I do worry about The Telegraph sometimes. Edmund Conway, Economics Editor, is presumably the best person working for that newspaper to explain how this whole complicated economy thing works. In fact, the author blurb of his blog proffers, in a charmingly self-deprecating way, “Come join in as I try to get my head, and hopefully yours, round what on earth is happening in the financial crisis”. Ah, how sweet. Unfortunately, based on the evidence of his recent article, “America: the least generous unemployment system in the world“, I think he first needs to get his head around basic analysis.
Conway begins by saying:
How is it that the American economy manages year-in-year-out to outperform its European neighbours in economic terms? There is no simple answer, of course, but this chart might hold some of the clues. It shows the comparative generosity of long-term unemployment benefits around the world – and guess who is right at the very bottom?
I would, of course, agree with the statement that there is no simple answer. But, as someone who isn’t economics editor for The Telegraph, I would also suggest the answer to the first question is that the US has almost four times (3.8 2DP) as many people living there than the biggest European country, Germany. I would also then take a couple of minutes to do a little bit of research and, judging performance in economic terms by the widely accepted measure of GDP, discover that the US is also almost four times (3.9 2DP) as strong economically. Interesting. But, Conway wants to talk about unemployment benefit, so let’s have a look at that graph…
Well, that certainly does show that the US has one of the least generous unemployment benefit systems in the world. Does Conway really expect us to read that much into it?
If you were after some evidence of how the US has managed to enshrine hard-working values in its citizens, this chart is probably a good place to start. And these figures matter.
It looks like he does. Conway asserts that this graph is evidence that the United States’ miserly unemployment benefits are a driver for its economic power. Likewise, “European-style statism” results in the opposite.
On the face of it, this is obviously stupid. You could take almost any measurement that shows the US at one extreme or another and assert that this relates to a similarly extreme measurement. For example, America manages year-in-year-out to outperform its European neighbours in terms of total recorded rapes. Does this mean there’s a relationship between the number of rapes in a country and economic performance?
I know, I know. This is a flippant comparison. Conway does at least attempt to provide a theory suggesting cause-and-effect, but it’s absolutely pointless if he doesn’t illustrate the relationship in hard terms. Continuing the tradition of his Telegraph blogging brethren, Conway clearly hasn’t bothered to actually check the validity of his claims. It’s far easier to just say stuff, safe in the knowledge that your readers have already decided that it’s true.
Fortunately, I’m here to do all the hard work – unpaid, of course – and help this successful journalist and expert on all matters economical wrap his head around such tricky things.
Let’s start by testing his theory that less generosity results in greater economic performance. Here’s a far more useful graph than the one provided by Conway (click on it to enlarge). I’ve compared the generosity of unemployment benefit figures with GDP by country:
If Conway’s belief that less generous unemployment benefits drives greater economic strength was correct, we’d expect to see the countries who are most generous (the ones on the left) experience lower GDP. Similarly, those that are least generous (the ones on the right) would have higher GDP. A correlation isn’t immediately obvious, but the flaw in this comparison is – specifically, the countries with the biggest economies are also the countries with the highest populations. In fact, if you were to rank the top seven in terms of population and GDP, I think the relationship would be clear. So, let’s take that bias out of the equation and use GDP per capita (click on the image to enlarge).
Now this is much more interesting! Well, not really. There’s no clear correlation here. Although, if anything, countries with a generosity percentage around 20% or lower seem to come off worse. But I’d hesitate before doing a Conway and jumping to conclusions, the countries with the lowest GDP per capita were all members of the Eastern Bloc, which must go some way to explaining their weaker economies. What’s most intriguing is the fact that America actually underperforms compared to many of its European neighbours (including the bounteous Denmark – which is eleven times more generous than the US!)
So, what are the conclusions?
Despite what you’re undoubtedly going to hear repeated ad nauseum by right wing commentators and the coalition government over the next few months (possibly years), there is no obvious connection between generosity of unemployment benefits and economic performance.
Edmund Conway is wrong.
Update: based on a suggestion left as a comment, I’ve created a new graph comparing generosity with long-term unemployment. This is an interesting comparison to make and could show us that being less generous with benefits ensures people stay in work – even if it doesn’t exactly help GDP (as usual, click on the image to see the full size version).
Again, a disappointing graph showing little to no correlation: Italy has almost five times the rate of long-term unemployment as the US, but is only 3 points more “generous”. Among European countries, there seems to be a slight trend for countries offering lower benefits suffering from higher long-term unemployment. However, the US is a clear exception with the lowest long-term unemployment within this group of countries. This raises many interesting questions: what is it that keeps long-term unemployment low if not a tightfisted benefit system? Is there a unifying “work ethic” between countries with low long-term unemployment? Is it simply related to availability of jobs? In short, what the fuck is going on?
Out of interest, I compared long-term unemployment with employment protection (EP), the latter referring to the level of regulations concerning the hiring and firing of employees. This last gasp at finding something worth talking about could suggest a relationship between less “red tape” and greater workforce dynamism. This comparison is shown on the graph below (click to see full size).
This is almost very exciting, but sadly it’s a case of close but no cigar. If you kind of squint your eyes and look for a pattern you want to see, you could argue the correlation is quite convincing, but unfortunately it isn’t.
Just to be clear (mainly for you stat junkies out there), here are the correlation coefficients for all the comparisons shown in this post…
GDP vs. Generosity = -0.32
GDP per Capita vs. Generosity = 0.30
Long-term unemployment vs. Generosity = -0.22
Long-term unemployment vs. EP = 0.36
If you’re not familiar with how you measure the correlation between two datasets, here’s a brief explanation of what these results mean:
Correlation is measured on a scale of -1 to +1
A perfect positive correlation will achieve a score of +1 (the higher the values of one dataset, the higher the values of the other dataset), while a perfect negative correlation will achieve a score of -1 (the higher the values of one dataset, the lower the values of the other dataset)
Any number in between these two “perfect” scores shows the strength of correlation (0.87 is a strong positive correlation, while -0.25 is a barely significant negative correlation)
Alone, these are pretty useful figures, but we can make judge our results even more accurately by taking into account the critical value – that is, the strength our correlation has to reach before it becomes significant. Using this handy table with a degree of freedom (df, the number of subjects in the study minus 2) of 18 and a level of significance of 0.05 (this means we can expect a fluke result 5 times out 100), we can see that we’d need the correlation to be at least 0.44 before jumping to any conclusions. And even then, a correlation does not necessarily mean a cause and effect.
Y’know, I might persist at this until I actually find a correlation. In the meantime, don’t trust anything you read that boasts of relationships unless they give you the critical value of r. That’s some good advice.