On explaining productivity

An oldie by Patrick Schneider at the Bank Underground: There are two productivity puzzles, 17 Nov 2016.

Two productivity problems: Schneider says there is a level problem and a growth rate problem. His analysis hits home because "levels and rates" are almost as central to economic analysis as "supply and demand".

Six graphs. He looks first at "Labour productivity relative to trend" long term, then just at the years after the GFC where a gap opens as productivity falls behind the trend. In the remaining graphs, Schneider analyzes the gap and finds that

... the level puzzle is almost all technology (Chart 5), but the growth puzzle is due to both slower technology growth and less capital per person employed (Chart 6).

In an update he adds:

Using the updated capital services series (chart below [previously Chart 6]), slower capital deepening can explain nearly two-thirds of the growth puzzle from 2011-2016 (compared to one-half previously).

It was somewhere along in there that it suddenly struck me: This is nothing like what I say about productivity. I say productivity will be rising soon, if it isn't already. I say productivity will rise when the economy starts to grow, and that the economy will start to grow when accumulated private debt starts to grow faster than GDP. And I say things are moving along right on schedule.

But I'm no economist. My expectations of improving productivity arise from graphs: from similarities visible in different time periods and different debt ratios. I look at debt-to-money ratios, debt-to-income ratios, and private-to-public debt ratios.

Schneider's evaluation of the productivity problem arises from "an aggregate, constant-returns-to-scale, Cobb-Douglas production function".

Schneider links to a page by Robert Solow, where Solow opens by saying "it takes something more than the usual 'willing suspension of disbelief' to talk seriously of the aggregate production function." Schneider uses it anyway. He rearranges the Cobb-Douglas and finds that

we can attribute changes in labour productivity to growth in either technology or capital depth, and we can also attribute deviations from trend to deviations from the trends in these two terms.

Based on the production function, Schneider attributes the fall in productivity to problems of technology and capital deepening. And his graphs seem to confirm this.

It is interesting stuff. I like the graphs. But is it right to take a formula some guy developed back in 1927, rearrange it, and conclude from the rearranged formula that there is no other factor influencing productivity? You would have to put a lot of faith in that one formula.

In the link Schneider provides, Robert Solow reminds us that the economy consists of production and consumption, both:

the aggregate production function is only a little less legitimate a concept than, say, the aggregate consumption function...

Supply and demand. Or as Keynes said:

there cannot be a buyer without a seller or a seller without a buyer.

Pretty much the whole trouble with economics today is that some people favor the supply side and others, the demand side. Some are Big-Endians; others, Little-Endians. Meanwhile, every egg comes with big and little ends. Every transaction involves supply and demand.

Cobb-Douglas is a "production" function. A supply side function. Demand concepts like the Kaldor-Verdoorn Law are ignored. Schneider's "simple growth accounting decomposition" shows clearly that slow productivity is caused in part by slow technological growth and in part by capital shallowing. From this, one could conclude we need policies that boost tech growth and policies that enhance capital deepening. But to reach such conclusions without also considering the demand side would be a grave error.


I don't mean to be overly hard on the Cobb-Douglas or the people who use it. I think that most valid explanations are valid, even conflicting ones. I like Schneider's story. I think it explains a lot, and provides good insights. But I don't think it tells the whole story. Here is his first graph:

Productivity Runs Flat Since 2008

Here is my graph from yesterday:

Non-Federal Debt Runs Flat Since 2008

See the similarity? Cobb-Douglas doesn't explain it.