Tyler Cowen on the Great Stagnation, pro-progress policy, metascience, and more | Hacker News Reader

Tyler Cowen on the Great Stagnation, pro-progress policy, metascience, and more

fasterplease.substack.com

"There's a commonly noted productivity slowdown in the Western world starting in 1973 that I and some other people here have written about."

We don't remark on this enough, but it is truly surprising (and worrisome) that the era when computers took over the world has also been an era of failed productivity growth and stunted economic growth. Every prediction suggested that computers would take us in a happy direction: that computers would increase productivity. More so, the spread of knowledge was supposed to enable an increase in the speed of innovation. Simply having Wikipedia exist should have contributed to the uptick in some kinds of progress. But now, we look back on the last 50 years, and we can see how wrong all of those predictions were.

Productivity stalled in the 1970s and has never again regained its previous speed on a sustained basis. New business formation stalled in the 1970s, the number of new startups has steadily declined since then. We've seen the greatest wave of corporate consolidation in history: as recently as 1999 there were still 8,000 publicly traded corporations on the various USA stock markets, but now there are only 3,400 such companies. And economic growth has generally been weak, despite a few good years, such as 1995-2000. The period since 2000 has been especially weak, even as computers became universal.

There is no technology in history where the predictions were so wrong, compared to the actual historical result. The decades when computers took over the economy were supposed to be boom years, instead we now call them "The Great Stagnation."

Counter argument: we are averaging together two things, which includes a lack of natural world progress with massive deflation in anything computer bound. ChatGPT vs. healthcare costs. Taken together it is worse because the latter is more important, but there's been real deflation in amount of work per capital dollar.

You are simply describing the fact that when we speak of growth of the GDP, we are averaging millions of things together. How else would you measure economic growth in a modern, diverse economy? That some sectors boom while others fade is well known, but we only care about the final aggregate. That final aggregate did well for most of the period from 1870 to 1970 (with the exception of the 1930s). But since 1970, that aggregate has done poorly

> That some sectors boom while others fade is well known, but we only care about the final aggregate.

Why? I really don’t think it’s a good idea to try to measure something as complex and multi-faceted as the evolution of an economy with a scalar.

>> I really don’t think it’s a good idea to try to measure something as complex and multi-faceted as the evolution of an economy with a scalar.

Right? Even if we look at a single number, why not a distribution of that rather than just a total/average?

I think economics is an area that needs a LOT of work. Why we expect the fed to achieve 2 objectives (regulating the economy and inflation, and maybe unemployment) by turning a single knob (interest rate) is really strange to me. Meanwhile, congress thinks they can simply legislate outcomes...

GDP going down because things we want in the form of software getting cheaper is a good thing. The assumption that 'GDP going up is good' is not true in good deflation.

What happened is the productivity increases caused by the interwebs cause us to nestle into a local maximum (i.e., where profits are maximized as opposed to productivity).

I'm always puzzled why the basic underlying assumption in this context is "we should have limitless exponential growth in productivity". For millennia, there was barely any growth at all and the very concept of progress was dubious.

We have only so much population growth and people can work only so many hours a day (and many prefer to work less if they can). This way of looking at GDP growth figures and then going "hey, we have technology X, so why isn't growth the same as when we were rebuilding after the last war?" seems like a category error to me. Our lives are changing at an unprecedented speed and nobody can even tell what skills will be needed in the job market a few decades into the future. Is that what "The Great Stagnation" looks like? We should be concerned with how to stabilize the economy at the high levels of prosperity we have and how to enable poorer societies to achieve the same while surviving the challenges ahead. It's pretty clear to me that computers would be a great help doing that, if we really wanted to.

> For millennia, there was barely any growth at all and the very concept of progress was dubious.

This is known as the Malthusian Trap: https://en.wikipedia.org/wiki/Malthusianism

It's not that there was no growth for millennia. It's that people reproduced until any increase in productivity was eaten away by having more mouths to feed.

Once birth rates started to go down, and the industrial revolution happened - population growth stopped exceeding productivity growth - and living conditions started improving instead of declining.

As long as birth rates keep going down - I don't think we have to worry about Malthusianism returning.

But that's their very point: birth rates can't go down forever. The productivity gains of declining birth rates (which themselves are a product of scientific advancement and productivity gains leading to greater health and quality of life) might also be a relatively short-lived phenomenon. What if the population replacement rate starts to stabilize around 0? it's already happening in some places, and there are arguments to be made that is a desirable outcome.

