Is the rate of scientific progress slowing down?

That is the title of my new paper with Ben Southwood, here is one segment from the introduction:

Our task is simple: we will consider whether the rate of scientific progress has slowed down, and more generally what we know about the rate of scientific progress, based on these literatures and other metrics we have been investigating. This investigation will take the form of a conceptual survey of the available data. We will consider which measures are out there, what they show, and how we should best interpret them, to attempt to create the most comprehensive and wide-ranging survey of metrics for the progress of science.  In particular, we integrate a number of strands in the productivity growth literature, the “science of science” literature, and various historical literatures on the nature of human progress. In our view, however, a mere reporting of different metrics does not suffice to answer the cluster of questions surrounding scientific progress. It is also necessary to ask some difficult questions about what science means, what progress means, and how the literatures on economic productivity and “science on its own terms” might connect with each other.

Mostly we think scientific progress is indeed slowing down, and this is supported by a wide variety of metrics, surveyed in the paper.  The gleam of optimism comes from this:

And to the extent that progress in science has not been slowing down, which is indeed the case under some of our metrics, that may give us new insight into where the strengths of modern and contemporary science truly lie. For instance, our analysis stresses the distinction between per capita progress and progress in the aggregate. As we will see later, a wide variety of “per capita” measures do indeed suggest that various metrics for growth, progress and productivity are slowing down. On the other side of that coin, a no less strong variety of metrics show that measures of total, aggregate progress are usually doing quite well. So the final answer to the progress question likely depends on how we weight per capita rates of progress vs. measures of total progress in the aggregate.

What do the data on productivity not tell us about scientific progress?  By how much is the contribution of the internet undervalued?  What can we learn from data on crop yields, life expectancy, and Moore’s Law?  Might the social sciences count as an example of progress in the sciences not slowing down?  Is the Solow model distinction between “once and for all changes” and “ongoing increases in the rate of innovation” sound?  And much more.

Your comments on this paper would be very much welcome, either on MR or through email.  I will be blogging some particular ideas from the paper over the next week or two.

And here is Ben on Twitter.


Is scientific literacy going up too? Or is there still only a small group of people that can understand the vast majority of complex science?

Devil's Advocate: would it not make sense that scientific literacy could be going down? The more science delves into the particular and the nuanced, the farther removed it is going to be from the lives of everyday people and more difficult the findings may be to communicate to the layman.

Ηello, yսp this piece оf writing is гeally good and I have
learned lot of thingѕ from it about blogging. tһanks.

Bellissimo lavoro,bravo continua cosi!!

In worst case it's stagnating, but there must have been an increase in number (which may or may not translate into a % increase). But it also depends on what you mean by 'complex science'. Science has never been as complex as today.

Sounds about right for a human endeavor in which the low hanging fruit has been picked? Definitely a change, but not one to worry about.

But if the tree is growing, why doesn't the ratio of low hanging to high hinging fruit remain the same?

A lot of modern social science is a case of science going backwards rather than just slowing down.

The quantity of data that social scientists like Raj Chetty have gotten their mitts on in this decade is astonishing. Unfortunately, Professor Chetty's insights into his vast dump of data aren't as impressive, but independent analysts have made a lot of progress at explaining what Chetty can't or won't.

The very name, social science, is an oxymoron.

You got that Right!!......just like Jumbo Shrimp and Government Service.....

An interesting research question! I'm looking forward to your paper.
If you want (that is, need) to include a discussion and, as I expect, working definition of terms like 'science", 'scientific progress' and so forth, you should widen your approach from "productivity growth literature, the 'science of science' literature, and various historical literatures on the nature of human progress" to the pertinent literature on the methodology/philosophy of science. Otherwise, you are bound to walk into a lot of walls once you try to implement these terms.
To give an example:
It's hard to quantify a scientific advancement like general relativity (Eddington's experiment was performed 100 years ago, by the way) as opposed to a technological improvement like a lithium-ion battery. I'm not arguing that you necessarily need to implement a framework like Thomas Kuhn's, but I think it'd be worth considering how philosophers of science (often scientists themselves) operationalize the difference between fundamental and incremental change. Especially, because it's often incremental change that's quickly converted into technological improvement.

> philosophers of science (often scientists themselves)

I disagree with this characterization. They often aren't, and philosophy of science in general has bad reputation among scientists.

