Overstretching an argument


Over at Burning our money, they are abusing statistics:


Every so often I like to draw up the chart above. It shows the latest (2005) OECD figures on GDP per capita plotted against the size of government for each of the G7 major economies.

The negative relationship jumps off the page.

If we draw a simple line between the two, it says that for every one percentage point increase in the share of government in GDP, GDP per capita is lower by 2.9%. On average, for every one percentage point extra of governmnet spending they get, Italians are worse off by 2.9%.

That's one helluva price to pay for... er, whatever it is Big Government provides.

What's that? Laughably simplistic? Correlation does not imply causation?

Give me a break. Excluding the obvious outlier (the US), all I can see is a horizontal line. And even if there was a downward slopping line indeed, drawing inferences from 7 observations is at best non-sensical and at worst dishonest.

Now, Burning Our Money also quotes (hopefully better quality) OECD research:

As regular readers will know, the OECD itself has done stacks of work on this. And they've got people who have taken econometrics far beyond anything Tyler has forgotten from his LSE Masters. Their conclusion is:

"Taxes...seem to affect growth both directly and indirectly through investment. An increase of about one percentage point in the tax pressure... could be associated with a direct reduction of about 0.3 per cent in output per capita. If the investment effect is taken into account, the overall reduction would be about 0.6-0.7 per cent." (The Sources of Economic Growth in OECD Countries, Section 2.3)


0.6% of GDP for every 1% increase in government spending? So, if government reduces its share of GDP by 10% (a whopping change), all we lose is 6% of GDP - a mere two years worth of economic growth. These numbers, if anything, provide evidence for increasing government spending: the price we have to pay for doing so is very low indeed.

Now, let me make clear I'm not pro-big government, but I'm definitely pro-good argument. There are plenty of reasons to reduce the size of government. But please don't present graphs like the one above as evidence: they make a mockery of a good case.

1 comments:

  1. Anonymous Says:

    Is it not also problematic to plot GDP on both axes? When GDP per cap increases, the Govt spending/GDP ratio will automatically fall.

    Possible indirect effects include increasing income leading to households coming off benefits, thus reducing the ratio further.