Population growth stopped exceeding productivity growth when we figured out how to do everything with energy taken from non-renewable natural resources. We grow people from oil, cities from oil, etc. Malthusian limits may not be an issue as long as we have oil, but of course we also are making earth inhospitable to human life as we go.

That doesn't make any sense, with a larger population you can just grow more food as long as there's space. One person has been able to farm far more food than they could eat from the early bronze age onwards and population growth has been mostly linear since, up till the last century when it became exponential with mechanization making food ridiculously plentiful.

The problem with tech progress is that it's cumulative and you need to figure out the prerequisites first and with a small number of people. After that it just snowballs faster and faster. With a larger population you also get ever increasing chances for someone to invent something new.

IIRC until a couple of centuries ago a super high percentage of the population was employed in agriculture. Like 50 percent or more.

It was >60% almost everywhere until ~1600: https://ourworldindata.org/employment-in-agriculture

> That doesn't make any sense, with a larger population you can just grow more food as long as there's space.

A smaller population only lives in the most productive parts of their living space. A larger population has a lot more people living in the marginal parts, and if growing food is more difficult in them, the people on the margins will be doing more work all-in-all.

Meanwhile, in the more productive parts, some king, despot, warlord, or relative thereof will happily steal all the surplus.

> One person has been able to farm far more food than they could eat from the early bronze age onwards

Until textile production was mechanized around by the industrial revolution, you also needed something like 3,000 hours of manual labour a year to clothe a family. You needed someone to go do all the work of collecting firewood (As it turns out, all the easily-accessible-to-your-community firewood has been cut down a long, long time ago, so you'll be walking a fair ways to get it.)

Oh, and that king, despot and warlord? He will happily take every surplus penny, shirt, and turnip that you have.

And the one across the river will happily take what's left.

"After that it just snowballs faster and faster."

That is the theory. The problem is the actual historical record: the slow down after 1970. If the theory suggests that the process is supposed to snowball faster and faster, why did the theory fail?

I'm talking about technology progress specifically there which has continued to evolve at a rapid pace since then, even despite the lack of world or cold wars that would cause intense money throwing at problems. And as this post points out, productivity doesn't seem to always scale with technology. Especially when most of that technology is just passive entertainment that expends lots of resources just to make us feel less bored.

Regardless it's far too early to say it failed given that we're talking millennium sized scales where a 50 year slowdown or even dip is less than a blip. Of course real life also isn't a magical place where exponential graphs go up into infinity, so it has to plateau somewhere. I'd assume at the very least sometime after the local group of galaxies is colonized.

"people can work only so many hours a day"

All productivity numbers are per hour, so they are not effected by how many hours in a day people work.

"why isn't growth the same as when we were rebuilding after the last war"

Tyler Cowen, and myself, are both talking about the USA, which did not have to rebuild after the last war.

> All productivity numbers are per hour

which numbers? GP was talking about numbers of companies and the article is pretty low on data either way. In any case, the points I was trying to make were: there are eventually diminishing returns on individual productivity and looking at the number of startups is a very weird way to evaluate how useful a technology is.

> the USA, which did not have to rebuild after the last war.

So the post-war US economy should be directly compared to today's while completely ignoring the war?

The great distraction. A colony of humans that does not feel an external stressor, real or imaginary, will not seek to expand/evolve and instead spend its lifetime with dopamine. I welcome the new Cold War as an opportunity

> We've seen the greatest wave of corporate consolidation in history:

This here is the problem there used to be the physical barrier of distance, and associated time, these things used to delay the consolidation we are seeing.

this is also the years when real wages for workers dropped, disparity between the classes widened, and the concept of a company job pretty much disappeared. Maybe there is also some correlation there between these things and lack of productivity gains.

Perhaps the computer just allowed people to stay at the levels they were at when they were being rewarded for their productivity.

Frankly that would seem to fit with how the modern corporation tends to structure potential gains, use them to stay where you're at while cutting other things.

Except the incentives are unexpected. Back when labor unions were strong then labor was getting 75% of all productivity gains. In Belgium, from 1950 to 1970, labor got 110% of all productivity gains (so national income shifted from capital to labor). But after 1973, in the USA, capital gets 90% (and then later 100%), of all productivity gains. Given such a huge reward, you would think capital would then want to maximize productivity gains. It is odd that when capital finally got all of the gains from productivity, productivity growth slowed.

my suggestion is that productivity growth is in the hands of the people who must be productive. As they lost incentives they became less productive, but because the companies provided productivity boosters they were able to stay at the same level of productivity. In order for capital to increase productivity if this were correct they would have to give the shares of the rewards to the labor.

on edit: it seems weird to say we gave them all these tools to increase productivity, and they didn't increase productivity, but not note at the same time we removed rewards for increased productivity.