Many of the most famous, influential, and innovative philosophers of science were themselves scientists or had a scientific education with at least some practical experience. Then, there are the scientists who 'dabbled' in philosophy of science and produced valuable work. Of course, there are also some very good philosophers of science without a scientific background. The same is true, by the way, of philosophers of mathematics.
Most scientists I've met over the years who are reasonably well-read in the philosophy of science (either in general of in their respective fields) are either interested in some of the foundational work done by philosophers or at least acknowledge that 'somebody' should do that kind of work. This is true, by the way, for both experimental and theroretical scientists (especially physicists).
Lastly, there are those scientists who may not hold philosophy in high regard but fail to realise that they themselves often engage in methodological or epistemological musings whenever they reflect on foundational issues, thus becoming philosophers themselves. That's always amusing - and a nice starting point for further exchange of ideas.
The only scientists I've ever met who disparage the philosophy of science in total - I'm not talking about particular arguments - are those who are largely unlettered or whose only window into the foundational issues was Richard Feynman - whom they tend to misunderstand.

This is the golden age of tech. Does this mean so-called tech isn't delivering the goods? Tech may be selling the goods, but tech isn't delivering scientific progress. We are banking the future on so-called tech, but it seems tech is stealing from the bank. Here's an interesting correlation, a negative correlation: scientific progress and inequality. Solve for the equilibrium.

Another article on the failure of America's business to invest:

Not without some pushback -
"The New York Times published a distorted and factually incorrect story on the front page of the Sunday, November 17 edition concerning FedEx and our billions of dollars of tax payments and billions of dollars of investments in the U.S. economy. Pertinent to this outrageous distortion of the truth is the fact that unlike FedEx, the New York Times paid zero federal income tax in 2017 on earnings of $111 million, and only $30 million in 2018 – 18% of their pretax book income. Also in 2018 the New York Times cut their capital investments nearly in half to $57 million, which equates to a rounding error when compared to the $6 billion of capital that FedEx invested in the U.S. economy during that same year."

Agreed. Not the best article to support the point that corporations aren't investing in productive capital as was supposed to happen as a result of, and to justify, the corporate tax rate cut - the article makes that point but then drifts off into an argument about tax fairness (i.e., FedEx paying zero tax) which became the focus of the article. Using record profits and tax cuts to redeem stock, make dividend distributions, and pay bonuses to executives won't produce scientific progress. The absence of investment in productive capital is not a new phenomenon, and it wasn't fixed by the corporate tax rate cut and the ability to expense such investment that was added in the 2017 tax act along with the tax rate cut. Scientific progress is slowing for the same reason productivity and wages have stagnated: the absence of investment in productive capital. If tax cuts won't do the trick, what will? That's a question that economists need to answer, an answer that differs from the answer economists have been repeating for years - more tax cuts. It ain't working.

The quote above says "....compared to the $6 billion of capital that FedEx invested in the U.S. economy during that same year."

$6 billion seems like a lot of investment, though probably little scientific.

"Scientific progress is slowing for the same reason productivity and wages have stagnated: the absence of investment in productive capital."

Scientific progress is NOT slowing - in fact it's exploding in virtually every scientific field. Having been in the field for almost 70 years, I've never seen the pace and diversity so exciting. What MAY be misleading you is the plethora of new (digital-based) consumer products that came on the market in the '90s. Those days are gone and are never coming back. The breakthroughs you'll see coming in the future will be in the background - new medical cures, new communication methods, new efficiencies in learning, ... And they won't be slaves to "investment in productive capital". They'll be the result of intelligent motivated individuals benefiting from the explosion of new knowledge and methods we're seeing every day.

Of course, the poster child for the failure of America's business is Boeing, which cut corners on the 737 Max in order to increase profits and share buybacks and executive bonuses. Given the disastrous results of Boeing's self-driving aircraft, I'm not confident in the safety of the self-driving cars being produced by so-called tech. Are you?

Golden Age of Tech? Yes, if "tech" is confined to better electronics and better software. And non-electronics tech that can be improved with better electronics.

Thus, we might get self-driving cars that are still powered by that fine 19th-century technology, the internal combustion engine.

The period 1880-1930 remains unique in technological change, due to widespread availability of electric power, devices enabled by internal combustion (cars, airplanes, steam->diesel), and the beginnings of electronics. It's difficult to see any such widespread technical change happening again (unless someone figures out how to build a starship, or how to extend the productive human lifespan well beyond a century or some way to store consciousness and individuality after death).

