes are everywhere. Civilian victims of "rebels" or Western forces are rare. Obtaining figures is not easy but the gap seems to indicate either neglect or intent.

Easily spotted bias
Sometimes, the way numbers are provided shows fairly easily detectable bias. You can find a small example in the post Even short news articles are biased: the 2017 hike of German basic social HARTZ IV benefits.

Compounded figures
One outstanding example of compounded figures was the early reporting on the German social security retirement benefits reform in 2014.
Almost all outlets had headlines like "Reform to cost 160 billion Euros". Only when you read the whole article would you find - in most cases - the added information on the time frame: until the year 2030.
Ten billion a year sounds a lot less scary. Given that the loss to the German economy caused by sanctions on Russia was reported as "6 billion Euros a year" at more or less the same time, one can wonder why the compounded 160 billion was so prominent.

No base figures are provided
Not providing base figures is so common that laziness is really the only sane explanation. For example, a recent article in one German paper reported 500 terror dead in "the OECD" countries.  Another article reported that Twitter had deleted 100 000  hate speech accounts in the past year. The number of total accounts on the platform was not provided.

Percent values should set off a warning
Percentages deserve a special note because they are less intuitive than absolute numbers. Few people think about what a small inflation rate means, but percentage values without base figures are truly evil.

The most egregious example the blogster has seen in recent months was an article in the big German daily Frankfurter Allgemeine. The article reported "an increase in online child abuse by 1000%".
No base figure at all was given.

Don't believe the effect of percentages? Try for yourself. Sit down and work out some percentage values, then try them on friends and family.

Let the blogster know what you found.

And remember, we didn't even get started on really huge and really tiny figures and logarithmic numbers.


* Gender neutrality is a thing at the K-Landnews. 
[Update 11/20/2016] Changed sentence structure of "Don't believe the effect of percentages" for clarity. 
[Update 11/22/2016] Added  the example paragraph "The real origin and meaning of some popular figures..."

No comments:

Post a Comment