Re: your bubble point at the end. There are two scenarios and I’m trying to figure out which is more true:
1. The work at these corporate labs was incredibly valuable for the world but was hard for the companies to capture the value of, so rationally they stopped doing it once they figured that out.
2. The work these corporate labs was on net not that valuable at all, despite a few charismatic successes.
Yeah, this is a key question. I think you can probably make a strong case for 1) based on a lot of these discoveries being key for other, later discoveries (it seems like ultra-sensitive magnetic field detectors and electron microscopes are the sorts of thing that unlock a lot of subsequent scientific progress). But I really don't know.
If (1) is accurate, it seems like that argues for boosting government investment in this type of research, to make sure that this non-excludable good gets produced at the optimal level.
(I imagine the Institute for Progress would have Opinions on this topic.)
This presumes government investment is well targeted, instead of being captured for other purposes. Much such investment seems to go into long term projects of questionable value, with the policy formulation process being subverted to sustain interests that have coalesced around these projects.
There was also the matter of tax policy. Corporate income taxes were about 50%, so the government effectively paid half the bill for a research lab. Personal income taxes could be as high as 90%, so shareholders would rather fund research on the long shot chance it paid off big rather than giving the government all but 10%. When they started cutting taxes in the 70s and even more in the 80s, research labs were no longer as good a deal as a company could just zero out research and cut a check to shareholders.
P.S. In the 1960s, Ford developed the first intercalating battery using sulfur and alumina. They even developed a prototype electric car. The downside was that the battery had to run at 450F. There are all sorts of premature technologies like this that got put on hold for decades before making a splash. For example, researchers at Pfizer demonstrated that GLP-1 cut sugar levels in diabetics, reduced hunger and slowed gastric emptying back in 1990, but that family of drugs wasn't rolled out until recently.
There is no reasonable argument that the corporations that made the capital expenditures on these R&D labs were able to fully internalize the economic benefits of the innovations produced, but the economic benefits are real and profound. For me, a very interesting question is *why* more of the economic benefits were not internalized by the corporations that funded the R&D.
One answer has to do with market timing, and in particular synchronizing the development of a new technology and commercialization of a new product with the gradual obsolescence of an old technology and old products. The disincentives of a monopoly on the old product to invest in development of a new product have gone by many names, but I tend to think of Kenneth Arrow and the Replacement Effect.
Having lived through the tailing off of investment in corporate R&D in the 1990s and made several personal visits to IBM Almaden Research Center at that time, another observation that's relevant here is that most of the researchers who were working at those corporate R&D labs moved to academic appointments at research universities, which were simultaneously — thanks the Bayh-Dole Act and some early successes in biotechnology — building out their technology licensing teams. For researchers, an academic job that allowed — even encouraged — both publishing and patenting was a great deal compared to working for a for-profit corporation R&D lab where the emphasis tended to be more on the latter. Not to mention the fact that the constant flow of new students through their labs could be more fun than seeing the same old faces day in and day out at the corporate labs.
And perhaps not surprisingly venture capital backed startups were taking off around the same time, many of them seeded with not only the capital from funds backed by research university endowments but with intellectual property and talent from research university labs.
It's hard for a for-profit corporation at any scale to compete with the pipeline of venture capital backed startups working in collaboration with federally funded and tax advantaged research universities. Empirically, a geographical cluster of a couple of top tier research universities combined with a healthy local market of talent, venture capital, and professional services seems more capital efficient than Bell Labs at its heyday.
The problem with VC is time horizons and portfolio size.
To take long steps through a fitness landscape of ideas, sometimes you need to traverse valleys or you need to take a step without knowing whether the step will yield gradient ascent. VCs aren't able to fund either of those things, because they have short horizons for the funds (so they need to generate 25% within 3-7 years) and because their funds aren't large enough to let true portfolio effects dominate (i.e. they're sampling a very limited distribution where they want every company to return threshold returns within that horizon, rather than being able to take bets on things where they simply don't know whether it will yield something).
That's why we still have government funding for so much fundamental research: commercial incentive structures and markets are quite bad at doing things that don't appear to yield 15+% IRRs within 5 years *as assessed at the time of investment*, and if that's your search strategy, you're sampling a very small set of the underlying distribution of possible critical inventions indeed.