During the golden age of scientific progress owners of capital either invested in productive capital or consumed it (thus, the gilded age mansions). Today, we have markets in everything, markets for the present, markets for the future, even markets for the past (I made that last one up but with all the rewriting of history, why not). A speculator's paradise.

Surely no one actually thought that borrowing a lot of money to give to stockholders would induce corporations to invest more. Why does the MSM pretend to be surprised that this did not happen?

Late in my academic career I decided that the candidates I was interviewing for tenure track posts had better credentials than their equivalents twenty years earlier but seemed to be less intelligent and high spirited. If I was right, and if the phenomenon was widespread, it would be no surprise if the rate of advance fell. Science would become just another bureaucratic job.

Where the bright and high spirited go instead I don't know. Presumably a disproportionate number of such people - compared to twenty years earlier - would have the sense not even to do a PhD.

I would support this. Academia seems mostly about rent seeking nowadays, and the winners are the sloggers and the diligent. Someone willing to put up with the endless BS for the sake of a prestigious safe well paid job is probably not the person who will come up with genius ideas that will change the world.

"Where the bright and high spirited go instead I don't know." International Finance, it seems to me from my experience as a consultant.

The bright and ambitious go into finance, tech, start-ups, anything but academia.

Low relative pay, cliquish and political, ideologically skewed, bureaucratic.

A few machine learning researchers (often outside academia…) and genetics seem to be the only fields actually advancing these days…

Academia has become very bureaucratic these days. Also the curse of metrics - universities want to boost their rankings and you do that by having teams of faculty publish a lot of papers that at best are small advances on the margin. An academic who takes risk in their research will probably fall short on the quantity of published papers.

Per capita relies on human capital being equal. The demographics of the world have shifted and are older. For revolutionary scientific work I suggest adjusting by non-geriatric high IQ population.

The now dominant Theory of Intersectionality exalts the feelings of demographics who tend to not be very creative scientifically, but who do have a lot to say about their feelings about society's opinion of their hair.

Baby Boomers tended to have more high spirits and confidence in their freedom of intellect. Nowadays, in contrast, everybody is constantly aware that the Volunteer Auxiliary Thought Police are lurking.

Most boomers were prelingual during HUAC, if they were around at all.

If you are interested in genomics, the last few years have been enormously productive.

Progress in space science is a good proxy as it encapsulates much of science except biology.
Another proxy is body count, how many folks do we kill globally.

There's hope (on campus):

Nice spot. Here is a summary of the basic idea:
Several expressions for the j-th component (xk)j of the k-th eigenvector xk of a symmetric matrix A belonging to eigenvalue λk and normalized as xTkxk=1 are presented. In particular, the expression
where cA(λ)=det(A−λI) is the characteristic polynomial of A, c′A(λ)=dcA(λ)dλ and A∖{j} is obtained from A by removal of row j and column j, suggests us to consider the square eigenvector component as a graph centrality metric for node j that reflects the impact of the removal of node j from the graph at an eigenfrequency/eigenvalue λk of a graph related matrix (such as the adjacency or Laplacian matrix). Removal of nodes in a graph relates to the robustness of a graph.

Note they transform the symmetric matrix into a graph, likely a directed graph. Then local operations become local node exchanges and much of graph theory can apply. This is the same technique that Matt Damon used when impressing the math profs at MIT in the movie Good Will Hunting.

In graph theory and network analysis, indicators of centrality identify the most important vertices within a graph.
Finding the tree trunk in our abstract algebra tree.

The solution has a common form. A symmetric transformation matrix that preserves relative distance can be described as a reflection plane of dimension N-1, plus the summation of flux over the plane and that gets the hologram at dimension N. A very common theme, though I cannot say how widespread. It is in calculus a lot, computing a sphere from the flux passing through its surface, for example.

Something that impressed me is that the three physicists (Stephen Parke, Zining Zhang, and Peter Zenton) who "stumbled" (the word used in the article) on the mathematical discovery went to a mathematician (Terence Tao) for confirmation of their discovery. I say it impressed me because the discovery was then shared by all four. Is that kind of collaboration common? An aside, it seems that a couple of other scientists had previously "stumbled" on the mathematical discovery but didn't quite know the significance. Read the article (and subscribe to The Atlantic). It has many gems like this article.

