Over at the Ottawa Citizen there is worthwhile attempt to keep the economy in perspective. The watch phrase seems to be “do not panic.” The implicit premise being that panic is driving bogus assessments of the extent of the turmoil unfolding in the economy and the labour market. At least that was the premise of Gardner’s blog post: StatsCan, you got some splainin’ to do .
For my part I am quite happy to the see that someone in the press is actually taking a look at the numbers. The problem is, in this case, Mr. Gardner does not seem to be generating his own numbers. He claims that Stats Can erred when it stated that:
This drop in employment exceeds any monthly decline during the previous economic downturns of the 1980s and 1990s.
Mr. Gardner argues that based on the Stats Can press release that reporters went about generating seriously flawed headlines. I quote:
In one news story after another, we were told that the monthly job loss figure was worse than any in the recessions of the early 1990s or the early 1980s. This isn’t just bad, we were effectively told. This is unprecendentedly (sic) bad.
Mr Gardner goes on to say:
But that conclusion is based on a grade-school mistake: The labour market isn’t the same size today as it was in the early 1990s or early 1980s. It’s bigger. A lot bigger. For the stat to tell us how bad this drop was relative to those in other recessions, it has to take that into account.
Do that and it’s not the worst drop since the data were first gathered in 1976. It’s the third worst.
Now when I first read this I thought that Mr.Gardner’s conjecture was likely right. The labour force is a lot bigger so all things being equal a large drop-off in unemployment might be relatively less severe than at other periods in the past. It would have been nice if Mr. Gardner had presented his readers with the data or the formula by which he arrived at his conclusion.
So this morning I took a virtual walk over to Stats Can and pulled three series to be exact: Labour Force 15+; Employment; and Unemployment. All monthly and seasonally adjusted going back to 1976.
Why unemployment when this relates to Stats Can’s claim about employment? Charity and balance is the answer.
So let us see if we can verify Gardner’s claim that Stats Can got it wrong and the press is full of a bunch of gloomsters.
In this graph two measures of employment loss are plotted. One is the absolute loss of jobs (in black). Here Stats Can is correct the 129,000 jobs was the most jobs ever shed since 1976. But what of Mr Gardner’s point about the changing size of the Labour market? The second measure captures this. The pink line is the percent change in employment month over month. Given that employment already accounts for 92% + of the total labour force it is by definition an ok approximation of the relative size of the Labour force. And here again we are in Historic Territory: the fall in employment was the largest relative fall since 1976.
But what if something really funky was going on with the labour force? That is, what if we used the labour force to generate our relative assessment. This seems to be Mr. Gardner’s acid test. Always dutiful I did that too. Here the change in employment was calculated as the absolute change in employment divided by the labour force.
Here again it would seem, according to the Stats Can data, that January was indeed a unprecedented happening. No matter how one cuts-it, the Stats Can conclusion that “this drop in employment exceeds any monthly decline during the previous economic downturns of the 1980s and 1990s” was exactly right.
To be fair to Mr. Gardner we could, however, take a different look at the employment situation and see if one could generate a much less novel conclusion. For that I took a look at unemployment. What I did was simply take the monthly unemployment rate and subtract Ti from T-i. This will tell us if the jump in unemployment in January was higher than during any other time since 1976 (remember this has nothing to do with the Stats Can claim; that was purely about employment).
Clearly by this metric no records were being set (yet). The problem is, however, this metric does not take into account the changing size of the labour force. Well it in fact does, it is just that if the labour force is contracting it understates the unemployment level.
How to account for Mr. Gardner’s claim that if we adjust for labour market size “it’s not the worst [employment] drop since the data were first gathered in 1976. It’s the third worst”?
I must confess I do not really know because Mr. Gardner does not present any data nor the formula he used to arrive at his conclusion. I am racking my brain to see if I can find some exotic data manipulation exercise to get to that conclusion based on the Stats Can data. But even if I did find a way and could manipulate the data I would be far away from dealing with the basis of Stats Can’s claim.
Now, if I’ve misunderstood something, I’d be delighted to be corrected. But if I’m right, I think Mr.Gardner got some splainin’ to do.