Interesting. I consult a lot of early-to-mid-stage startups.
There's typically a sharp distinction made between "scaling" and "post-scale" startups. I've learned that this is because a pre-scaled startup is quantum physics; a post-scale startup--a business--is Newtonian physics. Most lenses for looking at businesses are fundamentally Newtonian. There's little awareness that they're Newtonian. Trying to apply them to quantum stuff is inefficient at best, disastrous at worst.
Government is not a business, first of all. Its mandate is to serve *every* customer; if I were handed a business like that, the first thing I'd do is fire most of the customers. Not an option for government. Second, the understanding of "risk" a government brings to the idea of "innovation" or "flexibility" or "spontaneity" or whatever "startup" quality they're trying to capture is based on Newtonian principles, the only ones they (imperfectly) have. It's what's in all the books.
In my experience, "like a startup" from a post-scale org usually means "a way to work out undesirable/risky/weird elements as you would a splinter," i.e. gently and indirectly. Not everyone is in on the scheme, alas. The track record of such efforts is incredibly poor.
Besides (1) lucky and (2) never taken existential risk, isn't there the option that (3) it has become resilient to a degree that it would take not one but many catastrophes to destroy it? I guess we can argue whether or not a threat you can survive counts as "existential", but I'm not sure that matters since my understanding is that you're asking about an organization's willingness to take big swings, not its suicidality
Is attempted risks gone bad really the main reason for organisational failure though?
It feels like even Stafford Beer himself would argue that most organisational failures aren’t down to voluntary but stupid decisions, they are down to external threats that emerge from the environment regardless of what organisations do, with survival instead predicated on the capacity of the organisation to take steps to adapt? Certainly this is the general assumption that seems to baked into most of the management cybernetics literature I’ve read.
Assuming that’s the case, it seems like that’s going to exert much more of an impact on organisational natural selection than the few instances of an organisation that does something so spectacularly stupid it annihilates them entirely. What’s the equivalent for this there?
It depends on whether you consider "existential risks" as a superset with members "existential risks taken on purpose" and "existential risks taken by accident, without knowing you are doing so"
Sounds like Stafford Beer's System 5 is the missing link here. So a long-running organization isn't necessarily hostile to all kinds of novelty, only the ones that aren't compatible with its identity function.
Thing is, that "threatening" novelty doesn't map cleanly to existential risk or non-existential risk. If the organization can't cope it, maybe it technically survives but with a different System 5. Like Doctor Who regenerating, metaphorically speaking
Football clubs feel like a useful edge case here. They are often wildly risk-seeking in the short term, especially around managers, wages, transfers and promotion gambles, but intensely risk-averse around institutional continuity. The owner can be replaced or sell up, the creditors can take a haircut, the manager can be sacked after six weeks matches, but the club’s name, colours, rituals and identity are treated as sacred. Look at MK Dons, look at the European Super League implosion.
So perhaps football clubs are Lindy institutions wrapped in fragile corporate shells: reckless with money, conservative with their meaning.
Is the defining feature of startup culture (as against government/mature enterprise) "risk-taking," or is it more about individual autonomy to act on decisions? More autonomy may equal more risk-taking, but it doesn't follow necessarily. I've interacted with startups that made oddly risk-averse decisions, but where the principals still act with a fair bit of autonomy.
Eg, in Jennifer Pahlka's book, it seems like the benefit of the "product manager" archetype is not so much that they are empowered to take risks as much as they can make decisions at all.
