Correlation and Causality

We've now got through the course content for econ101b in lectures, so enjoy the vacation! I thought I'd link up an article by John B Taylor, who has recently become a staunch proponent of government austerity. He's supported/encouraged by Greg Mankiw, who is another Republican poster child.

For those more averse to all things US, the Republicans are much more right wing than any of the mainstream UK political parties, although their closest equivalent would be the Conservative Party.

John B. Taylor is the man who proposed the Taylor Rule, a rule that monetary policymakers should follow when setting monetary policy. He strongly believes the sole cause of the Financial Crisis was that monetary policy was too loose, according to his model (scroll through old posts on his blog and you'll get a sense of this, I can't find the original article).

In the main linked article though, Taylor takes us through some scatter plots. Now these are interesting scatter plots - they show that government purchases are positively correlated with unemployment, and that investment is negatively correlated. From this, Taylor draws the conclusion that austerity is fine and should be encouraged (since it means lower government purchases) while at the same time investment should be encouraged.

This is all well and good, but a scatter plot shows correlation and not causality. Why does this matter? Well, Taylor proposes austerity (cutting back government purchases drastically) because in his mind it will cause lower unemployment. But what if the causality is the other way? What if high unemployment causes higher government purchases? Now this is hardly controversial really, since if people become unemployed, the government will have more to do: Higher benefits, likely higher other costs too, and naturally we might see some attempt by government to stimulate the economy by increasing purchases (data is 1990 on). If this happens, the purchases happen at the same time as the unemployment exists, hence we get a correlation like in the plot.

So does Taylor's plot really tell us much? Even if the government purchases worked, this plot wouldn't tell us that since it's a dynamic picture - i.e. the reduction in unemployment wouldn't necessarily come instantaneously! So the plot has ignored causality and also the dynamic nature of cause and effect in the macroeconomy.

It's another example of why it's very hard to know who to trust when doing economics. Blogs are great - they contain the opinions of top economists who can comment on real world events as they are happening. But they are not what we call peer reviewed. For a paper to get into a journal, it must be read by a number of referees who decide on its quality. Shoddy data work like that shown in this blog post, would not get past the referees and editors at a top journal.

The conclusion to draw - be careful, and in particular if you do read blogs (and I'd recommend it!) try to read a balanced selection of them - something like the list given on the right-hand side of this blog.