IPPR seem to have a dedicated CPS output response wing broadcasting on Left Foot Forward these days. Today, their Senior Economist Tony Dolphin has ‘unravelled’ the Centre for Policy Studies ‘voodoo economics bigging up small government.’ He is referring to our report, released today, which provides further statistical evidence that small governments with low tax burdens are associated with higher economic growth. The beauty of the regression analysis that we undertook is that it enables you to a) control for other factors that theory suggests might be important, and b) obtain information from within countries over time.
Whilst I recognise that these sorts of blogs where people within the Westminster village debate amount to little more than navel gazing, I thought it might be worth responding to Tony in detail here (the italics are his words).
“This is not the first report to argue lower taxes lead to higher growth
Always good to start with an understatement. There’s a huge range of economic literature with evidence of this. See for example: Smith (1975), Hanson and Henrekson (1994), Engen and Skinner (1996), Barro (1997), Liebfritz et al (1997), Bleaney et al (2000), Folster and Henrekson (2000), OECD (2003), Pak Hung Mo (2007), Afonso and Furceri (2008), World Bank (2010), Afonso and Jalles (2011). In addition, there are many papers which look at the level effects: Koester and Kormendi (1989), Barro (1991), Cashin (1995), Bassanini and Scarpetta (2001), Alesina et al. (2002), Johansson et al (2008), Furceri and Sousa (2011).
“and it will not be the last. However, there are also many reports, also backed up by statistical analysis, that refute this claim. At the most basic level, in advanced countries there is no correlation between tax revenues (in relation to overall GDP) and growth in GDP per head. As Graph 1 shows, looking back over the last 40 years (the results are very similar if a shorter timeframe is used), and across the 22 OECD nations for which data are available, the correlation between average tax revenues as a percentage of GDP and GDP per head growth is zero.”
I’m glad that Tony has used the ‘at the most basic level’ term here. The graph that Tony has put on the blog is a pure cross section of 22 OECD countries since 1970. For each country, he has averaged the size of their state and their average rate of annual GDP growth per head across the full forty years. He has concluded that zero correlation between government size and growth here means that government size does not have an effect on growth in any particular country.
This, of course, is a complete straw man. By averaging by country, Tony has eliminated all of the within country information that our pooled cross section regressions are looking to examine. What’s interesting is to pick out how growth changes as government size changes within different countries given other variables within that country. That’s why our regression analysis split each country’s results into five year periods, and why Tony’s chart is meaningless.
So how does Tony’s chart change once you divide each country’s experiences into five year periods?
As the chart above shows, you get a negative correlation of magnitude of about -0.35.
The historical record shows that for every country with a small state and above average growth, there is a country with a small state and relatively low growth: Switzerland has the third smallest state in the OECD but also has the lowest per capita growth rate of all 22 countries. Similarly, countries with the highest tax revenues in relation to GDP also have a mixed record on growth. The divergent paths of Denmark and Sweden on the one hand and the US on the other are illustrative. In the early 1960s, tax revenues in all three countries were around one quarter of GDP. By the mid-1980s, tax revenues in the two Scandinavian countries were close to half of GDP, whereas in the US they were only a little higher than they had been in 1960.But per capita growth rates have been very similar in the three countries over the last 50 years.
Again, this is another straw man argument. As Matthew Sinclair has pointed out, there are a billion other things going on when you compare countries in cross section across huge time periods. All Tony has proved here is that small government doesn’t automatically mean high growth. Our regression analysis aim was to observe whether or not small government led to higher growth having controlled for a range of other factors.
The authors of the CPS report will reply that this is a partial account and that their analysis is more sophisticated because it takes into consideration other variables that theory suggests affect growth, such as levels of investment and labour force growth. This would be a valid criticism, were it not for the fact that their results (p.15) show there is no statistically significant relationship between any of these variables (other than the initial level of GDP per capita) and growth. This must have been a totally unexpected result and yet, surprisingly, there is no mention of it in the report. The authors could have concluded from their analysis that governments are wasting their efforts when trying to increase levels of investment in the economy because doing so does not boost growth. This would have attracted a lot of attention – but it would also have attracted a lot of criticism because it flies in the face of accepted economic wisdom.
Investment is clearly important in long-run growth. But what Tony has missed here is the fact that we include country fixed-effects in our regressions. This means that if there is little variation within a country over time for investment then we would expect the coefficient on investment not to be statistically significant – the country fixed effects dummy variable would be picking up most of the effect. That doesn’t mean that we are wrong to include the investment to GDP ratio in our regressions – it is worth checking whether there is still significant variation even after controlling for the individual countries within the specifications.
This result highlights the danger of drawing conclusions from regression analysis of this type. If the results suggest a relationship widely believed to be true – that higher investment spending leads stronger growth – does not hold, then any other relationship that is identified must be thrown into question. The explanatory power of the equations is also poor, suggesting important explanatory variables have been omitted from the analysis. Some of these may be correlated with the size of tax revenues and this throws further doubt on the purported relationship between taxes and growth.
The first part here is incorrect for the reason outlined above. On the explanatory power – with most fixed effects estimations of this sort you are lucky to get high R-squared values. But what Tony has failed to highlight here is that we include other specifications in our Appendix, including adding period fixed effects (for shocks affecting all advanced countries) and other measures that might be important, like the openness of the economy (which incidentally, has a strongly positive effect on growth).
Once these are added, the explanatory power of the regression increases significantly and the coefficient on the tax to GDP ratio gets stronger. The reason I chose to report the more basic specification result was because I did not want to be accused of adding variables merely to improve our result, which remained pretty robust following all these changes.
Many things determine per capita growth rates in advanced economies. Some people believe the level of taxation is one and would like robust statistical evidence to back up their view. This report, despite its claims, does not provide that evidence. A convincing case that a higher level of taxation leads to lower growth has yet to be made – probably because it is not true.
Our report certainly doesn’t prove that low tax to GDP ratios will always lead to high growth. However, it suggests that controlling for a range of factors within a country, reducing the burden of taxation as a proportion of GDP is likely to have a significantly positive effect on growth. IPPR’s knee-jerk reaction to this result shows the Fabian mind-set which the mainstream British left have settled into – unwilling to accept any critique of big government or large tax burdens. There is a tonne of evidence that bigger government leads to lower TFP growth and lower GDP per capita growth, given other variables. The fact they choose to reject all of this evidence on an evidenced-based blog is a shame.