If you can't compete with VC-funded startups, join them. Xerox decided it couldn't internally develop the work of PARC, so it encouraged the employees to spin off startups. At least, it said that was the plan, but it only managed to get one pair of people to do this, founding Adobe. Talking about efficiency is premature when it's just not happening at all.
To add to your list of Major Bell Labs Developments; they basically developed and standardized the way you interact with telephony and other information systems globally, which has enormous ramifications to this day on the omnipresent devices we keep in our pockets, as well as what we'd more generally term "Industrial Design" but then would have been a sort of psychological engineering.
In 1983 I hired into Bell Labs Research (Area 11); Materials Physics. I came from 2 years in a startup founded by 2 guys from Area 11 and was very familiar with Bell methods, processes, terminology. I came in as an STA - Senior Technical Associate. This was, generally, someone with a Bachelors in Physics, Materials, Chemistry and no experience required.
The fundamental Area 11 researcher was the Member of Technical Staff - Principal Investigator. They were a PhD in Physics, Materials, or Chemistry from Stanford, MIT, or Cornell. Current PIs would stay in touch with their old groups and ask who the top 2 people were from that year - thesis topic not important. Like the sports team that drafts the "best available athlete". When hired they'd be given someone like me - it was very much a Professor / Grad student relationship, albeit with good pay and benefits. An STA could take as many courses at night as we wanted; so long as you didn't get a PhD - then you would exit research. It was a tacit understanding none of us had an issue with; we didn't want to "dilute" the PI pool.
Area 11 was remarkably flat. You could walk into anyone's office, plop down, and talk shop. Arno Penzias, VP of Research, had a rule that everyone - from him to secretary - was on a first name basis. "It's not important where you were - only that you're here." Arno also remarked that if one of his people won a Nobel Prize in particle physics he would be embarrassed, as there was a general sense of where the research arrow should point.
Executive Director Kumar Patel (X-ray laser) said "If you're not failing 70% of the time, you're not asking the hard questions .............. but that other 30% better be world class." It started out that Research should be thinking 10 years down the road ..... then 5 .......... then 3 .......... then "what product do you support" ? That's when people began leaving. The stated 60/40 split of hard science/software was going to be reversed. PIs were told 'If you can do your work at a University you should go do it." By the mid-90's it was over.
The reason that labs like this are less common now is that there's an incentives problem: it's established in the economic literature that the fraction of the value of an invention that companies capture is low.
There's a compounding, perverse effect that when an inventor makes something that's particularly valuable (i.e. perceived to be particularly valuable by society at large), there's a negative feedback loop, as with GLP1 inhibitors and mRNA vaccines: the public's instinct is "that's extremely valuable to me, but also I don't want to pay what it's worth to me in order to have it, so the government needs to regulate prices".
They're the primary reasons why, as societies, we have so little fundamental research: because as societies we have a deep instinct to dislike seeing someone who's done something extremely valuable to society actually profit from it; as societies, the balance of IP protection and the ability to enforce rights vs. allowing others to develop a new technology isn't quite right (so firms can't capture enough value from innovation); and because, as societies, we have deep instincts for price regulations, which is an instance of concealed self-interest in wanting valuable things without actually having to pay for them.
The problems of inventions taking years to commercialise and the rate of success being low are more solvable: they can be done solved via "portfolio effects", whereby you look at 100 programs of fundamental research and maybe only 10 of them yield a commercial product, but you know that, on average, your "portfolio" still yields a 15% IRR or whatever. This is basically what VC firms and biotech firms do every day.
In sunmary, if you're a company and you think you can make commercial returns as an R&D lab like Bell Labs by doing fundamental research then commercialising or licensing it, you'd do that all day long. The fact that we don't see that anymore is an indicator that our political mechanisms for compensating invention aren't quite right, so that companies have defaulted to taking small steps - incremental things that can be commercialised quickly - through the option space.
We know that these labs can produce things of extreme value - everything on that Bell Labs list is a critical modern technology. We just aren't letting the inventors have enough of what they just developed.
Great point about incentives - One example is radio. In the 1920s and 1930s the development of radio caught the eye of congress and the US bought the rights for a complete set of patents for AM radio, and later kept going with FM radio and B&W/color TV. Congress (I think it was Congress) formed a "clearinghouse", RCA, where a manufacturer could go and buy a single license, allowing the production of a radio/TV with competitive technology.
