In The Eternal Inferno, Fiends Torment Ronald Coase With The Fate Of His Ideas

What have US healthcare, British railways, the shipwreck of Carillion plc, and the F-35 got in common, and why should you care? Well…none of them work terribly well, they all cost vastly more than expected, and nobody can put their finger on why. Cash seems to leak out of them by a thousand cuts, without necessarily ending up with a well-defined villain. At the same time, productivity is stubbornly terrible and improvement forever delayed.

These diverse phenomena actually have a lot in common. I think of them as Coasian hells. Ronald Coase observed that an organisation could be considered as a collection of contracts, and asked why, in that case, did organisations even exist. His answer was that contractual relationships have transactions costs. When these transactions costs outweighed the expense of organisation, organisation would predominate. Also, there were limits to transaction; it might be actually impossible to specify what was wanted in a contract, or equivalently, it might cost too much to write it.

As often happens, the first half of this insight was more successful than the second. Since the 1980s, there has been a global trend towards replacing organisations with networks of contracts. The idea that a firm could be considered as a network of contracts was taken up by the management consulting industry, and strengthened from a positive observation to a normative statement that firms should become more so. In as much as anyone bothered with Coase’s corollary, it was simply to say that there was some sort of “core business” in there – presumably it was thought to be the zone in which transactions costs got high enough to demand organisation – and everything else must be contracted out.

In many ways, we’ve lived through a giant experiment in proving Ronald Coase wrong, which has now failed.

One very large, and especially purist, example of this was British railway privatisation. This did not just transfer a firm from the public into the private sector. Much more importantly, it transformed one large firm into a large number of smaller ones that interacted on a contractual basis. This system further interacted on a similar basis with the government.

When a management problem arose – for example, a train was late – a claim would be raised by one actor on another. For example, the Department for Transport might invoke a contractual penalty because the trains were late. The train-operating company would immediately claim against the infrastructure operator, which might counterclaim. Because the train leasing company might have guaranteed a certain on-time service level under a total outsourcing arrangement with the operating company, it too would then try to claim against anyone else it could think of.

This had important consequences. First of all, the claims-management process was itself costly. This is Coase’s basic argument. Second, because the prices of services exchanged between the component firms were often determined after the event, through the claims process, they were no longer informative about the marginal costs involved, but rather about the contract-management process. As a result, costs overall rose substantially although nobody could put their finger on who was coining it. Thirdly, it simply became enormously complex. A contract, after all, is executed between parties. The number of pairwise interactions within an organisation rapidly becomes very large – in fact, it increases by the factorial of the size of the organisation.

These three phenomena will become very familiar. The first is just the administrative overhead of the contracting process. The second and third are actually much more important. It will always be very difficult to get more efficient if you don’t know what your costs really area. This is a source of long term dynamic inefficiency. A major motivation of taking track maintenance back in-house was just trying to get an idea of what it actually cost. And Coase’s logic interacts in an interesting way with the economics of knowledge. If you believe a lot of relevant knowledge in an organisation or market is implicit and tacit, well, that’s by definition the sort of thing you can’t write into a contract. Either the firm has to exist in order to be the vessel of this knowledge, or else we don’t care.

Also, as we will see, in Coasian hell it is usually impossible to finger any particular guilty party, because its problems are system-level properties, driven by the interactions between firms in the system. Reductionism just leads to finger-pointing.

Healthcare in the United States is an especially egregious example of this. Americans, notoriously, spend much more than any other nation, have worse results, and leave lots of people uncovered. People blame, variously, insurance companies, doctors, drug companies, intermediary organisations, public policy, and patients themselves for getting ill. But none of this has ever solved anything. Everyone who has tried to nail down exactly what costs so much money has ended up concluding that the whole system is weirdly expensive and wasteful. That is, of course, the point. Its awfulness is an aspect of the system, not any one component or group of components.

For a worked example of this, let’s turn to this Harvard Political Review piece about insulin. You’ll observe that all three elements are in play. There is enormous complexity. There is enormous administrative overhead. And there is nothing as simple as a meaningful price.

If the patient is enrolled in a high-deductible plan, like 29% of workers with employer-sponsored insurance, they will pay full list price out-of-pocket until their deductible is met. After a set amount of time, the manufacturer will remit a rebate to the PBM worth around $200 at the time of writing. PBMs keep about 10% of this for themselves and pass the rest to the insurer.

From there, the story gets murky. Insurers, according to the PBM CVS Health, are trusted to use this rebate “to lower overall member benefit cost.” But no one enforces this theoretical insurer benevolence. What we do know is the patient often does not receive that rebate directly, even though they paid for the drug in full.

The purpose of the system is what it does, as they say, but it’s worth noting that Stafford Beer’s aphorism refers to the system, not to any particular actor within it. Even trying to bear down on the pure administrative overhead is likely to run into the problem that, although hordes of claims managers, lawyers, and claims-management software developers are a parasitic load on the whole system, they are vital for any given hospital, insurer, or whatever. Therefore, the system is likely to unite in homeostatic self-defence against change unless some drastic triggering event intervenes.