Autonomy isn't necessarily risk taking but it is novelty-producing and my intuition is that there is a strong tendency for risk aversion to mature and sour into hostility to anything new
I believe that the commenter is formulating the following argument:
"Never taken an existential risk" implies Number of existential risks taken = 0
Therefore, if Number of existential risks taken = 1 and still in existence, the statement is false. My own feeling is that this should be counted under (1) "lucky" but I can see the argument that (1) "lucky" only applies if both Number of existential risks taken > 1
and probability of failing to survive an existential risk is > 0.5
ie. 1 only applies if the number taken and probability of failure per risk are sufficient to make the aggregate success probability 'small'. 2 implies that the number taken must be exactly zero which is too strong
So, mmm. I'm not sure this is the reason. I mean, can there even BE an existential risk for a government department? As long as someone in the UK wants a car, and that car's got a number plate on the back and a road to drive on, there pretty much needs to be some sort of government organisation that handles that: a department for transport, if you will. Whatever you do while running it, you can't crater it so badly that it actually goes away.
There is the argument that sure you can; screw up badly enough and you and your staff will get reorged out of existence, but... it's not like the DoT has been an unchanging banner of solidity for a hundred years or anything. It's only existed since 2002; before that it was the department for the Environment, Transport, and the Regions, in the 70s it was a bit of the Department of the Environment with no separate existence, way back it was the Ministry of Transport (there's staying power for you, name-wise at least). If the reorgs that caused those were in response to taking an existential risk and failing it, then if anything the argument would support being OK with existential risk because they've survived a bunch of them. And if those weren't caused by shooting the moon and failing, then what more existential risk could there possibly be?
I certainly agree that government departments are disinclined to take risks and trying to do "startup culture" in them is probably doomed to failure, but I don't think it's because of institutional fear of destroying the org if you fail. I think it's either (charitably) "we shouldn't do this because it's not our money so we should be very cautious with it" or (uncharitably) "we shouldn't do this because if we screw up we'll be pilloried on the front pages of all the papers", neither of which are a problem to the org itself, although they might be to the person who runs it.
I think the answer to this irritatingly is something like "the system does not necessarily or even usually correspond to the org chart". The functions get performed but the decision-making units can be reformed, removed and restructured. So, by "the department of transport" I'm really speaking metaphorically and euphemistically; it's not the same long-lived organization in the way that the Catholic church is, but it's a congeries of decision-making systems, most of which have been around long enough to develop the risk averse tendencies, some of which are recent enough to still have some vitality, and some of which in this second group are currently in the process of taking the risks that will do for them.
I am not sure that is very clear; I hope that the book when it is published will have a clearer explanation of what I mean.
I think there's a level below this worth pulling out. If the unit isn't quite "the Department for Transport" but something more like "the decision-making systems that sit within it," then for programme-delivery purposes the relevant unit is the programme itself. And once you look at the level of programmes, something interesting happens.
Programmes are risk-averse upstream — a hugely ambitious project with an unconvincing case for success probably won't get funded. But once a programme is funded (often because, as Stuart points out, government has limited choice in what it delivers — there's no market to pick up the slack), departments will tolerate extraordinary levels of operational and financial risk to keep it running. The minister facing the cancellation decision sees the political risk of cancelling as larger, for them, than the risk of continued failure, for the public. So failing programmes can run for a decade or more, hoovering up money, missing every milestone, and never quite being killed.
I don't think this is really a story about risk appetite though. It's a story about risk legibility. A private-sector equivalent would be marking these risks to market quarterly; auditors would be asking pointed questions; the board would be demanding cancellation triggers. In government, the same risks accumulate on the public balance sheet but aren't booked as risks anywhere the system actually responds to. You can give your own example of UK government project failure that limps along indefinitely here.
So I'd push the conclusion further than your post does. What looks like government risk-aversion isn't really less risk-taking — it's a particular distribution of risk-taking, where small legible risks are avoided and large illegible ones are routinely run. The fix isn't cultural, it's mechanical: instruments that make the slow accumulating risks visible, and cancellation mechanisms credible enough that stopping a programme isn't a career-ending political event.
I view it through the balance sheet. A startup is an all-equity firm, which seeks upside and doesn't care much about insolvency. As it grows, it starts to rely more and more on debt, which means that its governance becomes more focused on the downside.