This leveled the playing field for companies. Innovation continued but consumers had a variety of reasonably prices sets to choose from.
This model is capitalist enough to be interesting, but the RCA deal probably needed a better payout for the original innovators to justify a lab model. (Still making your point here.) We'd never get to it now without a crisis and a breakout technology that looked to have staying power. People/companies are so intent on their own ways of beating the other players, they never consider some place for an agency to manage the commons. We still have NIST, but they've been pushed back to the most fundamental things. Forget a set of more secure but also rugged protocols for communication networks...
I don’t think I’ve commented here before, but I’ve been reading your articles for a while, maybe 6 months, maybe a year. Really appreciate your excellent work and deep dives. Thanks!
Bell Labs also contributed hugely to human perception and memory research, which isn't on your list. I work in speech and hearing where ALL the fundamental knowledge comes from Bell. They also did the same for vision and other senses and memory. After they closed down, there's a lot less progress in these areas.
The influence spread to smaller companies as well. Before Wall Street bombed the country down to bedrock in the 80s, oil refiners, oil service firms, and wheat processors had major research labs in smaller cities like Enid and Ponca City and Bartlesville. All are gone now thanks to centralizing and financializing.
I can confirm from personal experience that PARC became more focused on product development in the 1990s. I worked at the Xerox XSoft division, which was literally down the hill from PARC. Our job was to take research projects and try to turn them into products.
Even though the scientists and engineers at Bell Labs didn’t apply for grants, I imagine there was a healthy competition for ideas to fund (or at least to choose how much to allocate to each idea, which is probably what happened there?). Do we have access to any proposals for bell labs? For example, I read that Shockley’s transistor idea and George Mueller version of it were going on in parallel so there must have been some forum for discussing ideas before diving into it.. maybe there was some other avenue for it?
I’m personally familiar with PARC. It had a lot of diversity in some demographic areas: people from all over world, gay people, religious diversity. Not so much racial diversity.
Re: your bubble point at the end. There are two scenarios and I’m trying to figure out which is more true:
1. The work at these corporate labs was incredibly valuable for the world but was hard for the companies to capture the value of, so rationally they stopped doing it once they figured that out.
2. The work these corporate labs was on net not that valuable at all, despite a few charismatic successes.
Yeah, this is a key question. I think you can probably make a strong case for 1) based on a lot of these discoveries being key for other, later discoveries (it seems like ultra-sensitive magnetic field detectors and electron microscopes are the sorts of thing that unlock a lot of subsequent scientific progress). But I really don't know.
If (1) is accurate, it seems like that argues for boosting government investment in this type of research, to make sure that this non-excludable good gets produced at the optimal level.
(I imagine the Institute for Progress would have Opinions on this topic.)
This presumes government investment is well targeted, instead of being captured for other purposes. Much such investment seems to go into long term projects of questionable value, with the policy formulation process being subverted to sustain interests that have coalesced around these projects.
There was also the matter of tax policy. Corporate income taxes were about 50%, so the government effectively paid half the bill for a research lab. Personal income taxes could be as high as 90%, so shareholders would rather fund research on the long shot chance it paid off big rather than giving the government all but 10%. When they started cutting taxes in the 70s and even more in the 80s, research labs were no longer as good a deal as a company could just zero out research and cut a check to shareholders.
P.S. In the 1960s, Ford developed the first intercalating battery using sulfur and alumina. They even developed a prototype electric car. The downside was that the battery had to run at 450F. There are all sorts of premature technologies like this that got put on hold for decades before making a splash. For example, researchers at Pfizer demonstrated that GLP-1 cut sugar levels in diabetics, reduced hunger and slowed gastric emptying back in 1990, but that family of drugs wasn't rolled out until recently.
The evolution of R&D into its present form is also a story of tax policy.
There is no reasonable argument that the corporations that made the capital expenditures on these R&D labs were able to fully internalize the economic benefits of the innovations produced, but the economic benefits are real and profound. For me, a very interesting question is *why* more of the economic benefits were not internalized by the corporations that funded the R&D.
One answer has to do with market timing, and in particular synchronizing the development of a new technology and commercialization of a new product with the gradual obsolescence of an old technology and old products. The disincentives of a monopoly on the old product to invest in development of a new product have gone by many names, but I tend to think of Kenneth Arrow and the Replacement Effect.
