R&D, Transport, Skills, and Growth.

The last few months have been hectic. You’ll see our new product announcements at The Data City next week which explain why. In a time where I haven’t had much chance to reply, I’ve read two fantastic blog posts on regional development in the UK.

Both pieces are important. Henry and Paul along with people like Diane Coyle and Cécile Maisonneuve are always worth paying attention to when they have different opinions on something in regional economic development. Those differences are usually rare and small, or at least usefully understandable if deep.

The blog posts

In both cases I agree with much of what the authors say. We just don’t know. We don’t know if far higher transport investment in London and South-East England is justified by higher returns on that investment. We don’t know if far higher investment in London and South-East England on R&D is justified by higher returns on that investment.

But I feel like we disagree on what action that implies we should take.

To me, not having a clear justification for far higher spending in one place, after thirty years of spending more there, is an argument to stop spending more.

I happen to think that there are good justifications for spending more on both transport and R&D in London and South-East England, and that we should continue to do so. But I have not yet found sufficient justification for the scale of the disparity in spending, and so I think it should decrease. In both cases, since the UK spends comparatively little in these areas and has an ambition to spend more in the future, we can reduce current imbalances in spending without cutting absolute spending anywhere. This is a rare opportunity and so there is some urgency to making it happen.

Transport investment and local economic growth.

I can’t comprehensively respond to the first blog post, on transport investment, yet. I need to understand the regressions used in the paper that informs the blog post. That will mean reproducing them from the raw data, which I haven’t had time to do yet. So I can only make these basic points,

  1. I agree that the UK government has not systematically collected data on the transport investments it chooses to fund and not fund that would allow thorough analysis of any possible preferences in funding allocation. Most work continues to rely on the Eddington Report of 2006 where Tim Leunig and colleagues made what is still one of the best efforts to assemble this data.
  2. Given the poor data, I agree that it is not possible to prove that transport investment funding is biased in favour of London and South-East England. I have my opinion about Leeds, but with such a small number of considered schemes we can’t prove anything. But the flip side of the lack of evidence is that it is not possible to disprove bias either, and given clear bias in political assignment of spending the default position to hold is not clear. If accusations of bias are unproven then so are frequent political defences of current spending patterns as “picking the best schemes wherever they are” especially since it is those making the claims that could release the data to back up their claims, but choose not to.
  3. Within the limited data that we do have to analyse this issue, making different assumptions (whether to include wider economic benefits or not, and whether to weight schemes by their total value in regressions) can “give the answer we want”. This means that all analysis risks confirming our biases rather than challenging them.

If we can’t prove that investment in transport infrastructure is providing better value for money in London than in North England I cannot see how we justify spending about three times as much on it. In my view the burden of proof after at least thirty of imbalance is not on those who propose more equitable spending, it is on those that propose the continuation of current imbalances.

R&D investment and local economic growth.

The second blog post is one that I can respond to more fully. It argues that we should be more cautious than the author imagines we will be in reducing regional inequality in government R&D spending because,

  1. Too little is known about why there are differences in innovation across the country.
  2. The mechanisms to encourage innovation aren’t well understood.
  3. We risk focusing on R&D and taking our eyes off skills and agglomeration, two areas that are more important to stimulating economic growth.
  4. A place creating innovation doesn’t automatically capture its value so we shouldn’t worry too much about dispersing innovation funding and thus innovation.

Points 1 and 2 are “we don’t know” arguments. So my simple response is as before. If we don’t know enough about what encourages innovation or why there are differences in innovation across the country, then we cannot justify the current and long-standing inequality in spending on innovation by the state.

But some people want complex and nuanced responses with a rich narrative, so here’s one of them.

I disagree that we don’t know much about innovation. About one third of total R&D spending in the UK is spent by the public sector and almost all of that is assigned to universities in proportion to their research strength. An extra weighting is applied to universities in London, with no economic justification. The other two thirds of total R&D spending in the UK is spent by the private sector. We don’t know precisely why they spend money where they do, but in a market economy we don’t have to. We can reasonably assume that they spend their money where they think it will generate the greatest financial return on their investment. One of the most powerful tools for understanding market economies is that we don’t have to understand every part of them, we can merely observe the collective wisdom of interested parties working in their own interests. Like private businesses investing in R&D and failing where they make poor choices.

We have two traditional and well-established measures of innovation output; patents and scientific publications. In the past decade work by Microsoft has made these easy to analyse. And by assuming that the private sector optimises economic output from its inputs in R&D as I just described, we can also use private sector expenditure on R&D as a proxy for innovation output.

These are, like all economic measures, flawed, but they are also surprisingly useful and the least worst option we have. I think that they are better than Paul’s proposal to use productivity as an innovation output measure. In the long-run I would expect this to be true, but the long-run is a long time in R&D. Berlin has been an innovative city for decades but though its productivity is growing quickly it remains a comparatively low productivity city in Germany. Seoul and Tokyo had low productivity compared to European and US competitors as they boomed. Many of China’s most innovative cities are still relatively unproductive, but few would gamble against them becoming productive in coming decades. In all these cases we see the outputs of innovation as measured by patents, scientific publications, and business spending on R&D long before we see it in productivity data.

We should of course be looking at better ways of measuring innovation, especially in service industries and in software, where patents are rarely filed. I tell the UK government every few months what they could do with R&D tax credits data to increase our understanding. One day we’ll get there.

We may also want to look at economic complexity as a way to understand non-patented innovation, but I don’t understand that. Importantly, the absence of these measures doesn’t stop us from understanding quite well what the differences in innovation are.

I also think that we know quite a lot about what encourages innovation in a place. If you want rigorous economic studies then Enrico Moretti & Claudia Steinwender & John Van Reenen have that. They measure the crowding-in effect of public sector spending on R&D. Private companies join in with public spending. Or you can see the same in the long run data on R&D spending in the private public sector across countries and regions where public spending in a place is strongly correlated with private spending (at two to three times the rate), and where there is increasing evidence that public sector spending can lead private sector spending, most famously in East Germany, but in lots of other places too.

But what I think is most convincing is to look at examples and talk to innovators. This lets me respond to Paul’s third and fourth points.

Let’s start with the Oxford-AstraZeneca vaccine that Paul mentions. The story lets me explain most of what I think about R&D and its role in regional economic growth.

AstraZeneca is a company that Richard Jones and I know well as scientists. My current career is the result of being trained, effectively, to join the company. It moved before I could and I did not want to move with it. Richard and I have also looked at AstraZeneca in detail in much of our work on regional development. In our The Missing £4bn report we looked at the incentives for its move from Cheshire, the UK’s most intensive private sector R&D cluster at the time, to East Anglia (Cambridge), the UK’s most intensive public sector R&D cluster at the time and a much less intensive private sector cluster.

AstraZeneca and its private sector R&D spending followed the state money. State investment in innovation incentivised private sector spending in innovation in this individual case, just as we see in aggregate data and properly controlled economic studies.

But what did this decision look on the individual firm level?

Fortunately Richard and I know people at AstraZeneca and it’s interesting to talk to them. In my experience, none explicitly mention public R&D spending in Cambridge as the top motivation for their location choice. The reasons given are “access to talent”, “an ecosystem of similar innovative companies”, and “access to London [for regulatory, financial, and scientific reasons]”. “Access to great science at the university” is mentioned, and initially rejected as being about relying on public sector funding, but eventually conceded as that.

It is tempting when we listen to these stories to say “see, it’s skills and agglomeration not really R&D spending”. We do that a lot in the UK and I think that it is a mistake.

The reason that the talent that AstraZeneca wanted to hire and work with in Cambridge existed in Cambridge is in part because of the accumulated impact of the University and the private sector ecosystem that developed around it over many decades. It is high UK government and UK charity investment in the city that has funded the work in the labs, that led to more people relocating to the city and staying in the city and being trained in the city. Many of these people then go on to found companies, create a cluster of excellence, and begin the virtuous cycle that AstraZeneca rightly wanted to join and that many companies had joined earlier.

R&D spending is supplementary to many more reasons in a complex system, but it remains a large, separable, and observable part of Cambridge’s success.

To briefly return to thinking about the first paper, the boost to Cambridge’s attractiveness from UK government R&D was clearly topped up by Cambridge’s excellent public transport links to London, links that were recently improved enormously by the Thameslink project, a very large UK government investment with a very low modelled return on investment. Thameslink’s modelled return on investment was far lower, for example, than a tram for Leeds which was cancelled and the Northern Hub Rail project which was downscoped in North England, both of which would have increased the attractivity of the North to AstraZeneca.

With specific reference to AstraZeneca, it’s useful to imagine an alternative future where,

In this alternative future, AstraZeneca would have been much more likely to stay in Cheshire, UK regional economic inequality would be lower as a result, and house prices in Cambridge and London would be slightly lower. We would still have our brilliant Covid-19 vaccine, but the scientists in Oxford would have looked North, not East. And I bet that if we asked AstraZeneca in this alternative future why they stayed in Cheshire and expanded into Manchester they would say it was because of the better access to talent, the better access to an ecosystem of similar companies, the better access to world-class research, and the better access to the state’s regulatory framework.

R&D and transport spending wouldn’t come top of their list, but that would be a big part of why.

To be blunt, access to talent is by far the top feature of a place that attracts investment, including investment in innovation, but it is not proven that investing directly in skills (which we spend vastly more on than R&D and transport) is the best marginal way to create, attract, and retain that talent and thus increase economic growth. So I believe that greater R&D and transport investment, especially in a country which under-invests in both, are in many parts of the UK the better marginal investment today.

That’s 3 dealt with, so now onto point 4.

Point 4 has a simple response. If innovation disperses quickly and a place creating innovation doesn’t automatically capture its value then that is an argument to do innovation in places where it is cheaper, specifically away from London, Oxford, and Cambridge. India has excellent scientists willing to work for low wages, yet there is little push for the UK government to fund science in Chennai instead of bringing people from Chennai to Oxford to do that work. Place does seem to matter to us quite a lot nationally, which suggests to me that it should matter internally too.

Once again, there is a more complex and nuanced argument, with an accompanying narrative.

Paul is right to say that the discovery of the successful Oxford-AstraZeneca in Oxford and Cambridge will generate a huge amount of economic value, but with only a vanishingly small fraction retained within Oxford and Cambridge. The same will be true in Mainz, home to BioNTech, the creator of the first proven Covid-19 vaccine and in Boston, home to Moderna.

But focusing on the big successes of innovation is the wrong way to look at this. With all successful vaccines we are observing a single flower at the end of a long branch of innovation and investment, on a tree where most branches will never produce a bloom. The brutal truth is that the vast majority of time spent working on innovation achieves nothing. This applies in all areas of innovation, but most brutally in basic research. Most scientists spend most of their careers working on dead ends. Tens of thousands of brilliant economists wasted their whole careers trying to make communism work better in the USSR. Millions of IT consultants are today installing business software and associated processes that will never pay back on their investment. For every Skype there are hundreds of companies you’ve never heard of that went bankrupt incurring huge losses for their investors and backers. RIP Jaiku.

But all of these companies, while trying to innovate, paid their staff good salaries for years. And those wages let people live good lives and power local economies. When we look at innovation in this way we see that, on the private side, the shiny successes generating huge profits for lucky investors are recycled by those investors into hundreds of failures. On the public side it is the public goods of innovation, such as not living in a never-ending pandemic, that are redistributed by the taxes raised in a strong economy enabled by those public goods into hundreds of failures, and the odd success.

Before any MMT fans get in touch, it is thanks to the innovations in our banking system that the spending on innovation paid for by its rare successes can actually occur because the success is achieved.

Since the economic value of innovation is pre-distributed from its few successes to its many failures it is reasonable to measure its inputs (wages), which are highly localised, as economic success not just its innovation outputs, which quickly disperse. We should also consider some portion of the skills created, attracted, and retained in an area as outputs from public investment in R&D.

The skills point is so important, and so regularly missed when looking at R&D spending. Innovation spending lets people stay in places and develop their talents, ready to be hired by a company looking for a place to grow. And sometimes that company has the better idea and the better luck that lets a lifelong innovation failure achieve something amazing. A failed innovation in a new context is often a key to success and the failures live on mostly in the minds of people.

One key to the economic success of California’s innovation ecosystem, something increasingly appreciated elsewhere but never quite as fully, is the value of people who have tried and failed, or who are trying and failing elsewhere. A town full of people who have tried to innovate and not quite made it and are now working on the next thing is much more attractive to a company picking somewhere than a town full of people who never tried.

A few examples

I find that examples help me to explain these theories best, and thankfully we have a lot once we start looking. Here are just two in Manchester.

The first is ARM, a great UK success story. It was founded, to a decent approximation, in Cambridge by a PhD graduate of that University, with help from the UK government and the UK’s state broadcaster. It relied heavily on talent trained and retained in the city by the university’s world-leading computer science department, which was well-funded by the UK government.

More quietly than AstraZeneca moving the other way, ARM has been gradually expanding to Manchester. The draw is the city’s talent pool in computer science, deeply linked to the University. In 2018, the number of ARM employees who had studied at Manchester University went above the number who had studied at Cambridge. ARM’s offices in Manchester continue to grow, with employees dipping in and out of research projects at the university, ARM, and other businesses like IBM and a broad collection of cybersecurity companies you should hope you never hear about. Manchester has a measurably strengthening cluster of talent, businesses, and research in this area.

As Paul says in his piece, innovation disperses quickly. Here’s an example.

When Apple relied on Manchester University spinout Transitive technology to power their transition from PowerPC to Intel chips in 2006 the innovation jumped from Manchester to California almost instantly. And certainly the profits that the transition enabled were mostly booked in Palo Alto.

But the pool of talent was far less mobile. It remains in Manchester, with the university’s computer science department and its other spinouts providing a reservoir that people with talent can retreat to if innovative companies fail and that new innovative companies can draw from if they need. ARM and dozens more companies draw on this pool of talent, partly supported by the state.

One such company is bet365 who have been expanding in the city from nearby Stoke. They explicitly hire from the universities, support projects at those universities that create more talent, and form part of the ecosystem that keeps skilled people who command high salaries and thus pay a lot of tax. They hire lots of people who’ve previously worked at the universities, at ARM, at IBM, and elsewhere, because experience of trying to innovate is a good qualification to innovate even in fields which may not seem very related. In fact it turns out that calculating live betting odds on thousands of markets at once and delivering them instantly to millions of mobile phones is exactly the same core problem as ARM engineers struggle with and which Manchester University’s researchers were working on before either realised they were hard problems.

I think I’ve shown that R&D spending plays a big role in training, attracting, and retaining the talented people that attract innovative companies and increase productivity in an economy. The largest part of this spending, except in a few unusual places like London, Berlin, and Dresden where the state plays the major role, is by the private sector. But by far the easiest part for governments to influence is the public sector spend.

By spending ten times more on R&D in Cambridge than Cheshire, the UK government incentivised AstraZeneca’s relocation to Cambridge, largely via the mechanism of providing a reservoir of skills directly in the public/university sector and seeding a reservoir of skills in the private sector. By spending less on R&D in Manchester than in London today the UK government is incentivising the same. I want to reduce that.

Achieving some change

Me and Richard Jones have been looking at data, talking with people, and working in innovation for a while. In 2020 we published a report with Nesta looking at the issue of R&D spending by the UK government. I am really happy with the impact of the report for two reasons. First, because it is clearly changing UK government thinking. Second, because the questions we got in response have either been answerable leading to improvements to our arguments, or they have been unthinking and easily dismissed. So far we have not endured the “oh shit” moment where we made such a deep blunder in our work that it no longer makes sense.

Paul’s points on whether dispersing the state’s R&D investments is a good way to try and disperse economic strength within the UK are good. I hope that this blog post has responded to them in a convincing way. I now want to respond to three of the less useful challenges that I hear from worryingly high levels within government and the associated policy world. I think I deserve it after writing 3000 words. I’ll restrain myself to three, and my swearing to mild.

  1. Lots of people have told me to shut up about the London weighting component of QR, the method by which the UK assigns research funding. It’s a small amount of money, they argue, and I should look at the big picture and be much more complex, nuanced, and holistic in my thinking. This is as coherent an argument as not stopping a child from falling into a pond because what’s really needed is a fence around the pond. That arguments of this type are respected in UK policy discussions is frustrating. We should remove London weighting from QR. It would make a very small improvement, I have never argued otherwise, and it doesn’t stop us from doing anything bigger. That we claim to have tried to fix regional inequalities for decades while leaving in this economically illiterate distortion undermines claims that we have tried in the past.
  2. A similar argument which regularly leads to my smartwatch giving a warning for Atrial Fibrillation is that “R&D is just a small part of the puzzle”. No shit. No-one has ever claimed otherwise. Disassemble your straw man! This is similar to arguing that France shouldn’t bother playing Kylian Mbappé any more since football is a team game and it’s the whole squad combined with their training that counts. We can make some decisions in isolation and work on the rest later. My best guess is that poor UK government R&D spending patterns account for at most a third of the economic underperformance of our mid-sized cities (Manchester, Birmingham, etc…). My estimate is much lower if R&D’s effects in increasing skill levels as I’ve described are much lower than I think. Similarly, Mbappé playing for France probably accounts for just 5% of the extra probability of them winning a match. But just as Didier Deschamps puts Mbappé’s name on the team sheet without neglecting all the other parts of winning the world cup, the UK government should be working to improve how we assign R&D spending better without neglecting all the other important things it needs to do to lead a thriving country.
  3. The “not ambitious enough”, “too ambitious” death spiral. This is classic UK thinking. If I suggest increasing R&D spending in Birmingham, Manchester, and Liverpool by 30% lots of policy people will say “why not aim higher?”. If I then suggest anything more, they will say it’s too ambitious, unproven, and we don’t know enough to proceed. It’s pure mindrot at the core of our national policy debate and we need to snap out of it.

Richard and my proposals in The Missing £4bn are remarkably tame. I want the UK government to assign its R&D spending slightly differently to how it does it now. Specifically, I want it to focus on,

  1. Supporting R&D spending where market signals (private sector R&D spending) suggest that the state is spending too little.
  2. Concentrating higher R&D spending on agglomerations where the state can play a role in creating a large pool of skilled people that will attract innovative companies. (we can also increase the size of these pools by investing in transport).
  3. Making changes to the structures of governing R&D spending that will make it less likely that the provable geographical biases of the past in favour of R&D spending in certain parts of the UK (London) are less likely to be repeated.

It’s not a big change, but it’s one that I think we can be pretty confident will slightly increase UK growth while making that growth a bit more regionally distributed. A small change, with a small positive outcome, but one that in decades of claiming that we were dreaming bigger we somehow never tried.

And the nice thing about all these things is that Paul, Me, and Henry mostly agree. Even if I’m a little bit more desperate to get started.

blog comments powered by Disqus