I'd look to the military as the classic example of an organization that shifts from conservativism to radicalism. The triggering event is painfully obvious: a serious war. I note the US experience with the second World War. Marshall inherited a peacetime army, fired almost all of its general officers, and replaced them with colonels and those more junior. He (and later Eisenhower) fired these replacements when they could not produce results within 3-6 months. Stalin and Lincoln did something similar in their existential wars, although only Marshall fired peacetime generals before they could fail on the battlefield.
On the other hand, I have personally been through a counter-example: the Great Financial Crisis of 2007-9, working at the New York Fed. Again a painfully obvious triggering event, after several decades of sloth. Nobody got fired, but quite a number of people were temporarily sidelined. The active part of the organization became very lean and hungry for awhile. When the crisis passed, the organization reverted to what it had been, with only a slightly different cast of characters.
The balance sheet is useful for businesses with a stable, repeating business model. Any startup worth backing--any innovative business, not another T-shirt shop or retail bank or restaurant--is searching for that model. It's not so much that the balance sheet is useless or wrong, per se. It's like betting on a racehorse based on its weight.
Clayton Christensen in The Innovator's Dilemma goes even futher - he demonstrated that in mature organizations it is rational to search for innovations only in high-profit segments. The caveat is that in these segments disruptive innovations rarely occur, if ever. Therefore, organizations become more profitable, but also more prone to be disrupted. If/when this happens they also often lack skills to survive.
In the public segment, services provided by organizations are often unique (and monopolistic). Therefore the growth of 'injelitis' as C.N. Parkinson called it, might be difficult to be observed until it's too late. They also might attract personalities prefering safe havens and no change, even if balanced by lower salaries.
What I find very dangerous that lately (esp. post covid) big bosses learned how to use public institutions and regulations for their own purposes. The risk there is carried by those who have the least chance to influence it - citizens or tax payers. At worst, few scape goats are sacrified.
While I'm not a fan of Polly Mackenzie, I think (with reference to the DoT) point 7 in her latest stack deserves to be referenced here: "Government is not the private sector (and it’s definitely not a start-up)" - part of the reason it's hard to get the DoT (in total, we can of course go to the Skunkworks response - which you do, and I have sympathy with) to behave like a startup is that it has a lot of responsibilities that require it to not be like a startup.
A reasonably classic approach (Hofstede) to organisational culture posits that attitudes to risk are largely a by-product of the dimension: Means orientation <-> Goal orientation. i.e. Is the org (or sub-unit) focussed on "how we do things" or "what we achieve".
As ever, the key question is "does the position on the dimension fit the necessary strategy." So, for example, canonically, we quite like nuclear power organisations to be means oriented, as means orientation is rather compatible with safety culture. (There's an interesting industrial evolution to be traced through the North Sea Oil industry, where the typical org has sub-units with very different cultural imperatives.) As you've convincingly argued at other times, a lot of what helps the private sector is a government that is reliable (possibly even boringly so.) This even maps over (in the extreme) towards stuff like bribery etc. Reliable is not always/only "means orientation" but it's a lot of it.
You can map out existential risk as "change in the environment so great that the 'how we do things' becomes invalid." As Michael Christian notes, failures in this regard look a lot like an atrophied System 5. As Doug Clow notes, in the UK government context, a lot of this is because "System 5" is significantly reliant on the political sphere, which has (eg in my view in the Department of Health) not really accepted that role.
All of which is to suggest, maybe the DoT (and others) have not in themselves actually *evolved* to become more risk averse in any significant way, or at least not since they became organisations of a certain size, inside a governing structure that (in theory) thinks about 30-50 year projections of need, capacity, etc. Rather than "cultural evolution creates coarsening and sharpening" (which I think is not a well evidenced take) the pressure for increased risk aversion has been not a ground up phenomenon, but an ideological one around "System 5." (You knew my hobby horse was coming - "post Hayekian learned helplessness" is the central dogma which denies the role of System 5.)
