One of the running topics here has been the relationship of markups and measured productivity growth. See prior efforts here and here. Those posts talked about how markups of price over marginal cost may skew the calculation of the growth rate of productivity. And I speculated here or there about how the falling labor share of output might show up in slower productivity growth.

Here I want to draw on some new evidence regarding the rise in the share of profits in output, a complement to the decline in labor costs, and how that might influence measured productivity. Evidence that the profit share is rising is important, because it doesn’t necessarily follow that a falling labor share lowers productivity growth; that was always just a possibility. With a rising profit share, though, I think you can trace a more direct line to lower measured productivity.

Evidence on higher profit shares and markups

A recent paper by job-market candidate Simcha Barkai provides new evidence that not only is labor’s share of output falling, but so is capital’s share. This combination implies that profit’s share of output is rising. And by profit, he means economic profits (i.e. rents) as opposed to accounting profits. Prior work documenting a falling labor share either punted on whether than meant a higher profit share, or implicitly assumed that it meant a higher capital share. Barkai makes the case that there has been a significant uptick in the profit share over time.

His first figure shows the decline in capital’s share of output. This isn’t something as easy to observe (unlike wages and employment), so there is some mild theorizing used to generate the measure of the return to capital he uses. But Barkai understands this, and a good chunk of the paper is spent exploring whether the assumptions he uses are driving this result. The fall in the capital share seems to be robust.

Capital's Share

The second figure here shows the implied rise in profit’s share of output. It is fairly close to a mirror image of the capital share.

Profit's Share

This rise in the share of output being paid out as profits seems consistent with some other evidence on corporate savings. Chen, Karabarbounis, and Neiman document that the source of savings in the global economy has flipped from being primarily driven by households, to primarily driven by corporations. Given that investment behavior stayed relatively similar over time, this means that corporations have seen an improvement in their net lending position. That is, they went from being net debtors (borrowing money from households to buy investment goods) to net creditors (using their own money to buy investment goods).


Their figure gives you the basic story. On the left is the ratio of savings to total GDP for each sector (corporations, households, and government). You can see the rise in the corporate savings rate from 1980 to 2015 from almost 10% to close to 15%, matched by a decline in the household saving rate from almost 15% to under 10%. On the right, though, corporate investment as a fraction of total GDP has remained constant (and note that by 2015 it is smaller than corporate savings as a fraction of GDP). Note that these are global numbers, not US specific. What CKN also document is that the expanded corporate savings are not being used to pay dividends or interest. They represent a build-up of actual savings (think Apple’s cash pile).


This is consistent with the Barkai result on profits. You would expect corporate savings, as a whole, to build up if the labor share was falling, which it is. But in addition, the fact that these extra savings are not being turned into interest or dividend payments would suggest that they are not just being used to pay for capital. I can’t prove anything here, but the rise in corporate savings fits with the profit share being higher. They do some further analysis with firm level data to show that for every dollar of additional corporate savings, about 32 cents ended up as cash holdings by firms (as opposed to share buybacks or debt paydowns). Again, consistent with the story of higher profits.

I think you could go in an interesting direction here by asking what this says about the nature of the principal-agent relationship of owners and managers. If firms really are earning more in profits, why isn’t that all turned back over to owners in the form of dividends or share buybacks? Do the managers benefit from keeping a large portion of these profits inside the corporation as cash?

But I’m not going to go in that direction. Rather, I want to see what this rise in profits implies about measured productivity.

The effect of markups on productivity levels

Firms generate profits by marking up prices over marginal costs. So if profits are higher, it implies that the markup has increased. To say something about the effect that higher markups could be having on productivity, we need some conception of why firms are able to charge a markup over marginal cost, and also some conception of what drives those markups.

For the why, we assume that firms have some market power based on …. something about them. Maybe they have a patent on a product, meaning they are the monopoly provider. Or maybe they have created some kind of brand, or reputation for quality (deserved or not), that makes people want to only buy that product. Think Apple. Or maybe they have some kind of locational advantage, like being a coffee shop in a busy train station. Whatever the reason, these firms have enough market power to charge a markup.

What determines the level of that markup, and why would it change? In general, the price elasticity of demand for their product is what matters here. The higher the price elasticity (the “flatter” the demand curve), the less market power a firm has, and the smaller a markup they can charge. The lower the price elasticity (the “steeper” the demand curve), the more market power a firm has, and the bigger a markup they can charge.

