The future is here, just not evenly distributed

Chiara Criscuolo at HBR discusses a new research report from the productivity group at the OECD.

Our research shows that the slow productivity growth of the “average” firm masks the fact that a small cadre of firms are experiencing robust gains. OECD analysis shows that the productivity of the most productive firms – those on the “global productivity frontier” in economic terms—grew steadily at an average 3.5% per year in the manufacturing sector, or double the speed of the average manufacturing firm over the same period. This gap was even more extreme in services. Private, non-financial service sector firms on the productivity frontier saw productivity growth of 5%, eclipsing the 0.3% average growth rate. Perhaps more importantly, the gap between the globally most productive firms and the rest has been increasing over time, especially in the services sector. Some firms clearly “get it” and others don’t, and the divide between the two groups is growing over time.


What this means is that the problem is not a lack of innovation it’s a lack of innovation diffusion. Note that a failure of innovation diffusion is also consistent with the argument that much of the rise in income inequality can be attributed to greater inequality among firms.

Why this is happening is unclear. Patents and other intellectual property being locked up in the frontier firms is one possible answer. Greater diffusion of ideas and people could thus increase innovation.Another possibility is that innovations are embedded in capital so you need new investment to diffuse innovation and business investment has been low for some time perhaps for “cyclical” reasons.

In one sense, these views provide some grounds for optimism. It’s easier to spread good ideas than to create good ideas. My Australia report discusses many levers we can pull to increase innovation diffusion.

If the future is here just not evenly distributed then increasing the speed of diffusion holds the promise of a highly productive catch-up period in which the average advances to the frontier.


Eric Brynjolfson at the MIT productivity project has been reporting this for years.

For a lot of firms innovation is not "hard" intangibles like patents, but "soft" innovation like better branding (internal, culture like recruiting training and retaining; and external - like pursuing clients differently and offering them more differentiated value added services), better operational processes and so on. This stuff is harder to understand and implement than a new hard techonology.

Part of the reason Goldman and Skadden Arps can charge higher fees is because they've carved out profitable niches with high barriers to entry and have a thick rolodex of clients that they wine and dine. So are these service-sector firms really more productive, or is it just market power?

My thoughts jumped to high powered finance and law firms as well! Perhaps the gap can be explained in part by the top firms wringing the most "productivity" from the poor associates. I'm pretty sure the gap between what's billed and what's paid to employees is biggest at the biggest firms, and that's all counted as "productivity," I'm sure.

My first thought was market power too. The example that came to mind was iPhones, which technically exist in a highly competitive market but for some reason enjoy significant pricing power. Surely that would affect measures of "productivity".

Like how the Beatles for some reason had pricing power when you could buy essentially the same goods from the Monkees at a much lower price. Consumers are not rational!

I'm not arguing that people aren't rational. Preferences are subjective and there is nothing irrational about one's subjective preference.

Consumers may be rational, but Apple enjoys pricing power, and to the numerical measure of productivity will be higher than the underlying technological growth. After all, Apple would still earn profits if iPhones were 20% cheaper.

One of the main lessons of Behavioral Economics is that those old words, rational and irrational, only work in a rough sense, as a rough guide.

As Damasio demonstrated in Descartes's Error, emotions create goals, and rational machinery is put in service, imperfectly, to achieve them.

That doesn't make every emotion and goal right or wrong, but I think self-knowledge adds perspective.

Do you really think the iPhone 6s is best? Or are you accepting a peer group value that it is "best" (Never knowing what a Nexus 6p is, what it does, or what it costs)?

The 13 yo in this house answers that a Rose Gold 6s is best, because it is what everyone at school has. Perhaps.

One clear answer that it is very good to be in Apple's current position, where their products are the only choice for 13 year olds, and the technology of competition simply doesn't matter.

I think that would be an example of the cumulative advantage I mentioned earlier.

"One clear answer that it is very good to be in Apple’s current position, where their products are the only choice for 13 year olds, and the technology of competition simply doesn’t matter."

Apple's iPhone technology is very good. So, I don't think you can conclude "he technology of competition simply doesn’t matter."

To what degree does this method capture cumulative advantage as "lack of diffusion?"

I did like the Australia report.

How is productivity growth measured? I assume by revenue growth per hour worked. Well, that would mean that the firms with high revenue growth would be high on the list of productivity growth, even if productivity had little to do with revenue growth. Firms that are very good at advertising, such as Facebook and Google, have high rates of revenue growth, but does productivity have much to do with it, or is it a new technology that is very good at Phishing for Phools. Similarly, WhatsApp has an almost infinite rate of revenue growth, but what's the productivity. Sure, the authors make a valid point, that "productivity" growth isn't widely shared across firms, but is that because firms vary considerably in productivity growth or because standard measures of "productivity" growth are outdated.

"How is productivity growth measured?"

It says so in the paper. OP method (Olley and Pakes, 1996). It's not revenue growth per hour worked. It's essentially a Cobb-Douglas production function, but with some other considerations.

Either way, it's a pretty well established method. The only concern I have is their frontier approach. If you're doing this by industry, within a country, and measuring the frontier as "top 50 firms", there's likely to be a problem. Top 50 firms in any industry are pretty They're not really the frontier, they are the firms which likely dominate 80% of the market in any industry, within a country. That's way to broad.