So, given all that, why do I say "I agree, but" - well, I think your model of how organisations become risk averse is looking in the wrong places - but your response is right in the here and now: if you have a new problem that needs a new solution (ie old means are invalid), you need a new group to create it.
How you then manage tensions between the new group and the old group is challenging? an exercise left for the reader ? (or at least another day.)
What would it mean for the Department of Transportation to fail at an 'existential risk'? A new government voted in? An amended constitution? A revolution?
The two existential risks for government departments are 1) screw something up so badly that it gets taken apart and its functions given to other departments and 2) get swept away in a DOGE style deregulation
But doesn't that happen all the time? And not even because of the actual product? A friend worked on the incredible system for Kentucky's implementation of Obamacare (KY Connect). It was by all rational standards really well done. A Republican governor was elected and scrapped it and withdrew from Medicaid expansion.
I don't think it explains everything, though. Two motivating, currently-salient examples: universities and the civil service. In both those cases, existential risks appear genuinely small.
Universities don't fail - yet! There have been some hasty mergers in the past. It is notable that Dundee university made a risky bet, and it did indeed go badly for the individuals responsible, but the university itself was kept alive. Similarly for London Met back in the day.
Individual departments of state get merged and abolished and so on but it's (almost?) never the result of what the civil servants there did or did not do, it's overwhelmingly a political imperative, ranging from a bold new vision to grubbing around to balance competing factions' portfolios.
I think there are other factors as well. For instance, if the organisation has been around for longer than anyone's career, simple scale creates a degree of skepticism that the latest new idea is going to be transformational, and that leads to inertia. And it is going to look like an attractive career bet to those of a low-risk frame of mind.
Interesting. I consult a lot of early-to-mid-stage startups.
There's typically a sharp distinction made between "scaling" and "post-scale" startups. I've learned that this is because a pre-scaled startup is quantum physics; a post-scale startup--a business--is Newtonian physics. Most lenses for looking at businesses are fundamentally Newtonian. There's little awareness that they're Newtonian. Trying to apply them to quantum stuff is inefficient at best, disastrous at worst.
Government is not a business, first of all. Its mandate is to serve *every* customer; if I were handed a business like that, the first thing I'd do is fire most of the customers. Not an option for government. Second, the understanding of "risk" a government brings to the idea of "innovation" or "flexibility" or "spontaneity" or whatever "startup" quality they're trying to capture is based on Newtonian principles, the only ones they (imperfectly) have. It's what's in all the books.
In my experience, "like a startup" from a post-scale org usually means "a way to work out undesirable/risky/weird elements as you would a splinter," i.e. gently and indirectly. Not everyone is in on the scheme, alas. The track record of such efforts is incredibly poor.
Besides (1) lucky and (2) never taken existential risk, isn't there the option that (3) it has become resilient to a degree that it would take not one but many catastrophes to destroy it? I guess we can argue whether or not a threat you can survive counts as "existential", but I'm not sure that matters since my understanding is that you're asking about an organization's willingness to take big swings, not its suicidality
Is attempted risks gone bad really the main reason for organisational failure though?
It feels like even Stafford Beer himself would argue that most organisational failures aren’t down to voluntary but stupid decisions, they are down to external threats that emerge from the environment regardless of what organisations do, with survival instead predicated on the capacity of the organisation to take steps to adapt? Certainly this is the general assumption that seems to baked into most of the management cybernetics literature I’ve read.
Assuming that’s the case, it seems like that’s going to exert much more of an impact on organisational natural selection than the few instances of an organisation that does something so spectacularly stupid it annihilates them entirely. What’s the equivalent for this there?
It depends on whether you consider "existential risks" as a superset with members "existential risks taken on purpose" and "existential risks taken by accident, without knowing you are doing so"
Sounds like Stafford Beer's System 5 is the missing link here. So a long-running organization isn't necessarily hostile to all kinds of novelty, only the ones that aren't compatible with its identity function.