We saw such a thing at Carillion last week. Carillion plc was a pure creature of bastard Coasianism. In the 1990s, the British government fundamentally changed how it dealt with private suppliers, in this direction. PFI is the icon of this, but the change was much wider and deeper. Essentially anything a government department wanted had to be tested for contracting-out. To a large extent, the contracting process itself was handed to contractors via the prime contractor business model.

Carillion consisted, essentially, of a sales and contract management organisation that hunted public-sector service contracts and then hired subcontractors to carry them out. This is a fairly pure statement of the firm as a network of contracts. You’ll note that the customer-provider relationship has itself been outsourced here – the contractor, not the customer, chooses the provider. Therefore, any relevant information had better be within the contractor, because it’s not anywhere else. (Here is a fine example of this.) It grew largely through a succession of mergers and acquisitions, buying up the facilities management divisions of British construction companies to get their government contracts. This had an important effect on the company – it became a conglomerate that had only one real specialisation, bidding on government contracts. It is not surprising that it didn’t do a great job.

This model, known as a prime contractor, emerged in the 1980s in the context of the US defence budget. After the 1985 Goldwater-Nichols Act, the US Department of Defense was meant to do less development work of its own and rely more on the private sector. To this end, contracts for major defence equipment were no longer written on the basis of a manufacturer getting a contract to build aircraft or tanks or whatever as specified – instead, a prime contractor would manage the process of developing them, even of specifying them, and contract with other corporations to do the job. As a result, you wouldn’t necessarily go to Boeing to buy aeroplanes or Colt to buy guns. Instead you might buy armoured vehicles from British Aerospace, or ships from Marconi. The consequence of prime contracting was that the manufacturers became conglomerates, whose only specialisation was bidding on government contracts. They also became much bigger and more oligopolistic.

And you know what? It didn’t really work in defence, either. The tortuous and hellishly expensive development of the F-35, the first aircraft whose whole development fell in the prime contracting era, is a fine example of everything we’ve discussed above. The only difference is that it was too big to fail from the very beginning, so each setback resulted in more money. The next major project, the B-21, is being built mostly in a government-owned factory under a single manufacturer development contract.

But this didn’t stop the prime contracting model from invading the rest of the state in the 1990s. That’s what we’re going to discuss in the next thrilling instalment.

15 Comments on "In The Eternal Inferno, Fiends Torment Ronald Coase With The Fate Of His Ideas"


  1. I remember my then brother-in-law (who was a civil servant) explaining to me, about 20 years ago, how big construction companies, that you might think built stuff, were in fact specialists in managing contracts and getting somebody at two or three removes to actually do the work. Perhaps I should have listened more closely. It all sounded fine the way he described it.

    Reply

  2. I’ve been kicking around an idea for how to think about economic actors, namely that they can be divided into two groups: people and institutions. While institutions are made up of groups of people, in order to succeed (i.e., not go bankrupt), the people within institutions must learn how to operate less as people than as agents of the institution. In other words, successful institutions teach their employees how to maximize value for the institution itself. This by itself is not evil: a successful institution can and should reward it’s employees handsomely, in part to prevent those skills from leaving. But as institutions attain success, on the whole they behave less like people and more like the oft-postulated “rational actor”.

    A person can be appealed to on any number of levels (this is often called “marketing”). But as a rational actor, an institution can only be persuaded on the basis of self-interest. Or to put it another way, people within successful institutions act as fiduciaries of a fictitious legal-person and are legally compelled to behave as much like a rational actor as they possibly can.

    Reply

  3. Great analysis. There might be a fourth consequence, in your 7th para: the ‘network of contracts’ model was supposed to incentivise efficiency by creating liability for failure and reward for success. But because such failures are inevitable, it actually incentivises each actor to avoid the blame and create the appearance of success – ie it actually makes it harder to work out what’s really going on, and introduces a new source of inefficiency.

    Reply

  4. For those interested…
    This is how the US DoD sees the acquisition process itself

    https://www.dau.mil/tools/Lists/DAUTools/Attachments/203/Life%20Cycle%20Chart%20(Pre-Tailoring).pdf

    Here’s the parent link for those worried about going to a .pdf directly….
    https://www.dau.mil/tools/t/Department-of-Defense-Acquisition-Life-Cycle-Chart

    This may seem ridiculous on its face, but bear in mind that the DoD has different priorities than a commercial company, and it makes explicit many processes that a commercial company hides from consumers by large volume or doesn’t have to deal with (assuring multiple suppliers for two-decades, development costs, very stringent environmental tests)

    As an engineer (my bias), the real problem is not the Prime/Sub contracting structure itself, but there are several problems with the way the contracts/management relationship are structured and executed:

    1) You tend to be dealing with uncertain technical problems, but very specific technical performance requirements that gate payment. Thus there are massive incentives for subcontractors to hide, elide and even misrepresent technical issues they are experiencing to the prime, and for those that do make it through, the prime has those same incentives to hide, elide and misrepresent any issues to the government.