There's better ways of measuring this.

Take any point in post war history. Pick the top firms then go back and compare the cumulative difference to the average firm starting ten years ago. You will generate data that looks very similar to the above charts. There are probably many unique things happening in our time period (Frontier firms achieving higher market shares than in the past) but the highlighted data doesn't add much to the discussion.

Exactly. Post-hoc sorting is incredibly stupid - does the author pick stocks to invest in the same way?

I agree too. Any graph of this will probably look the same at any point in time. It shows nothing special.

A potential explanation, related to the dark matter problem:

Selecting for high productivity firms may be selecting for high "dark matter" firms.

Maybe because the productivity increases are partly due to shuffling labor intensive tasks elsewhere. For example we are doing asset management and much of the grunt work formerly done by payables. Only a brain dead bean counter with some strange motivation would come up with a system where low paid office clerks are replaced with highly paid technicians for the same work.

Another may be the fact that increased productivity in manufacturing displaces the common service tasks leaving more complex and time consuming one of tasks to the service companies.

As well highly automated manufacturing relieves the regulatory burdens of mass employment. The people remaining are highly specialized and far away from any marginal regulatory cost increase due to health and safety regulations or even health care c cost increases. The complicated hand assembly, factory refit or construction, etc, all more subject to higher regulatory burden are done elsewhere or by someone else.

Maybe productivity is something easily gamed.

"What this means is that the problem is not a lack of innovation it’s a lack of innovation diffusion."

That's a plausible hypothesis, but, as has been pointed out, there are other possible reasons for high productivity than innovation. Didn't Martin Shkell increase the productivity of Turing by raising the price of a Daraprim tablet from $13.50 to $750.00 with no innovation whatsoever? (Unless you consider price gouging innovation. ;)) Tom Murphy has recently pointed out that technological innovation during the past 65 years is much less than it was during the previous 65 years, so much so that someone from 1885 in an advanced economy would be much more bewildered by 1950 than someone from 1950 would be bewildered by today. ( )

In addition to locking up innovations via IP, it strikes me that the rise in non-competes and similar barriers to labor mobility (such as the Apple-Google non solicitation pact) is dramatically slowing down the diffusion of best practices across firms.

No. The thing is (just my opinion) with labour mobility talent is becoming increasingly more concentrated ...

Was this phenomenon happening in other decades? If so, isn't this just a statistical bias?

I'm not sure if I buy it, but as other commenters imply, isn't zero-sum rent capture a plausible alternative to non-diffusive factor productivity?

The latter implies that if the laggard firms adapted the practices of the fronteir firms, than both would benefit. The former implies that the gains of one come at he cost of the other, because the rents are fixed and finite.

Not that I agree with him, but Michael Church makes a fairly plausible argument that the bulk of modern Silicon Valley "value creation" is really just a glorified pyramid scheme in the form of underpaying software engineers through sexy companies and vague but unfulfilled promises of big payouts in the future.

Or is the past is not evenly distributed? Take Kodak as an example. The company knew that digital was the future 40 years ago, but it needed the cash flow from film to pay for decisions it made 40 years before that. Or any bank started in the last five years. It was a great time to start a bank, free from all those garbage loans made in the last decade.

100 bucks says I can get totally different results to theirs using a different method of measuring the frontier.

The "average" college football team wins 50% of its games. Some of them (FLA State, LSU, Alabama, Oklahoma, ND, Ohio State, etc) consistently win 80% or more. What does that prove?

Or to paraphrase Barry Goldwater 'there is always a top 10%'.

Wow, this is the most Tylerish Alex post I can recall

"...the problem is not a lack of innovation it’s a lack of innovation diffusion."
"The future is here, just not evenly distributed."


So, income redistribution is not enough. Innovation, productivity, growth, and the future in general, should also be redistributed.
Maybe IQ, beauty, happiness, love, too.

I guess the California Board of Equalization will have its jurisdiction drastically extended.

Earlier known as "the spread of innovation" in World Future Society futurist terms. The resistance to outside ideas was termed NIH - not invented here. GM was a good example, an engineer being quoted " our customers don't care about ergonomics".

The development of steam power and electricity caused the same divide between the frontier and the rest. Perhaps this issue is a feature of the system, not a bus.

Could they be observing the coincidental maturation of many manufacturing and service industries, leading to industry concentration/consolidation? That would not only concentrate revenues in a few firms but also would give them market power in some industries, allowing them to raise margins regardless of any technology or process gains. The lingering smaller firms would then bring the average down until they putter out of business. If that were the case the average would rise to the fronteir as those firms start dying out.

The assumption of the post that it is good to decrease "inequality" among firms puzzles me. Certainly this blog has made the argument in the other direction (correctly IMO) with respect to individuals. Why doesn't this data just show the sorting of winners and losers in the market? Why wouldn't we expect this? Would it be better if through instantaneous "diffusion" all the firms has the same productivity growth whether they were responsible for the innovation? Puzzling.

"The assumption of the post that it is good to decrease “inequality” among firms puzzles me "

It's a strange assumption, I agree. But you should not be puzzled by it.

KWSMTlDNJZeRisBckp 7533

Comments for this post are closed