Thing is, that "threatening" novelty doesn't map cleanly to existential risk or non-existential risk. If the organization can't cope it, maybe it technically survives but with a different System 5. Like Doctor Who regenerating, metaphorically speaking
This wouldn't be an "impossibility theorem", but an "improbability theorem" - lots of "not likelies" attest to the latter.
Fair point, well quibbled!
Football clubs feel like a useful edge case here. They are often wildly risk-seeking in the short term, especially around managers, wages, transfers and promotion gambles, but intensely risk-averse around institutional continuity. The owner can be replaced or sell up, the creditors can take a haircut, the manager can be sacked after six weeks matches, but the club’s name, colours, rituals and identity are treated as sacred. Look at MK Dons, look at the European Super League implosion.
So perhaps football clubs are Lindy institutions wrapped in fragile corporate shells: reckless with money, conservative with their meaning.
Is the defining feature of startup culture (as against government/mature enterprise) "risk-taking," or is it more about individual autonomy to act on decisions? More autonomy may equal more risk-taking, but it doesn't follow necessarily. I've interacted with startups that made oddly risk-averse decisions, but where the principals still act with a fair bit of autonomy.
Eg, in Jennifer Pahlka's book, it seems like the benefit of the "product manager" archetype is not so much that they are empowered to take risks as much as they can make decisions at all.
Autonomy isn't necessarily risk taking but it is novelty-producing and my intuition is that there is a strong tendency for risk aversion to mature and sour into hostility to anything new
Prop 2 isn’t right, D, as I suspect you know full well.
How so? In my brain it's pretty much true from the meaning of the words, what's the easy counter example I've missed?
I believe that the commenter is formulating the following argument:
"Never taken an existential risk" implies Number of existential risks taken = 0
Therefore, if Number of existential risks taken = 1 and still in existence, the statement is false. My own feeling is that this should be counted under (1) "lucky" but I can see the argument that (1) "lucky" only applies if both Number of existential risks taken > 1
and probability of failing to survive an existential risk is > 0.5
ie. 1 only applies if the number taken and probability of failure per risk are sufficient to make the aggregate success probability 'small'. 2 implies that the number taken must be exactly zero which is too strong
So, mmm. I'm not sure this is the reason. I mean, can there even BE an existential risk for a government department? As long as someone in the UK wants a car, and that car's got a number plate on the back and a road to drive on, there pretty much needs to be some sort of government organisation that handles that: a department for transport, if you will. Whatever you do while running it, you can't crater it so badly that it actually goes away.
There is the argument that sure you can; screw up badly enough and you and your staff will get reorged out of existence, but... it's not like the DoT has been an unchanging banner of solidity for a hundred years or anything. It's only existed since 2002; before that it was the department for the Environment, Transport, and the Regions, in the 70s it was a bit of the Department of the Environment with no separate existence, way back it was the Ministry of Transport (there's staying power for you, name-wise at least). If the reorgs that caused those were in response to taking an existential risk and failing it, then if anything the argument would support being OK with existential risk because they've survived a bunch of them. And if those weren't caused by shooting the moon and failing, then what more existential risk could there possibly be?
I certainly agree that government departments are disinclined to take risks and trying to do "startup culture" in them is probably doomed to failure, but I don't think it's because of institutional fear of destroying the org if you fail. I think it's either (charitably) "we shouldn't do this because it's not our money so we should be very cautious with it" or (uncharitably) "we shouldn't do this because if we screw up we'll be pilloried on the front pages of all the papers", neither of which are a problem to the org itself, although they might be to the person who runs it.
I think the answer to this irritatingly is something like "the system does not necessarily or even usually correspond to the org chart". The functions get performed but the decision-making units can be reformed, removed and restructured. So, by "the department of transport" I'm really speaking metaphorically and euphemistically; it's not the same long-lived organization in the way that the Catholic church is, but it's a congeries of decision-making systems, most of which have been around long enough to develop the risk averse tendencies, some of which are recent enough to still have some vitality, and some of which in this second group are currently in the process of taking the risks that will do for them.
