The Canadian Taylor Curve -- and the direction of causality

This is from my Carleton colleague Hashmat Khan:

There is a strong negative relationship between investment and unemployment in Canada. The data on private business fixed investment from the past 35 years confirm it. The measure of investment is non-residential structures + residential structures + machinery & equipment. The correlation between investment (relative to GDP) and the unemployment rate is -0.57.

This correlation is even stronger, nearly -0.8, since 1990. Without including residential structures, the correlations for the two periods are -0.40 and -0.73. And once consumer durables are included in investment, the correlation becomes even stronger: -0.65 and -0.83 for the two periods.

Figure 1: Investment-Unemployment Correlation in Canada: The 'Taylor Curve'

 

After investment, the component of Aggregate Demand that has the second largest correlation with the unemployment rate is net exports. This correlation also becomes stronger post-1990, moving from -0.10 to nearly -0.5.

The Canadian evidence on the negative relationship between investment and unemployment is consistent with the US evidence pointed out by John Taylor. Taylor suggests that the best way to lower unemployment in the US may be to promote policies that increase business investment.

While Greg Mankiw describes John Taylor's evidence as a "striking scatterplot", Paul Krugman argues that the housing bust led to a sharp fall in residential investment and this drives the negative correlation. Justin Wolfers argues that the negative relationship does not exist if one were to examine pre-1990 US data. Taylor rebuts these points in a later post.

So the evidence does establish a "Taylor Curve" for Canada. But a crucial question is whether investment causes unemployment or vice versa. Or maybe some other factor drives both?

Canadian data point to two things:

First, the unemployment rate leads investment by one quarter. The correlation between current unemployment rate and one period ahead investment rate is -0.58, slightly bigger than the contemporaneous correlation.

Second, the Granger-causality test rejects the hypothesis that past unemployment rates does not predict investment. It does not reject the reverse hypothesis, that past investment does not predict unemployment.

Taken together, these two pieces of evidence seem to suggest that in Canada a more likely channel for bringing about higher investment is via lowering the unemployment rate.

The above is by Hashmat Khan.

Comments

[Jim Rootham]

Makes sense, investment happens when customers start showing up with money. There should be a positive feedback component as well (more customers generates more employment generates more customers).

[Determinant]

There's also a link here, the investment is frequently in response to consumption in other countries, mostly the US. In manufacturing that is frequently the case.

Take a walk though any industrial park and count the number of American-oriented firms.

[Jon]

How do expectations fit into this model? Does the investment statistic report when projects complete? I would expect so. So looking for causality between expected investment and unemployment might change the conclusions.

[RSJ]

"The correlation between investment (relative to GDP) and the unemployment rate is -0.57."

If there is an increase in net exports, then the investment share of income, which includes NX, can be constant, or declining, even as actual investment is increasing. That could create the appearance of unemployment leading investment.

Perhaps it would be better if you measured investment as deviation from trend, rather than as a ratio.

[btg]

But isn't business investment also dependent on interest rates? and isn't unemployment also related to interest rates, in that if the B of C or the Fed tightens, the economy will slow down and unemployment should go up?

[tyronen]

Taylor has no specific policy proposals to increase investment, other than the usual conservative tripes of "reduce tax rates on small businesses" and "reduce the anti-business sentiment in Washington". What planet is he living on? Washington is among the world's most pro-business capitals. Wall Street is earning billions after having destroyed the world economy.

[Hashmat Khan]

@Jon: The correlation of investment (deviation from trend) and the unemployment rate is -0.55, and the correlation with one period leaded investment is nearly the same.

@RSJ: Yes, I agree that a variety "third" factors may drive the correlation. And you have pointed to one of them.

[Hashmat Khan]

Previous comments were for: @RSJ and @btg!

@Jon: I'm not sure. I'll have to check further on the details of how Stats Canada collects the information.

[Phil Koop]

"Taylor rebuts these points in a later post."

Please do not assume the conclusion; that completely vitiates your post. Taylor's post deals with pre-1990 correlations but does not in fact address the housing argument, or indeed any of the three posts Krugman made on the subject.

It may be that Krugman is wrong, but neither you nor Taylor have made any attempt to demonstrate this.

[Patrick]

Phil, what are you on about? I think you might need to read the OP one more time.

[Mark H.]

This is really interesting, particularly given the fact that changes in investment typically lead the business cycle while changes in unemployment lag it. The fact that the correlation strengthens with consumer durables is interesting as well, since it suggests the "investment" response to unemployment rates is not just limited to firms/employers (and their information sets), but households/employees as well... I'm curious how symmetric this relationship is (i.e., true both on the way up and the way down?), as it's hard to tell just from eyeballing the graphs. Worth looking into further, clearly. Nice post!

[Shan]

I've been missing this blog the past couple of days, just found out why: This post is breaking your rss/atom feeds. Both my feed reader and xmllint don't like it. There is a spurious ^L here, "private business ^Lfixed investment".

[Nick Rowe]

Shan: I cut and pasted from the pdf file Hashmat sent me. That might be what is causing the RSS/atom feeder problem. It did cause weird things to happen in the edit mode.

Funnily enough, I couldn't see the extra ^L you speak of, but when I went into Edit mode, the cursor was doing something weird at that point. So I deleted that bit and re-typed. Is it fixed now?

Has anyone got any ideas about why this post is breaking the feeds?

[Shan]

Yes, it's fixed now, thanks.

[Nick Rowe]

Shan: you deserve the thanks for telling us. It seems that glitch was holding up our posts from appearing in a number of feeds. Now it's freed up a logjam on Economics Roundtable.