Mathematical Innumeracy in the Media

Michael Round

November 5, 2009

The stimulus package has "created or saved" 650,000 jobs, proclaims the Administration.  It's immediately heralded as proof the stimulus package has "worked", and attacked as false.  Examples are brought forth, ridiculing the data.

Where is the validity in any of this?

Let's start with unemployment?  Surely, the argument goes, if jobs had been created or saved, we'd see a declining unemployment rate.  We don't.  We see a rate hovering under 10%, projected to go above 10%, and stay there.  The Administration counters the rate would have been even higher without the stimulus package.

We're not making any headway here.

What about looking at the 650,000 number?  With individual companies reporting in the data totaling 650,000 jobs, the evidence is right in front of us!  Let's take a look at it.

This is what the media has done.

One Texas public housing authority reported hiring 6 people with $26,000 in stimulus money.  Can that be the case?  Likely not.  They were challenged, and reported back the total hours worked by these people as 450.  This new figure made it to the 650,000 database as "jobs saved".

How did this happen?  If the original number was wrong, how could the revised number be right?

Who is responsible for maintaining - and challenging - the integrity of the database?

The Department of HHS reported "saving" 9,500 jobs, by giving pay raises to employees.  I guess without the pay raise, all 9,500 would have left.  Possibly so, in an aggressive job market and low employment.  However, who of these 9,500 would leave their current job, knowing the likelihood of finding a new job would be slim?

So there's trouble in these numbers.

Obviously.

We also see it's pretty hard to even define the terms in this murky arena of "jobs saved".

We see there's political motive as well by recipients of stimulus money to inflate data.  Can you imaging receiving stimulus money to create jobs, and reporting back you didn't save any?  Hardly!

But there's more to all of this.  The media is mathematically illiterate - innumeracy, as John Allen Paulos says - regarding the investigation of these numbers.

For example, taking into consideration the examples above and a few others, one media outlet says "At least 2,000 of the 640,000 jobs the administration claimed to have save or created do not exist."

Big deal.  That's merely noise.

What's the context of the data?

Did they investigate four companies, and find all four misreporting data?

Did they investigate all companies, and find only four misreporting data?

If the former is the case, more analysis is needed.  Something's seriously wrong, here.

If the latter is the case, the data seems in tact, and the number gains in credibility.

How should the media investigate a massive amount of data to reach a general conclusion regarding an important issue like this?

Investigating every company takes resources.  Massive resources.  Is this possible?  Timely?  Probably not.

Investigating a few companies and extrapolating to the population may be quick, but may lead to false generalizations.

This trade-off is the science of sampling.

The reporting of the sampling method is as important as the science itself.

Both are necessary to understand an issue better.  The absence of either - or tragically, both - is what we have today in the media.