I am not sure that is very clear; I hope that the book when it is published will have a clearer explanation of what I mean.
I think there's a level below this worth pulling out. If the unit isn't quite "the Department for Transport" but something more like "the decision-making systems that sit within it," then for programme-delivery purposes the relevant unit is the programme itself. And once you look at the level of programmes, something interesting happens.
Programmes are risk-averse upstream — a hugely ambitious project with an unconvincing case for success probably won't get funded. But once a programme is funded (often because, as Stuart points out, government has limited choice in what it delivers — there's no market to pick up the slack), departments will tolerate extraordinary levels of operational and financial risk to keep it running. The minister facing the cancellation decision sees the political risk of cancelling as larger, for them, than the risk of continued failure, for the public. So failing programmes can run for a decade or more, hoovering up money, missing every milestone, and never quite being killed.
I don't think this is really a story about risk appetite though. It's a story about risk legibility. A private-sector equivalent would be marking these risks to market quarterly; auditors would be asking pointed questions; the board would be demanding cancellation triggers. In government, the same risks accumulate on the public balance sheet but aren't booked as risks anywhere the system actually responds to. You can give your own example of UK government project failure that limps along indefinitely here.
So I'd push the conclusion further than your post does. What looks like government risk-aversion isn't really less risk-taking — it's a particular distribution of risk-taking, where small legible risks are avoided and large illegible ones are routinely run. The fix isn't cultural, it's mechanical: instruments that make the slow accumulating risks visible, and cancellation mechanisms credible enough that stopping a programme isn't a career-ending political event.
I view it through the balance sheet. A startup is an all-equity firm, which seeks upside and doesn't care much about insolvency. As it grows, it starts to rely more and more on debt, which means that its governance becomes more focused on the downside.
I'd look to the military as the classic example of an organization that shifts from conservativism to radicalism. The triggering event is painfully obvious: a serious war. I note the US experience with the second World War. Marshall inherited a peacetime army, fired almost all of its general officers, and replaced them with colonels and those more junior. He (and later Eisenhower) fired these replacements when they could not produce results within 3-6 months. Stalin and Lincoln did something similar in their existential wars, although only Marshall fired peacetime generals before they could fail on the battlefield.
On the other hand, I have personally been through a counter-example: the Great Financial Crisis of 2007-9, working at the New York Fed. Again a painfully obvious triggering event, after several decades of sloth. Nobody got fired, but quite a number of people were temporarily sidelined. The active part of the organization became very lean and hungry for awhile. When the crisis passed, the organization reverted to what it had been, with only a slightly different cast of characters.
The balance sheet is useful for businesses with a stable, repeating business model. Any startup worth backing--any innovative business, not another T-shirt shop or retail bank or restaurant--is searching for that model. It's not so much that the balance sheet is useless or wrong, per se. It's like betting on a racehorse based on its weight.
The republican and democratic parties came to mind.
But not, for some reason, the Labour and Conservative parties.
Clayton Christensen in The Innovator's Dilemma goes even futher - he demonstrated that in mature organizations it is rational to search for innovations only in high-profit segments. The caveat is that in these segments disruptive innovations rarely occur, if ever. Therefore, organizations become more profitable, but also more prone to be disrupted. If/when this happens they also often lack skills to survive.
In the public segment, services provided by organizations are often unique (and monopolistic). Therefore the growth of 'injelitis' as C.N. Parkinson called it, might be difficult to be observed until it's too late. They also might attract personalities prefering safe havens and no change, even if balanced by lower salaries.
What I find very dangerous that lately (esp. post covid) big bosses learned how to use public institutions and regulations for their own purposes. The risk there is carried by those who have the least chance to influence it - citizens or tax payers. At worst, few scape goats are sacrified.
