Lies, damn lies and unemployment stats

[Brazen]

So the double-dip downturn could be all my fault

How one woman can cause economic boom or bust

GO FIGURE
Different ways of seeing stats

With Britain's economic fortunes in the balance, decision makers are looking at all the economic surveys. But, as Michael Blastland writes in his regular column, just one specific woman losing her job could make the difference between boom and bust.

Let's say she lives in Durham. Today, the smart skirt hangs in the cupboard. Today, her shoes are not polished. Last week, by chance, she lost her new job when an unlucky few were made redundant at Bluggins & Co, shoe-makers to the gentry, and the axe fell among recent recruits. We'll call her Eve.

This week, by chance, someone knocks on the door as part of what is known as the Labour Force Survey, a huge wheel of the statistical bureaucracy that follows the working lives of very roughly 100,000 adults. About the same as interviewing every household in Oxford every three months.

She is just one of those surveyed. But Eve, unknowingly, is about to move mountains. She will make economies tremble with a 30-minute interview and a cross in a box on a laptop questionnaire.

Vast sums of money will lurch round the world's financial system. Politicians will reel and businesses be broken.

How? Because her unemployed status is noted. It enters the record. And, in our fiction, 5,037 people who took part in the survey last month were classified as unemployed. This month, in a group that changes partially each time, there are 5,038.

Calculations are made, the new total of unemployed in the sample is extrapolated to the population as a whole, the unemployment rate found to be 7.8%. At this, the world trembles. For last month - again, in our fiction - the rate had been 7.7%.

"Analysts" had expected a fall to 7.6% - continuing signs of recovery. They declare themselves "shocked". They fail to check the headline figure against others that might calm them down and the media pronounce that unemployment is rising again.

The headline writers go to town, speculating on "Double Dip Doom". "Failed" says another, as all the attempts at economic rescue, by Bank of England or government, are written off.

The prospects for the UK's public finances are revised by commentators everywhere as the recovery is doubted and, with it, the reviving tax revenues the government depends on to reduce its monthly borrowing. The markets are jittery. They have Greece in mind. A rating agency downgrades the nation's credit worthiness.

Panic sets in. Money takes flight. No-one wants UK gilts unless interest rates rise horribly. The UK public finances are said to face meltdown.

And at the bottom of this mountain of fuss lies Eve. Why she alone? Because the sample size on which the unemployment numbers are based is such that a change of 0.1% in the unemployment rate is equal to about 65 people.

But it might be even more finely balanced than that. For the numbers are unlikely to add up to anything divisible by a round 65. And if there are, say 32 left over, the number will be rounded down, but if there are 33, it will be rounded up.

Thus can one person make the difference between whether or not the official unemployment rate rises.

This is not a criticism of the way the data is gathered or presented at source. Check the ONS and it states clearly that the figure is accurate only to 0.2 per cent, most of the time. This means that a rise of 0.1 per cent in the unemployment rate could be consistent with an actual fall in unemployment across the whole economy of 0.1 per cent.

The range of uncertainty is equal to tens of thousands of people in the real economy. The sampling variability once you get down to the regions of the UK is about three times bigger than for the whole country.

Our little fantasy is improbable and depends on unlikely statistical happenstance and thoughtless public comment. But the point is that feverish times make attention twitchy. Every piece of evidence about the state of the economy is interpreted, explanations offered, forecasts recalculated, and much is made out of little, perhaps too much.

The difference between a rise and a fall is judged with solemn faces when the truth is the change we observe may not even be there. Economic data is never a set of facts; it is a set of clues, some of which are the red herrings of unavoidable measurement error.

The Labour Force Survey is one of the biggest and best around, but it is a question to ask of all surveys: on how many real people in those surveys does the appearance of change depend? Not how many in the survey, but how many make the difference.

The next day, before the figures were even published, Eve found a new job.

http://news.bbc.co.uk/1/hi/magazine/8538751.stm

[The Brain]

Er... I don't get it.

[Jaron]

If he doesn't get it, I'm not even going to bother trying.

[Brazen]

Er... I don't get it.

Quick precis: Unemployment statistics are based on a variable and statistically insignificant sample so one person reporting losing their job (even if they found one since) can make it look like the recession is worsening and lead to market uncertainty which in turn further lengthens the downturn.

[The Brain]

Er... I don't get it.

Quick precis: Unemployment statistics are based on a variable and statistically insignificant sample so one person reporting losing their job (even if they found one since) can make it look like the recession is worsening and lead to market uncertainty which in turn further lengthens the downturn.

Seems to me that there is nothing in the article. Yes people are free to disregard the well known uncertainty of a measurement (any measurement), and often take advantage of this, but so what? Even the article doesn't criticize the actual measurement or how its uncertainty is presented.

[citizen k]

The Brain: Swedish Grumbler?

[The Brain]

The Brain: Swedish Grumbler?

Nice strawman.

[Martinus]

The Brain: Swedish Grumbler?

Nice strawman.

How much is that nice strawman in the window?

[The Brain]

The Brain: Swedish Grumbler?

Nice strawman.

How much is that nice strawman in the window?

Not for sale.

[Martinus]

 

I wanted my own strawman to be as nice as the one grumbler is making all the time.

[Admiral Yi]

Quick precis: Unemployment statistics are based on a variable and statistically insignificant sample so one person reporting losing their job (even if they found one since) can make it look like the recession is worsening and lead to market uncertainty which in turn further lengthens the downturn.

One person is 1/1000 percent of 100,000.  So a lot more than one person changing their answer.

[Zanza]

So the point of the article is that with every statistic there is one marginal item that will change a percentage?

5000 persons surveyed is not bad as long as they are somewhat representative of the general population.

[grumbler]

I wanted my own strawman to be as nice as the one grumbler is making all the time.

I am too busy making unlikely claims about being a lawyer to make any strawmen. 

Hell, I don't even have time to tackle random people with rifles!

[Admiral Yi]

So the point of the article is that with every statistic there is one marginal item that will change a percentage?

5000 persons surveyed is not bad as long as they are somewhat representative of the general population.

The sample size is 100,000.  5,000 is the number of hypothetical unemployed people out of that sample.

[DGuller]

Er... I don't get it.

Quick precis: Unemployment statistics are based on a variable and statistically insignificant sample so one person reporting losing their job (even if they found one since) can make it look like the recession is worsening and lead to market uncertainty which in turn further lengthens the downturn.

I think you have gotten the wrong impression from that article.  The actual point of the article is that there is statistical uncertainty in the unemployment rate, and that many analysts may be reading too far into what may really be just noise.  It doesn't claim that the sample used is statistically insignificant, and it would be very wrong to do so.