It's nearly impossible to discuss without the political flamewars - but Krugman's quote about the internet and the fax machine was in the context of some of his contemporaries predicting that we would consistently double our GDP growth due to the productivity gains from the internet. Cowen's Great Stagnation was a formalized book version making the same point with ~15 years more confirmation of the fact.

This 1997 essay written a year before the infamous fax machine article makes the full case and holds up very well: https://hbr.org/1997/07/how-fast-can-the-us-economy-grow

I remember reading some comment by Peter Drucker, sometime in the early 90s, that at the time he estimated that the entire edifice of computing, viewed as an investment by the global government, academic, military and business enterprise as a whole, could only demonstrate superficial, locally improved productivity but had never in fact paid off the investment overall and did not appear (at that time) to be tending toward it.

There's probably a tangent with the hot anthropological debate over whether agriculture has ever paid off for humanity as well.

I'm not sure how you measure investment in computing "as a whole", but I'm pretty sure this prediction is demonstrably wrong in any metric you choose. Imagine comparing a modern economy with that of a country that had not invested in computing at all - maybe Cuba would be the best example?

I'm not an economist, so I don't know what productivity means in a technical sense. But I think optimizing for productivity may be missing the point, if it simply means some rate of product creation with respect to human and natural resource inputs.

This I can see ending in the same old questions - what exactly did Plato mean by the good life, and how do we optimize for it as a species, or at least minimize suffering at the same scope.

Productivity is simply putting more machines at work for every person. More machines require more energy and resources. Since 1973 (and even 1971, remember when the US reached peak oil and immediately removed the peg between the dollar and gold?), energy consumption growth has been limited by different, compounding factors. Peak conventional oil occurred in 2008. Peak oil (all liquids) is announced for 2018-2025 (maybe already occurred, time will say). Peak gas is coming in the 2030s. The great slid down is ahead, and it won't be pretty.

I kinda look at this on a regular basis. There's a strange inverse curve between computing power, and its effect on organizational efficiency. Back in the 70s a big computer was a clear leverage, today it's a swampy cost point.

I think the answers lies at the anthropological layer, average groups don't know how to work very fast especially on non-linear tasks.

Application were also designed as dumbed down black boxes that people had to use without getting smarter themselves.

Computing also became a trend, lots of reliable, lean, scriptable as400 or win3 era apps were replaced by webapps, with less features (but more portable and cuter to look at).

The only way this improved was by moore's law, every N years users got stronger machines that managed to make a bit of the bloat faster.

As an example I like to tell that at a courthouse, when network failed and printers couldn't print, the productivity doubled. Filling simple templates was done faster by hand by simply chatting with users and writing things down as they come instead of asking for data, walking to your desk, waiting for the apps, typing data (in triplicate some times), waiting for the server to maybe generate the document, document that may still not have the correct data, then print it, then review it with the user.

Also, comfort breeds stupidity..

For hundreds of years, energy got steadily cheaper. Then in 1973 that progress stopped, and for the next 50 years energy prices have been roughly flat.

This is my hypothesis for the main cause of the great stagnation.

If I'm correct, then we're in for a reversal -- renewable energy has been steadily dropping in price, recently became the lowest cost form of energy and its price looks set to continue dropping.

> For hundreds of years, energy got steadily cheaper. Then in 1973 that progress stopped, and for the next 50 years energy prices have been roughly flat.

Fortunately, this is artificial due to OPEC - and solar and wind will free us from this eventually. The effects should be pretty dramatic somewhere in the next 10-15 years.

But this might be one of those situations where things get worse before they get better.

I think this is the key insight.

Top of my head, I can point to a couple of real productivity things from even cheaper energy:

* Vertical farming (Energy must be cheaper than the sun) * Space travel * AI * Cheaper everything as the value chain becomes cheaper

However, to enable this we also need to get rid of energy-conservatism which is prevalent now and, this is the important one, find energy sources with less negative externalities.

correct, pretty much every single issue plaguing humans at the moment could be trivially solved with infinite free energy. obviously that isnt coming soon, but as we reach that target things get easier and easier. and we have some significant tools that would get us there (more nuclear in general, potentially a breakthrough in fusion, and of course actually catching a non-trivial amount of energy the sun is giving off).

for example, would we ever have water problems again if we could de-salinate the oceans essentially for free?