I think network analytics and the growth of information is a good way to look at the question of scientific progress.

Think of quantum computing and CRISPR technologies today, and look back 35 years and the advent of the personal computer.

From these, discovering, managing and manipulating information produced more information, er, scientific progress.

CRISPR and Quantum are the new nodes from which more discovery will follow.

As we will see later, a wide variety of “per capita” measures do indeed suggest that various metrics for growth, progress and productivity are slowing down.

"Per capita" measures? Ergo, a hypothetical population of 100 should come up with twice as much scientific advancement as one of 50? That's simply statistical BS.

"Per capita" explains why the Industrial Revolution happened in China and could not possibly have happened on a little island at the edge of the Atlantic.

It also explains the advances in philosophy, drama, poetry, history, and so on in Egypt at around 480–323 BC, which couldn't possibly have happened at some Hicksville near the Aegean.

but did they have iPhones?

There are some pretty clear wins. I had earlier than normal cataracts, and now have remarkably effective plastic lens in both eyes. At the same time, a little computer controlled laser sculpting of the eye to remove astigmatism. Lifetime fix, a few thousand dollars. Not too many years ago, this now routine procedure and result would have been considered miraculous.

GPS - cheap, reliable and precise navigation. Mundane now, but near magic compared to what was available before.

Fusion power however, is still “in 20 years”, just as it was 50 years ago.

Fusion was always 40 years away in Britain; clearly we are a less optimistic mob than Americans.

My cynical theory on why stuff like fusion isn't available: When 'they' find someone working on fusion, 'they' pay the guy $20 million to cease and desists and place his work where the Sun doesn't shine.

Then, the researcher spends half his millions on booze, fast cars, and women; and wastes the rest.

From the short passages above (and just my general take on how people think of scientific progress generally) almost all metrics are "positive result" oriented. For example, calculating the size of a proton, measuring the expansion rate of the universe, finding a compound that support super conductivity at higher temperatures, machines performing calculations faster....

It does seem to be me that misses an important aspect of science: the ability to ask good questions. In other words, the ability to highlight what it is we don't know and so should be exploring.

While a provocative way to say it, it seems we might be able to say we have a pretty good handle on how we think about 5 percent of the universe we exist within works. Not sure where the line might be drawn so just making up a line here, but we might say 80 years ago we thought we had a really good handle on something like 80 to 90%?

Does that change in knowledge represent progress and if so is it more important progress than say knowing how to provide 5G communication technologies or highly detailed MRI scans?

I read the original research paper and then invested time reading other research papers on machine learning and AI using statistical tool sets interpreting fMRI results. I am not a statistician and know little of this subject area but was intrigued by the variety of approaches used by teams interpreting and the divergence of their findings.

Other research papers point out that the statistical tool sets are new and the fMRI data sets can be noisy. I wonder if these two factors explain most of the disparity in approaches:

To wit, "Pattern Recognition Pipeline for Neuroimaging Data?" by Sharaev, Andreev, Artemov et al identifies the noisiness of fMRI datasets and the need for a unified approach to data cleansing:

"Cleaning the artifactual (irrelevantto the process in question) noise incidental to scanning deems necessary, as such fluctuations drastically hurt recognition performance, blocking the way to the identification of neuroimaging markers for mental disorders."

The full paper can be found here.

Clarification: I posted my comment incorrectly to this article on scientific progress. I was referring to one of Tyler's recent links to the following research papers: "Variability in the analysis of a single neuroimaging dataset by many teams" found here:

The other issue is new researchers often stick with working down existing rabbit holes, so brand new approaches are rarely being investigated.

For instance software engineering research has experienced a 30-40 year dark age, and is only now starting down the path of creating a real engineering discipline (some data here:

I am sure that this will be (or already has been addressed) but having done some study of the history of science as an undergrad, I believe that the existence of previously unobserved phenomena is important and underrated in scientific progress. The most prominent example being the transition from a Ptolemaic, geo-centric model of the solar system to that of a solar centric model. When people in the late Middle Ages/early Renaissance started arguing for a geo-centric model, the reason they did so was based upon previously unobserved occlusions of some of the planets, which only become observable with better lenses (and the invention of telescopes). Ptolemy’s model could not explain those occlusions, which were explicable in a solar-centric model. However, Ptolemy’s system was still far superior in predicting the position of planets (and stars) in the sky!