A high price elasticity would be the case if there were lots of close substitutes for the product, so people were willing to abandon a firm’s specific product if the price went up just a little. Think of gas stations, which have a tiny bit of market power due to location. But lots of people would stop going to a gas station that raised its price by 1 or 2 cents per gallon. A low price elasticity would be the case if there were few substitutes for the product, or the product was a strong complement to other things people buy. Then, no matter what the firm charges, people will buy some. Think of an iPhone. Apple has convinced people - whether you believe it or not - that there is no substitute for an iPhone; other smartphones just aren’t the same. So they can charge 600 bucks for one, while the marginal cost is something closer to 100.

When the evidence says that markups have gone up, then one interpretation is that the price elasticity of demand for different goods has gone down. I think of this as meaning that people have gotten very picky. I want an iPhone, not a Samsung. I want a double chai skim latte, not a coffee. I want a Lululemon pair of yoga pants, not the ones from Target. Just imagine that everyone has become a 13-year-old girl.

Now, if markups are higher because the price elasticity of demand for individual goods is lower, what does this imply for measured productivity? Well, in most typical settings, this would lower measured productivity. Why? The low price elasticity of demand comes from a low elasticity of substitution across different products. And a low elasticity of substitution across products means that most of our inputs (labor, capital) end up thrown at the least productive firms or sectors. Measured productivity gets pushed down in this case because of the concentration of inputs in these low productivity sectors.

If this is the story, then we should see this show up as Baumol’s “cost disease” symptoms. Inputs (capital and labor) should be moving into low productivity sectors, and the relative price of the output from these sectors should be rising. Here I think you’d appeal to the rapid rise in the price of child care, education, and health care, among others, to support the argument.

To sum up this line of thinking. The elasticity of substitution across products has declined over time. This results in (1) higher profits, as firms enjoy more market power when people are unwilling to substitute away from their products, and (2) lower measured productivity, as resources get concentrated in low productivity sectors. In this case, the higher profits are not a cause of lower productivity.

The problem with this explanation is that it depends on changing preferences. It isn’t clear why the elasticity of substitution between products within broad sectors would be getting lower. I can see that perhaps we are getting less willing to substitute between goods and services as broad categories. But what would be making the elasticity of substitution between different services decline over time? More effective branding?

Concentration and Market Power

An alternative explanation to make sense of higher profits and lower productivity would be the concentration of firms within industries. Let’s say that you have a number of industries: health care, transportation, manufacturing, education, etc. And preferences are such that the output of these industries are complements, with a small elasticity of substitution across them.

Within each industry, though, people are willing to substitute between different firms’ output. The firms engage in Cournot competition, meaning they try to pick a profit-maximizing amount to produce taking the others firms reactions into account. The basic result in a Cournot setting is that the markup charged by firms increases as the number of firms decreases. Fewer firms, less competition, higher markups. One could read the evidence of higher profit shares from Barkai as consistent with a decline in the number of firms competing with one another within industries.

And I think this would lead to lower aggregate productivity. For the industries with fewer firms, they would produce less output, and given a constant marginal cost as in a typical Cournot set-up, this means fewer inputs used. Because the output of industries are complements, we don’t like having less output of any given industry, so in order to convince the industry to produce more, we have to jack up the price we pay. A higher price for the output of a given industry is like a lower productivity level in that industry. And because different industries are complements, aggregate productivity is weighted towards the lowest productivity industries. Concentration in a few industries will thus drag down productivity in the aggregate.

I know that sounded a little tortured. I’m not sold I’ve got the logic right. But thinking of Cournot competing firms embedded in a larger economy turns out to be …. not easy. At some point I will try to digest some work by Peter Neary on this, and do a better analysis. For now, let’s just say that I think that concentration leading to lower aggregate productivity (and higher profits) is a working hypothesis.

Factor supplies and GDP

But wait, there’s more! Productivity might be affected by the rise in markups, but even if it was not, overall GDP is going to be affected through factor supplies. Higher markups and higher profits came at the cost of a lower share of output going to capital and labor. Which means that the return to providing capital and labor is lower. And if there is any elasticity to the supply of these factors, then presumably people provided less capital and labor to the market in response to the fall in capital and labor shares. Having less capital and labor would mean that total GDP (and GDP per capita) would be lower, even if productivity were the same. If productivity were lower as well, then the effect on GDP from higher profit shares would be even larger.

To sum the whole post up, a documented rise in the profit share (either directly or as an implication of the rise in corporate savings) could well explain part of the slowdown in productivity and GDP growth of the last 20-30 years. Putting a firm number on this is beyond my ability at this point, but it seems to be worth exploring.