I really hope that cheap energy enables a farming breakthrough.

Right now we essentially use approximately all arable land for food production. People see this and think we are on the edge of disaster, that if we had more people or land got less productive we wouldn't be able to feed everybody.

But the reason we use all the land is because that's the cheapest way to make food. Less intensive farming is cheaper. If we had more land we'd use it for growing food and food would be cheaper. If we had less land we'd use more fertilizer etc and food would be more expensive. If we had a lot less land we'd use a lot of greenhouses which can produce about 1000x as many calories/acre as a dryland wheat farm, and food would be ridiculously expensive.

But what if using more land wasn't the cheapest way to make more food? I don't think the answer is vertical farming, but it could be. More likely is precision fermentation IMO: https://www.rebootfood.org/

This is really exiting! Thanks for the link!

I completely agree. For me it is not so much about vertical farming either, more than it is the next industrialisation step in food production. Currently there is a plethora of candidate technologies that all need to be assessed and tried out!

Precision fermentation is, however, a new term for me.

This is also what I focus on. Energy is fungible for a great deal else in the economy. Energy is fungible for transport. Energy is fungible for food. Given enough energy, cheap enough, then energy is even fungible for salt-free water.

There are other possibilities, such as the breakdown of certain international trade systems, or less productive legal regimes (expanded judicial review and expansive due process) but energy seems like the most likely, most direct cause of the slow down.

> We don't remark on this enough, but it is truly surprising (and worrisome) that the era when computers took over the world has also been an era of failed productivity growth and stunted economic growth.

I have long suspected that computers help a ton in a few ways, but are net harmful in a bunch of other (more common) situations, and that organizations area really bad at telling which is which, while management is so enamored of the largely-useless "visibility" or "transparency" they provide that they can't resist computerizing All The Things.

Like, you talk to some office workers who've seen both, or who have the computerized system now but can still see how the paper version worked before, and you pretty quickly get the idea that this whole digital-workplace thing isn't all it's cracked up to be.

"I have long suspected that computers help a ton in a few ways, but are net harmful in a bunch of other (more common) situations"

I think this is key. Those of us who love tech tend to ignore how rigid it is. A database schema is rigid. We've gotten better at building tools that let us change a schema, but it is still rigid. The fact that computer programmers think a "green field" project is more fun than a legacy system should be a warning to us, that even those of us who love tech still hate how rigid and tedious each system is. And moving data around a large corporation means doing ETL work, which is almost anti-productivity: given the rigidity of each schema, and given multiple databases, we need to spend more and more time translating from one schema to another.

Or to put this another way, those of us who love tech tend to forget that corporations used to be run by armies of secretaries, and those secretaries allowed corporate systems to be highly flexible. A human secretary could:

1. pull meaningful data from an incomplete form

2. recognize high priority items without needing to be trained around rigid prioritization algorithms

3. easily understand a word even when it was misspelled (realize that "Albert Klien" referred to the customer "Albert Klein")

4. build commonsense hot caches (maintain easy access to the folders holding data for the most important clients) without needing complex programming. Or for lawyers, bookmark the most commonly cited precedents for the work the lawyer did. (Whereas several startups, in 2023, are still working to bring NLP techniques to precedent search.)

5. help customers navigate both internal and 3rd party bureaucracies

When we gave up on secretaries, we lost a lot of flexibility that we have only slowly and expensively been winning back with technology. Perhaps the cost of giving up on secretaries was underestimated?

And then of course, there are the ways that computers clearly damage productivity: Facebook, Instagram, Twitter; all distractions that burn up hours that people are at the office.

*Comment sent from your smartphone over WiFi. Technology which may as well be magic to the people of 1971.

1973 pretty close to 1971....

https://wtfhappenedin1971.com/

Debasing the currency was necessary, but not sufficient for what happened.