Based upon this, I would think that scientific progress should be slowing down. The reason being that the past century or two saw an explosion of scientific devices that introduced previously unobserved (or under appreciated) phenomena into the purview of scientific enquiry. The apparatus of tools to control chemical reactions, and to manipulate, and observe (and measure) electricity and magnetism, and to observe cells and cellular organelles, were all incredibly momentous leaps forward in being able to actually observe the world. I just don’t see anything comparable in the past 30 years (or more), and frankly find it hard to conceive what corner of the world might possibly see a revolution like that in the future. Although as others have suggested, the social sciences have seen an explosion of data in the past couple of decades.

Quantum phenomena might be the one area where we just haven’t yet developed the tools to even see what is going on, and yet it seems the tools to do this (like super colliders) are incredibly expensive, so progress will likely be very slow. Perhaps understanding how the physical brain and body relate to the mind is another area where we are groping about in darkness due to an inability to really directly observe the phenomena we wish to study.

Peter Woit's "Not Even Wrong" backs up this idea: as accelerators reached their practical energy limits, new evidence stopped coming along that they had earlier had to get creative to explain.

The idea of a solar centric model goes way back, before Copernicus. But it was useless sense all navigation was only as accurate as an earth centered. The the chronometer was invented and solar centric navigation became useful.

But the point about the telescopes is well taken. Key instruments, chronometer was one.

"On the other side of that coin, a no less strong variety of metrics show that measures of total, aggregate progress are usually doing quite well."

That's been my argument here for years, that this aggregate, and rate of diffusion, matter far, far, more than per capita navel gazing.

The first time diffusion appears on the page?

Skimming the paper, I think what is clearly good news, a high rate of total innovation and rapid diffusion, is still painted as bad.

This is not a silver lining on a dark cloud, this is a sunny day.

Scientists around the world are working on every identified problem, and we hear their results faster than ever.

Global duplicate effort is at an all-time low.

I largely agree, but duplicate effort isn't bad. Different approaches yield different results, some better. In medicine, the 'me too' drugs are denigrate, but seemly same drugs work differently on different populations.

To add to this, the lack of replication would also suggest that duplication is good.

Well sure, but I was thinking, what if you lived near the current site of GMU, and the Chinese invented gunpowder. How long before your heard about it?


And then the hard way?

There are many fewer innovations lost for centuries today. And that puts everyone on a more equal footing.

The "social sciences" that are about creating justifications for government expansion ("programs") and political power grabs (wealth taxes, quotas, speech codes, etc) will never slow down. The sciences that are about explaining newly discovered fundamental phenomena have and will continue to slow down. Engineering will not slow down.

I take Tyler to be asking: is the politics of science and technology (S&T) becoming constrained?

My answer is "yes", for sound anthropological reasons, apart from and in addition to facing the pending threats coming our way from Technogenic Climate Change, the advent of which will even more severely erode the standing of S&T (and for sound reasons).

Humanity (per capita and aggregate positions) begins to see just how friendly science and technology have become. Humans begin to sense that S&T researchers are much more interested in the systems and processes and phenomena they study, and how these systems and processes and phenomena perform in theory and in experiment, to the virtual exclusion of "humans and their problems" viewed in the aggregate and/or in per capita terms.

The studies of humans and their problems seem not to enjoy the high social status that Tyler observes in the academic and commercial studies of systems and processes and phenomena not directly connected to the human condition. (When humans are considered in aggregate or per capita terms by S&T researchers these days, the attention seems focused on how to manipulate and shove humanity around.)

"Anthropology" is no science I would credit ANY "progress" to over the past two hundred years, Darwin & Co. notwithstanding: humans were animals before Darwin and remain animals after Darwin, despite all the excavations that physical anthropologists and paleontologists have made--and medicine for humans (at least in the aggregate) has found its level since Darwin increasingly on the plane of veterinary medicine.

To hear TC go on and on, "scientific progress" and the economics supporting it are practically the only matters that matter in this world today: "systems analysis" enjoys much more attention today from S&T researchers than all analysis of the human animals these glorious systems ostensibly are devised to serve.

How is it that Holy Science and its dedicated priesthoods have been so forgetful of humanity itself, in terms of BOTH the aggregate and the per capita?

A relevant science-satiric appraisal and fictional illustration of some of these themes can be found here:

I want to make an anonymous comment so I don’t upset people who might be easily upset.

I work in a semiconductor fab. The R&D group tells me that it is getting more difficult to make progress. Projects require more people and more time, but have lower return on investment.

The people in R&D are incredibly bright and motivated, so IQ and education are not the problem. I think we are seeing the slowdown in growth that Robert Gordon documented in “The Rise and Fall of American Growth “.


There are a lot of technologies in the world. At any given time some are beginning, some are trending, some are plateauing, and some are surprising us by breaking out of their plateau.

The semiconductor guys did their job, afaik, and $5 UNIX computers (Raspberry Pi Zero) are cheap enough. So have the wired and wireless bandwidth guys. AI could be a bit better, but that effort is fully funded (and in a breaking out of plateau stage).

I'd look to GMO to be the next century's growth segment, but I try to temper that. Many futurists expected the electronics to biologics handoff to occur much sooner. Decades ago.

Trivia: when you(*) were born, there was no computer on earth as powerful as a $5 Raspberry Pi Zero is today.

* - for most values of "you"

we self identify as a pair of 150 lb swedish wolfhounds
our pronouns are attaboy& gogirl
remember when dartmouth graduate & presidential candidate klobachar recently&boldly declared there was no such thing as "infanticide"?
infant defined as a babyv up to 1 year of age?
this is gonna open up a lotta new breakfast possibilities!

Cowen's paper includes: "Overall, consider Robert J. Gordon’s simple take: “U.S. economic growth slowed by more than half from 3.2 percent per year during 1970-2006 to only 1.4 percent during 2006-2016."

Cowen might want to include an update through 2019, which brings the 14 year average to 1.8%.

Also, how would one measure progress in quantum computing?

Why bother caring about this?

After all, the most brilliant scientist ever - Greta Thunberg - has conclusively proven that the World will end in 12 years. Party Hardy everyone!

I guess you can prove your manliness by fighting little girls, but at least fight fair!

"Our house is on fire. I am here to say, our house is on fire.

According to the IPCC (Intergovernmental Panel on Climate Change), we are less than 12 years away from not being able to undo our mistakes."

Imagine thinking you have to misquote her to win..

Little girl? I thought she was "a brave young woman."

Either way, she didn't accurately say what the IPCC has said about 12 years.

So, let me understand. Anon's misquote is justified because Greta was imprecise?

Fwiw, it sounds like she was closer than he.

The IPCC states the difference in impacts between going from the current 1.0 degrees to 1.5 degrees and 1.0 degree to 2.0 degrees is very small. There would be more extreme hot days, which are very rare, and they would go up 1.0 degree. Extreme cold days would also be warmer.

I take it that the meta question is why has p.c. income growth slowed down. The technology side of this is, has "technological growth" slowed down or has its incorporation into economic processes slowed down? Maybe a still more meta question is what is the ROI of more taxpayer funded R&D and of what kind?

How the Heck..can anyone... say scientific innovation is slowing down,?? Artificial Intelligence, Renewable Energy (eventually!!!) Almost Free Communications,, Driverless Cars, Quantum Computers, Nuclear Fusion (someday!!!), Space Travel, Undersea Exploration, Al.....even Better it's just Not Western Countries..but Countries across the world.... eventually some aspects of most of these innovations will bear fruit...just like electricity, electrical motors and basic computes took 20 years to be integrated....

Unsurprisingly, TC confounds science with technology. The plow, according to him, was a SCIENTIFIC advance. I find this risible. It's accepted that economists use value as a nearly universal metric in their work. The generally unstated assumption is that there exists a metric with which we can compare between 2019, 2000, 1950 CE, and 1950 BC. I don't know of any such metric (that I find reasonable)(even if we limit the time range to a century or so). I can compare, marginally, a dollar today to one I held last year, maybe even 10 years ago, but I can't think of any way to validate the comparison of a dollar held by me today, to the inflation adjusted 'equivalent' in coinage held by my great-great grandfather (I'd guess at least one of them was alive) in 1919. (that's related to the fact that a dollar I hold has a different value to me than one you hold has to you.) How would we compare the value of a cell phone to the one we (might have) had in 2000? Good luck with that. How do we compare our knowledge of transposons today to what "we" knew in 2000? The "value" (i.e. size, importance, significance,...) is necessarily context dependent. As the context changes, the value must change. In some ways, it can be seen as analogous to evolution, context evolves. (which should never be, but commonly is, considered to "progress") Is evolution predictable? No, not outside of well controlled environments. This implies, to me, that we can do a better post-hoc job of creating metrics than we can at predicting how we should value/measure future developments. Hypothetically, what would be the progress if we suddenly found out that there is no Dark Energy? Would that be forward progress or regression 'backwards'? Is the EPA's shift away from using all available scientific evidence in its decision making progress or not? Seems to me studies of progress will always be monday morning quarterbacking, and of dubious ( or no) value in improving/navigating our future. Here, I don't mean to imply that R&D expenditures can't be valued, in the context of opportunity costs, for decision making purposes. In other words, at the margin.

Really interesting paper. Thanks for sharing.

One interesting question, which was paper hinted at but didn't really expound upon in great detail, is what we are to make of the apparent divergence between aggregate and per capita scientific progress. Is it a "Red Queen" scenario, where we would be seeing a collapse in scientific progress but for putting far more researchers to work, or is it a case where there are limits to how fast science can advance and, for the number of researchers we have, we are seeing sharply diminishing returns?

The paper focuses on two main phenomena: the apparent recent slowdown in productivity (which I think is moderated by the most recent years of data) and apparent diminishing returns to progress in some specific areas. I'm not convinced that those phenomena are closely related. Sluggish productivity has many potential causes. I also see very big gaps between where we are and where we could be using existing technology, though such gaps are probably not greater (and if anything may be lesser) than in the past.

The paper itself observes that in focusing on specific areas for which there are long time series--crop yields, life expectancy, and information processing--may neglect the important contributions of younger fields. There might also be some sort of punctuated equilibrium hypothesis. Radical progress in these areas has historically been driven by certain paradigms: crop breeding, public health, and miniaturizing silicon-based semiconductors respectively. Each of these methods of progress shows signs of diminishing returns. Each of these areas might see dramatic progress through new paradigms, such as, for instance, greenhouses, gene therapy, and quantum computing respectively. However, industry might tend to underinvest in the new paradigms, which multidecade lead times, as long as the long paradigms are bearing fruit. Hence a period of stagnation that we can hope will be temporary.

One can draw an analogy to peak oil predictions of the 2000s. There was in that decade a period of slow growth in crude oil production and a price rise. The 2010s saw hydraulic fracturing and shale oil come to maturity. High prices, and a lack of sufficient return to old technology, were needed to motivate the market acceptance of new technology.

Anyway, great paper. I'm looking forward to seeing the successor on the institutions of science.

"American TFP typically grew quickly in the 1920s and 1930s, ranging from between two to slightly over three percent per year. In more recent times, in contrast, TFP growth often has ranged between one and one and a half percent." p. 15.

Again, Cowen and Southwood aren't reading Gordon's graph correctly, which is reproduced in the paper below. "1930" gives average TPF for the 1920s, which was 1.3%, not 2% as is stated and the average TFP for the 1930s was 1.8% not just over 3% as is stated.

The top of the graph explains this: "The average annual growth rate is for the ten years prior to the year shown." ("2014" covers the average from 2001 to 2014)

They just need to change "American TFP typically grew quickly in the 1920s and 1930s" to "American TFP typically grew quickly in the 1930s and 1940s"

Yeah from your draft things look pretty dismal. You may want to include data on how much money it takes to bring a drug for approval too to hammer your point.

The Flynn Effect debates. I find it curious no one has mentioned that so far. To the extent that scientific genius is likely to be the most vulnerable to Woodley Effect (Or Lynn/Vanhanen or whatever you want to call it) 'Anti-Flynn Effects, then a stagnation of scientific genius seems highly relevant to those debates. Taking some of the issues from there and reading the Cowen Southwood paper from that perspective several thoughts emerge. Of course there is the usual demographic fertility transition issue of persons with higher IQ's delaying having children and lower fertility rates. (Various estimates/models are out there. Say, 0.4 IQ points per decade at the global level.) But more particularly in the context of the paper things like leaded gasoline seem to loom large as possible variables with a large degree of explanatory power. There are more possibilities of course and I'll just toss in one example- Mass public education, from the perspective of young geniuses, meant a considerable dilution dumbing-down effect of the teaching in High Schools. And then later on with universities as well. Genius development delayed is genius .. stunted.

Might productively gains in crop yields be depressed by anti-GMO policies in Europe and/or trend to organics in developed countries?

Comments for this post are closed