Why are wages tracking productivity in Canada, but not the US?

A well-known curiousity of this business cycle is that US wages are stagnant, notwithstanding quite impressive improvements in worker productivity:

But in Canada, real wages are tracking productivity fairly closely:

The Canadian counter-example provides a useful check for possible explanations for the US 'disconnect'. For example, consider the hypothesis that US real wages are being held back by competition from low-wage countries such as China. This is a plausible story - we'd expect wages to converge eventually - but it doesn't square with the Canadian experience. If anything, we'd expect the effect to be even stronger in Canada, what with the 40% appreciation of the CAD against the yuan since 2002.

So the challenge is to come up with an explanation for the US disconnect that wouldn't also predict a disconnect in Canada as well.

Update: Comments here and in Mark Thoma's link to this post have asked for data sources for the Canadian numbers. The productivity series is real monthly GDP (CANSIM series v14182657) divided by hours worked as reported by the Labour Force Survey (v4391505). The hourly wage is labour income (v1996473) divided by hours worked. Unfortunately, it'll likely cost you CAD $9.00 to reproduce that graph. I also looked at other series - notably real weekly earnings - and I saw that they were also increasing.

Update²: A follow-up post on using the GDP deflator vs the CPI to calculate real wages.
 

Comments

[AC]

Part of the difference could be US employers are diverting wage increases to cover increasing benefits costs. Canadian employers don't have to pay for health coverage, right? (Or are taxes levied to cover it?)

[satch mehta]

Could it be the use of more extensive hedonic adjustments in calculating US inflation?

[outsider]

Or perhaps hard working migrants in the U.S. lower wages?

[outsider]

Or the commodities boom rewards labor intensive mines etc. in Canada?

[Laurent GUERBY]

Hi, what's your data source?

[Tim Worstall]

I would want to rootle around in the definitions of those statistics for my answer. Are they in fact measuring the same thing? There’s a lot of weirdness in the way the US measures wages.

[tom]

Wrong deflator for Real Compensation! Real Compensation uses the CPI All Urban Consumers as the price deflator. I would like to see the author put up another graph. This one would compare productivity to hourly compensation deflated by the implicit price deflator. If one did this, one would see that productivity and real compensation (deflated with implicit price deflator) go nearly one to one. I have done this with quarterly data from 1947 to 2006 and the series are nearly identical!

[Stephen Gordon]

I just used the series for real compensation per hour (RCPHBS) from the St Louis Fed. The correspondence between the two over time has been quite good, I agree. The puzzle is the disconnect since 2000. Menzie Chinn at Econbrowser has also made this point:

http://www.econbrowser.com/archives/2006/09/productivity_an.html

[tom]

The puzzle is resolved if one uses the implicit price deflator. If productivity is deflated with an output deflator then to be consistent compensation should also be deflated with an output deflator. To deflate compensation with a consumption deflator (the CPI) is mixing apples and oranges.

[Stephen Gordon]

You're not the only one to have raised the deflator issue; I'm looking into it. At first glance, it doesn't appear to be much of a smoking gun, but I'll have a post up on it fairly soon.

[tom]

Using a CPI deflator the disconnect is in 2000 because you have indexed the two series to equal 100 in 2000. If you use 1992 as the index year, guess what, the disconnect is in 1992 using the CPI deflator. The problem is the deflator. By using an output deflator there is no disconnect what ever year you index to.

[Stephen Gordon]

I'm confused. I didn't deflate the compensation series; I used the Fed's series for real compensation. I haven't rootled around in the documentation - do you know offhand how it's constructed?

[tom]

The real compensation series comes from the BLS and in their methods handbook it states that the real compensation series is deflated by the CPI All Urban Consumers. I will look for a link.

[tom]

"Real compensation per hour (RC) is computed as hourly compensation deflated by the seasonally adjusted Consumer Price Index for All Urban Consumers (CPI-U):

RC = (C / H) / CPI-U "

http://www.bls.gov/opub/hom/homch10_e.htm#Calculation%20Procedures

[Stephen Gordon]

That's great - thanks very much. I'll make it a point to note that in the follow-up.

[Tessa]

Just curious, but your graph directly contradicts the findings of a Canadian Centre for Policy Alternatives report on the same issue, which shows no serious increase in real wages since the 1970's despite climbing productivity. That can be seen here: http://policyalternatives.ca/documents/National_Office_Pubs/2007/Rising_Profit_Shares_Falling_Wage_Shares.pdf

Now, both you and them cite CANSIM tables as your source, but come to drastically different conclusions about the compensation Canadian workers are getting. I was wondering if you knew what accounted for the discrepency.
Thanks.

[Tessa]

sorry, i made a mistake with the link: http://policyalternatives.ca/documents/National_Office_Pubs/2007/Rising_Profit_Shares_Falling_Wage_Shares.pdf

[Tessa]

that's odd, it won't write .pdf. Well, just add .pdf to the end of that.

[Stephen Gordon]

There has been some real wage growth in the last 5-10 years. See this post as well as this one, which try to understand why. My conjecture is the the appreciation of the Canadian dollar is helping here.