Having lived through the tailing off of investment in corporate R&D in the 1990s and made several personal visits to IBM Almaden Research Center at that time, another observation that's relevant here is that most of the researchers who were working at those corporate R&D labs moved to academic appointments at research universities, which were simultaneously — thanks the Bayh-Dole Act and some early successes in biotechnology — building out their technology licensing teams. For researchers, an academic job that allowed — even encouraged — both publishing and patenting was a great deal compared to working for a for-profit corporation R&D lab where the emphasis tended to be more on the latter. Not to mention the fact that the constant flow of new students through their labs could be more fun than seeing the same old faces day in and day out at the corporate labs.
And perhaps not surprisingly venture capital backed startups were taking off around the same time, many of them seeded with not only the capital from funds backed by research university endowments but with intellectual property and talent from research university labs.
It's hard for a for-profit corporation at any scale to compete with the pipeline of venture capital backed startups working in collaboration with federally funded and tax advantaged research universities. Empirically, a geographical cluster of a couple of top tier research universities combined with a healthy local market of talent, venture capital, and professional services seems more capital efficient than Bell Labs at its heyday.
The problem with VC is time horizons and portfolio size.
To take long steps through a fitness landscape of ideas, sometimes you need to traverse valleys or you need to take a step without knowing whether the step will yield gradient ascent. VCs aren't able to fund either of those things, because they have short horizons for the funds (so they need to generate 25% within 3-7 years) and because their funds aren't large enough to let true portfolio effects dominate (i.e. they're sampling a very limited distribution where they want every company to return threshold returns within that horizon, rather than being able to take bets on things where they simply don't know whether it will yield something).
That's why we still have government funding for so much fundamental research: commercial incentive structures and markets are quite bad at doing things that don't appear to yield 15+% IRRs within 5 years *as assessed at the time of investment*, and if that's your search strategy, you're sampling a very small set of the underlying distribution of possible critical inventions indeed.
If you can't compete with VC-funded startups, join them. Xerox decided it couldn't internally develop the work of PARC, so it encouraged the employees to spin off startups. At least, it said that was the plan, but it only managed to get one pair of people to do this, founding Adobe. Talking about efficiency is premature when it's just not happening at all.
To add to your list of Major Bell Labs Developments; they basically developed and standardized the way you interact with telephony and other information systems globally, which has enormous ramifications to this day on the omnipresent devices we keep in our pockets, as well as what we'd more generally term "Industrial Design" but then would have been a sort of psychological engineering.
https://www.nytimes.com/2013/02/09/business/john-e-karlin-who-led-the-way-to-all-digit-dialing-dies-at-94.html?searchResultPosition=1
In 1983 I hired into Bell Labs Research (Area 11); Materials Physics. I came from 2 years in a startup founded by 2 guys from Area 11 and was very familiar with Bell methods, processes, terminology. I came in as an STA - Senior Technical Associate. This was, generally, someone with a Bachelors in Physics, Materials, Chemistry and no experience required.
The fundamental Area 11 researcher was the Member of Technical Staff - Principal Investigator. They were a PhD in Physics, Materials, or Chemistry from Stanford, MIT, or Cornell. Current PIs would stay in touch with their old groups and ask who the top 2 people were from that year - thesis topic not important. Like the sports team that drafts the "best available athlete". When hired they'd be given someone like me - it was very much a Professor / Grad student relationship, albeit with good pay and benefits. An STA could take as many courses at night as we wanted; so long as you didn't get a PhD - then you would exit research. It was a tacit understanding none of us had an issue with; we didn't want to "dilute" the PI pool.
Area 11 was remarkably flat. You could walk into anyone's office, plop down, and talk shop. Arno Penzias, VP of Research, had a rule that everyone - from him to secretary - was on a first name basis. "It's not important where you were - only that you're here." Arno also remarked that if one of his people won a Nobel Prize in particle physics he would be embarrassed, as there was a general sense of where the research arrow should point.
Executive Director Kumar Patel (X-ray laser) said "If you're not failing 70% of the time, you're not asking the hard questions .............. but that other 30% better be world class." It started out that Research should be thinking 10 years down the road ..... then 5 .......... then 3 .......... then "what product do you support" ? That's when people began leaving. The stated 60/40 split of hard science/software was going to be reversed. PIs were told 'If you can do your work at a University you should go do it." By the mid-90's it was over.