    2) The actual design reviews are administered by the contracting and legal staff so nobody says anything that prevents payment. The government technical oversight (often ALSO contractors who may eventually want to be hired by the primes) is can be pretty far removed from the day-to-day problems of the subcontractors.

    3) As technical needs change over the course of a long acquisition, the contract language has frozen the technical program, making changes very costly. For some, this is a feature, not a bug, but for a development program it can be stifling.

    4) The government program management oversight, often uniformed military, are shuffled from job-to-job every several years. So just when they’re developing the feel for the gigantic administrative and technical problems they face, they move on, and a new person comes in to steer the boat with a different set of priorities. This results in the institutional knowledege of an acquisition program residing with government technical experts who don’t have the same mission as the the government. Further, the uniformed service status metric here is $$$/people managed, not exactly a cost-saving incentive.

    So all that is to say there are changes in how the government manages its contracts which, in theory, could be addressed before chucking the whole thing out.

    It’s not clear to me that having the government in-house more of its R&D is a good approach, i.e. turning it into pork, (Space Shuttle, anyone?) is a better approach.

    Alternatively when you have big defense companies being more vertically integrated competing tooth and nail for giant contracts there are other problems– look up the history of the formation of the ULA (United Launch Alliance). At least with the Prime/Sub structure there’s something of a hedge and if you lose a big contract, you’ll probably be sub for some chunk of it.

    Any other ideas on how to do it right?
    –Josh
    https://en.wikipedia.org/wiki/Augustine%27s_laws

    Reply

    1. Considering how few big defense firns are left, surely they are only competing “tooth” nowadays, the nails not being necessary.

      Reply

  5. So, anyone who has followed me for a while on Twitter, or back in the EuroTrib days (not many, after all I’m sort of boring) will know I periodically go on a rant about Hayek and Information.

    I won’t reprise here, I’ll just say: Economists are in love with Hayek’s notion of how markets aggregate knowledge – but as soon as you put his assumptions under scrutiny, they fall apart.

    As you say: “Coase’s logic interacts in an interesting way with the economics of knowledge”

    Reply

    1. @Metetone — Hayek’s assumptions are exactly that. Take them away (e.g., externalities) and yes, of course his point falls apart, but that means you’re missing the value of his thought.

      @OP — great blog post. I teach Coase ’37 and this is exactly Coase’s point — 40 years before “information asymmetry” became a thing 🙂

      Reply

  6. Very very good; anyone who has worked for any sort of large organisation since the 1980s knows exactly what you’re on about. One quibble: the number of duplex pairwise links is 0.5n(n-1) i.e. it scales as n^2 not n!

    Reply

  7. I’ve been thinking exactly this way since I was taught transaction cost economics 15 or so years ago. I work in the UKs privatised rail industry and your description pretty neatly fits my own.
    It’s notable that theres been a huge rise in similar comments about Coase and privisation in the last few months.

    For my part, I think someone should be able to stretch Coasian thinking all the way to a justification for a flexible mixed economy of the type most of Western Europe had from the war to the 1980s. Of course hierarchy comes in several guises- partnerships, limited liability companies and indeed Government. .

    His starting point wasn’t “why we should have more markets”, as the natural spontaneous outcome of an efficient economy. To my mind Coase was seeking to explain not the emergence of markets but the emergence of hierarchy. It was that heirarchy was a natural consequence in the economy where transaction cost are high. So looking to create markets (like your management consultants or PFI privateer) is only as valid as looking to create hierarchy (like taking track in house or creating the NHS). So I think that it shouldn’t be beyond anyone’s wit to extend Coasianism to provide a neat market based justification for the work of Government.

    .

    Reply

  8. Good stuff, and oh so familiar. Just a little push back, in the same vein as Josh Model above: I don’t think the alternatives are obviously better. They just have different failure modes.

    For example, I have seen in-house expert teams decide that they do not need a prime contractor. They get better prices if they organize it themselves, and better control as well. Usually, they underestimate the work of oversight especially in the latter stages of execution, and then everything falls apart.

    Or organizations do have their own, more or less capable in-house ‘prime contractor’, and this department develops its own bloat and monopolist’s behaviour. Etc etc., there’s no simple trick to get it right.

    Reply

  9. “The number of pairwise interactions within an organisation rapidly becomes very large – in fact, it increases by the factorial of the size of the organisation.”

    Is this right? If the organisation has n members, the number of possible pairs = n!/2!(n – 2)! = n(n-1)/2 – so it grows with the square of n, not the factorial.

    Reply

  10. Please can you tell me the name of the font you use for your title and the quote.
    Also a very informative article, thanks.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *