The federal government is taking its sweet time being sworn in (apparently the GG is on holiday in Germany or something), so it’ll be another week or so before we get new ministers and new mandate letters. These letters set ministers’ priorities in a more formal way than manifesto commitments.
My absolute dearest wish, when it comes to Science and Innovation, is that these letters should read “you should take our mandate commitments seriously but not literally”. That is, the government should be committed in general to interesting ideas to improve science, productivity, innovation, etc, but on absolutely no account should it tie itself to the deeply frivolous and unserious notion of creating a Canadian DARPA. It’s not just that it’s that there’s no evidence that a DARPA-like entity answers any of Canada’s evident problems in innovation: it’s that it is vanishingly unlikely that any Canadian government could establish a DARPA-like entity in a way that would give it any chance for success. Even if it might work, it won’t.
To understand what I mean here it’s worth talking about what people mean when they say DARPA “works”. There are two separate definitions of DARPA’s success. The narrow definition, which has technically been the only raison d’etre since the early 1970s, is: “does DARPA produce advanced technologies which are of benefit to the US military?” This is an important question, because the US military drops a couple of billions of dollars a year on this, and it’s good to know whether the military and its civilian overseers think they are getting their money’s worth. The answer, if a pile of Congressional Budget Office reports are to be believed, is yes. Which is great for the Department of Defense, but not necessarily a whole lot of benefit to anyone else.
But there is a second and much broader way DARPA is presumed to “work”, which is more eagerly embraced by the policy entrepreneurs selling this particular vision (that is, the folks who want to believe in some phantasm of an “entrepreneurial state”) than by people who know how DARPA works. And that is that somehow DARPA -like agencies can produce massive economic spillovers in the form of general-purpose technologies that can be commercialized once they have been proven through the DARPA process. The common success stories you hear are the Internet and GPS but these, it is worth noting, are several decades old now and date to the era when DARPA was ARPA and less-focussed specifically on military applications.
As I have noted before, I think success on that second score is extremely unlikely and that anyone who thinks that’s what some kind of CARPA is going to achieve is dreaming in technicolour. But let’s take the narrower definition of success, for a moment and see if we can imagine a Canadian DARPA having any success at all.
So first, the narrow DARPA implies above all that the government is an extremely good client. That is, it is putting out contracts to private industry to accomplish certain proof-of-concept tasks with a view to eventually putting technologies into use. DARPA isn’t about curiosity-based research, and nor is it just about funding crazy-off-the-wall ideas: it’s about ideas which may seem futuristic but still have very imaginable applied uses by the military. Do take a list at DARPA’s current portfolio of projects: it includes things like “Thermal Imaging using Material Physics”, “Atmosphere as a Sensor”, “Gamma Ray Inspection Technology”, etc. It’s a bit out there, but you can see the military applications of most of them. The military wants things.
Now I want you to think, just for a second, if there is any department in the Government of Canada that is capable of being a client the way the US DoD is. I’m not even asking if they have the savvy to manage crazy-off-the-wall projects, I’m asking if there is any department which has purchasing needs which absolutely require there to be private-sector agents going out and doing crazy sci-fi things as opposed to buying boring, proven off-the-shelf technology? I can think of a couple that probably should – Fisheries and Oceans being probably the top of the list – but in real life?
But leave that stuff aside: let’s get down to the brass tacks of how DARPA worked. Most analysis of DARPA’s successes which has ever been done (there are loads, but Ben Reinhardt’s Why Does DARPA Work is the best short version) gives an awful lot of the credit to i) opacity and ii) personnel management. The first is simple: DARPA, as part of the military, acts in secret. No freedom of information, precious little central oversight. This allows people to move fast, break things, and be certain they will not be made a scapegoat if something goes wrong.
Then there’s the personnel side. DARPA is small, with not much more than 100 program managers who decide on projects and distribute a couple of billion dollars every year. And these individuals – get this – rotate out every five years. They are not bureaucrats – they come in from industry or from universities, do a job and get out before they can be captured either by bureaucratic thinking or by the companies they fund. They work in a comparatively flat structure which allows program managers to make funding decisions with minimal oversight. And in general, they work in an environment which tries to put as few rules in place as possible, particularly when it comes to hiring.
Simply put, there is no way in hell an Ottawa-created DARPA will look anything like this. Any new agency will certainly be bound by the Official Languages Act and the Freedom of Information Act (ineffective as it may be). It will likely be bound by the Public Service Employment Act. It will almost certainly be answerable to Treasury Board on a regular basis. And I absolutely guarantee you that neither the Public Service Commission and nor the various public service unions are going to go along with a 5-year sunset clause on terms. No way, no day. Hell, half the reason official Ottawa likes this cockamamie idea is the sheer feeding frenzy that goes on whenever someone creates a new agency (New hires! New empires!).
So, let me recap this. The provable track record of the US DARPA relies on i) a department which has concrete purchasing needs and ii) can act as a good and reliable client, while at the same time having an agency which iii) is opaque, iv) rotates personnel frequently and v) operates free from many of the rules which constrain other government departments. Oh, and vi) the client needs to be patient with the agency: it must tolerate extraordinarily high failure rates, because it is only with a heck of a lot of failures that the occasional crazy futuristic technology ever happens.
That’s six key success criteria. And literally none of them are going to happen to a Canadian DARPA. So, let’s put this idea back in the closet where it belongs, turn our backs on gimmickry and get back to serious thinking about innovation and growth.
Tweet this post
The first thing to point out is that Canada already has organizations that function in a role similar to DARPA. One is the National Research Council. Another is the much less well known Defence Research and Development Canada (DRDC). How these work sheds some light on the argument made above.
The two-fold definition of ‘works’ is at play in both agencies, with some minor caveats. The NRC delivers research to ‘clients’, some of which include the military. DRDC has only one client, which is the military. But there is the expectation and hope that there will be wider benefits from both types of research, either directly, through the commercial or military success of the client, or indirectly through the production of new knowledge or new technology that is of wider benefit to Canadians.
There is a significant difference in how DARPA works, however. “DARPA implies above all that the government is an extremely good client. That is, it is putting out contracts to private industry to accomplish certain proof-of-concept tasks with a view to eventually putting technologies into use. ” This is actually pretty similar to the way DRDC works, but much less so about how NRC works.
There was an effort during the Harper years to change that, and that efforts hasn’t entirely disappeared. Both NRC and DARPA were organized into ‘programs’ (NRC first, then DRDC) led by program managers that would deliver technologies to clients by contracting either external or internal resources. They are still organized this way. At NRC, programs are (mostly) constrained to using internal resources. At DRDC there is more flexibility.
I have been an NRC program leader and have worked in both NRC and DRDC programs. The intent was to have programs contract to external agencies, including universities, private industry or other government departments, and I actually did this to a small extent on my program. However, it would take a major change to the way the budget is managed for this to be done on a widespread basis. DRDC, as I said, has more flexibility, but I’m not sure how much.
Would this work better? Probably not. Contracting to private enterprise, especially in the area of military applications, greatly increases the cost of whatever is produced. And you’re not buying greater quality or even greater control over the product, you’re buying speed. But this – as correctly pointed out – comes at a cost of transparency and control. My own view was that programs structured this way would be a better way to subsidize selected companies and industries (which is primarily how it functions in the U.S.) that do develop new and important research.
Canada’s NRC already has a way to subsidize research: through IRAP, which delivers millions of dollars to companies taking research risks. Applications are selected through IRAP programs (I have been involved in this vetting process) sometimes to help companies develop their own product and sometimes to help a product or service for other government departments. Viewed from the perspective of supporting commercial research and development, this is just one of numerous initiatives that spend billions of dollars annually, including the Strategic Innovation Fund, Business Development Band, Canadian Institutes of Health Research, and many more (there’s a good list here: https://www.venninnovation.com/en/blog/federal-budget-2021 ).
The other way to look at research programs is to look at them as a form of research management. That’s the thinking behind sentiments like this: “DARPA isn’t about curiosity-based research, and nor is it just about funding crazy-off-the-wall ideas: it’s about ideas which may seem futuristic but still have very imaginable applied uses by the military.” The concern, which was louder during the Harper years than today, but which never goes away entirely, is that government research should be in some way directed, not merely curiosity-based.
Canada accomplishes this through funding, both internally through programs (which still exist) and other projects, and externally through the funding agencies, including especially (but not in any way limited to) the tri-councils – the Canadian Institutes of Health Research (CIHR), the Natural Sciences and Engineering Research Council (NSERC), and the Social Sciences and Humanities Research Council (SSHRC). These agencies provide direction, but for the most part there’s no particular client (other than the government itself).
We could debate whether these programs are, taken collectively, ‘successful’, but on review there’s no particular evidence that shows that a program-based defense-oriented set of programs contracting to private industry for specific outcomes would be any more effective. Arguably, diverting resources from existing systems for subsidizing commercial R&D, or diverting funds from existing research finding programs, would represent a step backward, getting less value for more money. And it’s hard to see the case for allocating more, given that defense already has a sizeable research component.
The additional commentary on why a Canadian DARPA would fail is incorrect, and demonstrates a lack of knowledge of how programs are in fact managed (this isn’t a criticism, it’s just a fact).
Programs are director-level positions. These are not union positions; they are management. Five-year appointments are very common at the director level, and in the research agencies especially, directors can come from anywhere – from within the ranks, or from external agencies. I’ve worked with astronomers, engineers, former industry execs, former university administration, and more. The union (or, more accurately, the professional institute ( https://pipsc.ca/ ) never said a word, because they have no jurisdiction.
Why are these director-level positions? This has nothing do do with hiring or firing (other than the directors themselves, but this is pretty minor in the overall scheme of things). Programs directors do virtually no hiring or firing at all – why would they? They are contracting out research tasks, either externally or internally. No, it has everything to do with budgets – directors are responsible for portfolios of millions of dollars, and this makes them a management level position. When I led a program, we worked “in a comparatively flat structure which allows program managers to make funding decisions with minimal oversight.” This has changed quite a bit since then, because as it turns out, this isn’t a very good way to manage prortfolios of millions of dollars (now we’ve gone too far the other way, but that’s another story).
So let’s review. “The provable track record of the US DARPA relies on i) a department which has concrete purchasing needs and ii) can act as a good and reliable client, while at the same time having an agency which iii) is opaque, iv) rotates personnel frequently and v) operates free from many of the rules which constrain other government departments. Oh, and vi) the client needs to be patient with the agency: it must tolerate extraordinarily high failure rates, because it is only with a heck of a lot of failures that the occasional crazy futuristic technology ever happens.”
i. The government as a whole spends billions of dollars annually on concrete research and development purchasing needs, contracting either internally through other government departments and programs, or externally through programs that fund private enterprise, universities, and other agencies.
ii. These are good and reliable clients, they have agency, and they never go away.
iii. In some cases they are opaque (especially in contrast with the U.S., which treats government IP as public property that can be used by anyone) and bound by private contracts (you’ll have to trust me on this one), though in general opacity results in less efficient management, sacrificing efficiency and control for speed
iv. Director-level positions, which manage the large budgets needed for such programs, rotate frequently, and are only sometimes filled internally, being filled just as, or more, frequently from business and industry, usually on fixed-term positions or secondments
v. Granted, these agencies are bound by the rules which constrain other government departments, which have been demonstrated through time to produce more efficiency and agency in public spending.
vi. Trust me, there’s room for failure. 🙂
So – while I agree that there isn’t really a good case for a Canadian DARPA, it’s not because Canada couldn’t make it work. To a large degree, we have funding and programs in place that *already* implement most of what a Canadian DARPA would. Rules involving unions, bilingualism, treasury branch, and freedom of information simply aren’t barriers to success. But we *shouldn’t* more toward a system that is more opaque, with less oversight, that simply (and secretly) funnels government money to private industry.
Yes, there are voices saying we should. They were louder during the Harper years but they still exist. They should be resisted,