My opinion of John Taylor has cratered since I began following his blog. This post is another travesty. The data are fine. The interpretation is atrocious.
"Sustaining growth in employment requires sustaining investment, which requires government policy that encourages investment and innovation, not short-run stimulus packages that try to boost consumption and government purchases, which crowd out investment." -- John Taylor (http://johnbtaylorsblog.blogspot.com/2010/11/end-of-recrudescence-of-keynesian.html)
- Government spending occurs in response to expectations of unemployment, so you'd expect a negative correlation which time series regression might reverse the causality of (ie - government spending is caused by expected unemployment which would look like it IS caused by future unemployment in a model). If it was as easy as throwing in a few lags, there would not be as much disagreement among economists on this point.
- The purpose of monetary stimulus is to lower interest rates and increase investment. A positive relationship between investment and stimuulus doesn't say anything about stimulus unless we know the source of the investment - it may be in response to stimulus.
- This is a little more speculative, but I'm also troubled by the vertical lines within the government spending/unemployment plot. It looks like we have two processes going on - one that is steep and negative and one that is more gradual and positive. For reasons stated in the first post, the causality of either of these processes is unclear.
- Let's say government spending has no effect on unemployment but because of deficit spending and automatic stabilizers it stays much more constant than investment spending. In a recession, where output (the denominator for these plots) goes down, you would expect government share of the economy to go up simply because it is less volatile than investment. You would get a positive relationship even if government spending did nothing to unemployment, which suggests you should be wary about interpreting a positive slope even if you don't take the first point I made about causality into account.
These graphs alone don't tell you much of anything.
John Taylor frustrates me because he does this a lot - he throws up a scatterplot or a raw number and calls it a day.
"Sustaining growth in employment requires sustaining investment, which requires government policy that encourages investment and innovation, not short-run stimulus packages that try to boost consumption and government purchases, which crowd out investment."
ReplyDelete-- John Taylor (http://johnbtaylorsblog.blogspot.com/2010/11/end-of-recrudescence-of-keynesian.html)
Daniel,
ReplyDeleteCould you elaborate a bit on why his interpretation is atrocious?
- Government spending occurs in response to expectations of unemployment, so you'd expect a negative correlation which time series regression might reverse the causality of (ie - government spending is caused by expected unemployment which would look like it IS caused by future unemployment in a model). If it was as easy as throwing in a few lags, there would not be as much disagreement among economists on this point.
ReplyDelete- The purpose of monetary stimulus is to lower interest rates and increase investment. A positive relationship between investment and stimuulus doesn't say anything about stimulus unless we know the source of the investment - it may be in response to stimulus.
- This is a little more speculative, but I'm also troubled by the vertical lines within the government spending/unemployment plot. It looks like we have two processes going on - one that is steep and negative and one that is more gradual and positive. For reasons stated in the first post, the causality of either of these processes is unclear.
- Let's say government spending has no effect on unemployment but because of deficit spending and automatic stabilizers it stays much more constant than investment spending. In a recession, where output (the denominator for these plots) goes down, you would expect government share of the economy to go up simply because it is less volatile than investment. You would get a positive relationship even if government spending did nothing to unemployment, which suggests you should be wary about interpreting a positive slope even if you don't take the first point I made about causality into account.
These graphs alone don't tell you much of anything.
John Taylor frustrates me because he does this a lot - he throws up a scatterplot or a raw number and calls it a day.