The reason that labs like this are less common now is that there's an incentives problem: it's established in the economic literature that the fraction of the value of an invention that companies capture is low.
There's a compounding, perverse effect that when an inventor makes something that's particularly valuable (i.e. perceived to be particularly valuable by society at large), there's a negative feedback loop, as with GLP1 inhibitors and mRNA vaccines: the public's instinct is "that's extremely valuable to me, but also I don't want to pay what it's worth to me in order to have it, so the government needs to regulate prices".
They're the primary reasons why, as societies, we have so little fundamental research: because as societies we have a deep instinct to dislike seeing someone who's done something extremely valuable to society actually profit from it; as societies, the balance of IP protection and the ability to enforce rights vs. allowing others to develop a new technology isn't quite right (so firms can't capture enough value from innovation); and because, as societies, we have deep instincts for price regulations, which is an instance of concealed self-interest in wanting valuable things without actually having to pay for them.
The problems of inventions taking years to commercialise and the rate of success being low are more solvable: they can be done solved via "portfolio effects", whereby you look at 100 programs of fundamental research and maybe only 10 of them yield a commercial product, but you know that, on average, your "portfolio" still yields a 15% IRR or whatever. This is basically what VC firms and biotech firms do every day.
In sunmary, if you're a company and you think you can make commercial returns as an R&D lab like Bell Labs by doing fundamental research then commercialising or licensing it, you'd do that all day long. The fact that we don't see that anymore is an indicator that our political mechanisms for compensating invention aren't quite right, so that companies have defaulted to taking small steps - incremental things that can be commercialised quickly - through the option space.
We know that these labs can produce things of extreme value - everything on that Bell Labs list is a critical modern technology. We just aren't letting the inventors have enough of what they just developed.
Great point about incentives - One example is radio. In the 1920s and 1930s the development of radio caught the eye of congress and the US bought the rights for a complete set of patents for AM radio, and later kept going with FM radio and B&W/color TV. Congress (I think it was Congress) formed a "clearinghouse", RCA, where a manufacturer could go and buy a single license, allowing the production of a radio/TV with competitive technology.
This leveled the playing field for companies. Innovation continued but consumers had a variety of reasonably prices sets to choose from.
This model is capitalist enough to be interesting, but the RCA deal probably needed a better payout for the original innovators to justify a lab model. (Still making your point here.) We'd never get to it now without a crisis and a breakout technology that looked to have staying power. People/companies are so intent on their own ways of beating the other players, they never consider some place for an agency to manage the commons. We still have NIST, but they've been pushed back to the most fundamental things. Forget a set of more secure but also rugged protocols for communication networks...
I don’t think I’ve commented here before, but I’ve been reading your articles for a while, maybe 6 months, maybe a year. Really appreciate your excellent work and deep dives. Thanks!
Same here, I concur, excellent long-view articles.
It is excellent articles such as this one on Bell Labs that provide serious food for thought and build our continuing understanding.
Excellent article. Thank you
How great was Bell Labs? Two words: Claude Shannon.
This is a MASSIVELY important point!!!!!
Bell Labs also contributed hugely to human perception and memory research, which isn't on your list. I work in speech and hearing where ALL the fundamental knowledge comes from Bell. They also did the same for vision and other senses and memory. After they closed down, there's a lot less progress in these areas.
The influence spread to smaller companies as well. Before Wall Street bombed the country down to bedrock in the 80s, oil refiners, oil service firms, and wheat processors had major research labs in smaller cities like Enid and Ponca City and Bartlesville. All are gone now thanks to centralizing and financializing.
Wow! Great article!
I can confirm from personal experience that PARC became more focused on product development in the 1990s. I worked at the Xerox XSoft division, which was literally down the hill from PARC. Our job was to take research projects and try to turn them into products.
Even though the scientists and engineers at Bell Labs didn’t apply for grants, I imagine there was a healthy competition for ideas to fund (or at least to choose how much to allocate to each idea, which is probably what happened there?). Do we have access to any proposals for bell labs? For example, I read that Shockley’s transistor idea and George Mueller version of it were going on in parallel so there must have been some forum for discussing ideas before diving into it.. maybe there was some other avenue for it?
Funny. No mention of DEI. 😎
XXX
I’m personally familiar with PARC. It had a lot of diversity in some demographic areas: people from all over world, gay people, religious diversity. Not so much racial diversity.