"I agree, but"
While I'm not a fan of Polly Mackenzie, I think (with reference to the DoT) point 7 in her latest stack deserves to be referenced here: "Government is not the private sector (and it’s definitely not a start-up)" - part of the reason it's hard to get the DoT (in total, we can of course go to the Skunkworks response - which you do, and I have sympathy with) to behave like a startup is that it has a lot of responsibilities that require it to not be like a startup.
A reasonably classic approach (Hofstede) to organisational culture posits that attitudes to risk are largely a by-product of the dimension: Means orientation <-> Goal orientation. i.e. Is the org (or sub-unit) focussed on "how we do things" or "what we achieve".
As ever, the key question is "does the position on the dimension fit the necessary strategy." So, for example, canonically, we quite like nuclear power organisations to be means oriented, as means orientation is rather compatible with safety culture. (There's an interesting industrial evolution to be traced through the North Sea Oil industry, where the typical org has sub-units with very different cultural imperatives.) As you've convincingly argued at other times, a lot of what helps the private sector is a government that is reliable (possibly even boringly so.) This even maps over (in the extreme) towards stuff like bribery etc. Reliable is not always/only "means orientation" but it's a lot of it.
You can map out existential risk as "change in the environment so great that the 'how we do things' becomes invalid." As Michael Christian notes, failures in this regard look a lot like an atrophied System 5. As Doug Clow notes, in the UK government context, a lot of this is because "System 5" is significantly reliant on the political sphere, which has (eg in my view in the Department of Health) not really accepted that role.
All of which is to suggest, maybe the DoT (and others) have not in themselves actually *evolved* to become more risk averse in any significant way, or at least not since they became organisations of a certain size, inside a governing structure that (in theory) thinks about 30-50 year projections of need, capacity, etc. Rather than "cultural evolution creates coarsening and sharpening" (which I think is not a well evidenced take) the pressure for increased risk aversion has been not a ground up phenomenon, but an ideological one around "System 5." (You knew my hobby horse was coming - "post Hayekian learned helplessness" is the central dogma which denies the role of System 5.)
So, given all that, why do I say "I agree, but" - well, I think your model of how organisations become risk averse is looking in the wrong places - but your response is right in the here and now: if you have a new problem that needs a new solution (ie old means are invalid), you need a new group to create it.
How you then manage tensions between the new group and the old group is challenging? an exercise left for the reader ? (or at least another day.)
I’d point to the difference between a founder-investor and a salaried employee. See also: charities vs NGOs.
What would it mean for the Department of Transportation to fail at an 'existential risk'? A new government voted in? An amended constitution? A revolution?
The two existential risks for government departments are 1) screw something up so badly that it gets taken apart and its functions given to other departments and 2) get swept away in a DOGE style deregulation
But doesn't that happen all the time? And not even because of the actual product? A friend worked on the incredible system for Kentucky's implementation of Obamacare (KY Connect). It was by all rational standards really well done. A Republican governor was elected and scrapped it and withdrew from Medicaid expansion.
Yeah that seems very plausible.
I don't think it explains everything, though. Two motivating, currently-salient examples: universities and the civil service. In both those cases, existential risks appear genuinely small.
Universities don't fail - yet! There have been some hasty mergers in the past. It is notable that Dundee university made a risky bet, and it did indeed go badly for the individuals responsible, but the university itself was kept alive. Similarly for London Met back in the day.
Individual departments of state get merged and abolished and so on but it's (almost?) never the result of what the civil servants there did or did not do, it's overwhelmingly a political imperative, ranging from a bold new vision to grubbing around to balance competing factions' portfolios.
I think there are other factors as well. For instance, if the organisation has been around for longer than anyone's career, simple scale creates a degree of skepticism that the latest new idea is going to be transformational, and that leads to inertia. And it is going to look like an attractive career bet to those of a low-risk frame of mind.