1971: World Population 3.7B

  |
  V

1972: Nixon opens US-China relations

  |
  V

1973: Kissinger's failure causes the oil crisis

  |
  V

Oil gets more expensive. Quarterly profits necessitate making up for the increased cost

  |
  V

Humans recast as "human resources": Outsourcing; wage stagnation

  |
  V

1989: Roger & Me

  |
  V

2023: World Population 7.8B; lots of cheap "human resources"

An alternative point of view is that computers provided a lot of increased productivity, but that the increased productivity didn’t reach the enormous levels of productivity increases caused by mechanization. In other words, it’s not that computers have been a dud (we’ve actually seen an impressive amount of productivity growth), but that mechanization was such a game changer that other technologies pale in comparison.

At least, so far. We should also consider the possibility that certain technologies might take years to fully mature. Look how long it took for mechanization to happen during the industrial revolution. For instance, it was about a century between the first steam engine and the first commercial steam powered train.

Peter Thiel talks about this also:

https://news.ycombinator.com/item?id=24607896

Computers help generating gigabytes if not petabytes of (garbage) content that is consumed by billions of people in HD+ 24/7. AI will soon increase that even further. Or that doesn't count as a productivity boost?

It's interesting that this doesn't mention the productivity / wage divergence graph that also splits at 1973 that shows that productivity still really outperforms wages. https://www.epi.org/productivity-pay-gap/

I tried to Ctrl+F-ing for some themes:

Not one mention of: "wages, income, poverty, inequality, etc." as progress to this group is purely a tech phenomenon, and "never" by policy failures or from any kind of class issue.

Here are some non-technical things that were mentioned though: "civilization, West, Venice, Florence, American, United Kingdom, Pentagon, etc" - telling...

I think the best explanation for what happened was globalization and its accompanying political movement in the West (neoliberalism). Productivity kept growing in the West at the cost of shifting labor-intensive jobs to China and other emerging economies. This in turn depreciated the value of labor back at home. With the majority of economic value being produced by large corporations and a small elite of individuals responsible for managing them, this elite was able to strong-arm the government into lowering taxes, eliminating tariffs and installing all kinds of preferential treatment to whoever could extract value of this global supply chain.

I was also disappointed that the discussion about politics and globalization was very superficial.

I had been reading Marginal Revolution and following Cowen for probably 5 years but I finally completely gave up on him.

At first I was deeply impressed and considered him someone who was unafraid to confront challenging issues with thoughtful analysis.

Now I see him as a Malcolm Gladwell style entertainer.

Anybody thinking of taking him too seriously should do some meta-research of their own before spending a lot of time on his ideas.

This reads rather damningly. There a rundown on why you are so negative on him?

I recognize the name, but I have not followed him directly. Amusingly, I thought he was more of an editor of folks with decent opinions, such that I don't think I know his.

His food blog on restos in the DC / NoVA area is pretty good. That's about the only thing he writes that I'd trust.

I'd be interested to read a version of this comment that substantiates its criticisms.

I've spent a lot of time thinking about him and his ideas over the years but I really don't want to spend more time on him honestly.

A few thoughts:

* His view of UFOs is very unscientific and sensationalist

* His view of Cryptocurrency has always been extremely credulous and I never saw him confront the failures and problems honestly (IIRC he loved SBF)

* His tendency to post very cryptic items ("Straussian") that you would only understand if you've been following him religiously for years is very frustrating

I do think his fast grants program is a pretty cool idea though.

I don't find this persuasive at all.

Thanks for substantiating. I still read him avidly but appreciate the points you made. I also rate UFOs being aliens as less likely than I believe he does. About cryptocurrency I have no interest or knowledge so I can't comment on that. As for the Straussian stuff I don't see the downside and am not frustrated. Agree that fast grants is the most exciting thing he's currently working on. On a personal level I might enjoy Conversations with Tyler more because I do think he's successful at his stated goal of "extracting the best stuff" out of his guests. Although more than a few of the episodes fall flat for me, the majority are enjoyable.

I think Tyler suffers from a form of the curse of knowledge....if one is 5x smarter / more knowledgeable than others, it can result in thinking one's opinions are necessarily knowledge.

One thing in particular that he gets wrong is that he seems to have a heavy bias toward ~neoliberal economic policies.

That's funny because I was reading him from perhaps 2006-2011 and came to similar conclusions. I was never sure if it was more because my economic views had changed, or because his writing has changed. I always suspected that latter was the larger factor

Some blame our divorce from sound money as the cause. This website is very insightful: https://wtfhappenedin1971.com/

How random - exactly as I saw this on the HN front page, I also saw Tyler Cowen's name in the subject of one of the unread emails from Derek Sivers.

What are the chances of that happening?

Reply on news.ycombinator.com