The Price of Being in the Room
When a rule can change, firms stop building and start lobbying for protection. The bill is far larger than anyone counts — and the obvious fix makes one part of it worse.
There is a moment, familiar to anyone who has worked near a standards body or a platform’s developer relations team, when a firm stops asking what it should build and starts asking who it should know.
The trigger is always the same. Someone circulates a draft. The draft would change something the firm has already built against. Nobody has decided anything yet, and the change may never happen — but the possibility is enough. Engineering time is reallocated to reading the draft. Someone senior starts attending the working group. A budget line appears for the trip. And a firm that had been building begins, instead, to manage its exposure.
Economics has a lot to say about the moment before this one. There is a large and excellent literature on how a group of firms comes to agree on a common standard in the first place — whether committees do it better than market races, whether firms get stuck on an inferior technology, whether talking to each other helps. What that literature mostly does not address is the moment after. Once a rule exists and firms have sunk money against it, who bears the cost when it changes?
The answer turns out to be stranger than it first appears, and the strangeness has consequences for how these institutions should be run.
Two things happen when a rule changes, and they are usually counted as one
Start with what a rule change actually does to a firm that has built against the old rule.
The first thing is obvious and everyone counts it. The firm has to adapt. Code gets rewritten, hardware gets requalified, documentation gets redone, two versions get supported in parallel for a while. This is real resource cost. It shows up on a budget. If you asked a CTO what a base-layer change costs, this is what they would describe.
The second thing is less obvious and almost nobody counts it separately. While the firm is busy adapting, its position in the market moves. The ground it occupied is briefly vacant, and somebody else occupies it. Sometimes that somebody is a competitor. Sometimes — and this is the case that matters — it is the party that decided to make the change.
These two things look similar from the firm’s side. Both feel like “the cost of the rule changing.” But they are economically opposite. Adaptation destroys value: the resources spent rewriting code are gone, and nobody gets them. Displacement moves value: what one firm loses in position, another gains. In a welfare accounting, the first is a real social loss and the second nets to zero.
Collapsing them into a single “cost of revision” is the mistake that makes this whole area confusing. Keep them apart and a set of results falls out that are not obvious in advance and, in a couple of cases, run directly against the intuition.
Standing is a purchased good, and it is pure waste
Now add the thing firms actually do about this.
They buy standing. They send delegates. They fund a secondee to the working group. They hire someone who used to work at the standards body. They pay for the membership tier that comes with early access to drafts. They cultivate a relationship with the people writing the specification.
None of this makes their product better. None of it makes anyone’s product better. It buys one thing: when a draft is written, it is written by people who are aware of what this firm has built, and the draft comes out in a shape that does less damage to it.
I want to be careful here, because it is easy to read this as an accusation of corruption. It is not. No money changes hands with the rule-setter. Nothing improper occurs. A specification written in a room where a particular firm’s constraints are well understood will simply tend to accommodate those constraints — not out of favouritism, but because the people writing it know about them and the alternative constructions are less salient. Representation works even when everyone in the room is behaving impeccably.
But from society’s point of view, the money is burned. It produces no output. Its only function is to shift where a loss lands, and the loss it shifts is largely a transfer that nets to zero anyway. This is a rent-seeking cost in the strict sense that Krueger (1974) gave the term: resources consumed in the pursuit of a redistribution.
So the full bill for a mutable rule has four parts, not one:-
projects that were worth doing and were not done, because the exposure made them unattractive;
-
adaptation expenditure by firms that proceeded anyway;
-
standing expenditure by firms that bought protection;
-
the cost of actually effecting the change.
The transfer of market position — the part that feels most dramatic to the firms involved — is the one part that is not a social loss. It is also, as we will see, the part that drives everything.
The rule-setter revises when it pays, and that changes the calculus completely
Here is where it gets counterintuitive.
Suppose the party that can change the rule also holds a position in the same product space. Not necessarily a sinister one — a platform that also ships first-party applications, a certification body whose members compete with those it certifies, a licensing authority with commercial interests, a consortium dominated by firms that also sell into the market. When it changes the rule, some of the position that dependent firms lose ends up with it.
Then the decision to revise is an economic calculation. Revising costs something to effect. It yields a gain proportional to how much exposed position there is to displace. The rule-setter changes the rule when the gain exceeds the cost.
That single observation drives the result. Because now ask what happens when rule changes become more damaging — a more brittle architecture, a deeper dependency, more specialised investment.
The naive answer is that firms are worse off. The actual answer is that the rule-setter revises less often, because a more damaging change drives more firms into protection, which shrinks the base it can extract from.
The two things move in opposite directions. As rule changes become more damaging (moving right), the frequency of change falls — but the exposure each remaining builder carries rises. The product does not fall.
The two effects go opposite ways and they largely cancel. Which means the obvious remedy — make each rule change less damaging, build in better migration paths, be gentler — reduces the damage per event and increases how often events occur. The burden on builders barely moves.
This is not an argument against good migration paths. It is an argument that good migration paths are not a substitute for governance. If a rule-setter has a reason to revise and revising is cheap, it will revise more often when each revision hurts less, and the firms downstream will end up in roughly the same place.
What does move the burden is the other margin entirely: how much the rule-setter gains from revising, and how much it costs to effect a revision. Those are not technical variables. They are procedural and appropriative ones — who is entitled to what when the rule changes, and what it takes to change it.
There is a nice historical footnote here. Farrell and Saloner (1988), studying committee standardisation, noticed that a firm proposing a standard often offers to license its related patents at nominal royalties, and that this facilitates agreement. In the framework here, that practice is precisely a reduction in what the proposer stands to gain from the change. Committees appear to have discovered the right instrument by trial and error, decades before anyone wrote down why it works.
The same change helps large firms and hurts small ones
The second result concerns who bears what, and it is the one with teeth.
A firm’s total expected burden has the two components we separated: expected adaptation, which rises when changes get more damaging, and expected displacement, which falls because changes become less frequent. Which effect dominates depends on how much position the firm has.
A small builder has little position to lose. Its burden is mostly adaptation, which rises. A large builder has a great deal of position to lose. Its burden is mostly displacement, which falls with the revision probability.
So the same institutional shift moves them in opposite directions.
Total expected burden against severity, for three project sizes. The small project’s burden rises. The large project’s falls. There is a project value at which the two effects exactly offset, and it can be located.
The practical implication is uncomfortable. An architecture that makes rule changes more consequential — deeper integration, more specialised tooling, tighter coupling — hurts the small builders in the ecosystem and helps the large ones. Not because of favouritism. Because the large ones’ exposure is dominated by a channel that shrinks.
And this compounds with something else. The firms that can afford serious representation are the large ones. So the large firms get both effects: their burden falls with severity, and they are the ones who can buy protection. The concentration is not designed. It falls out of the arithmetic.
Greenstein (1993) found something adjacent to this in federal computer procurement, and his finding has a twist worth repeating. He documented that agencies stuck with incumbent vendors even controlling for how well the vendor matched their needs, and that incompatibility between what was installed and what was being bought independently drove vendor choice. The twist is his IBM result: IBM’s apparent disadvantage with federal buyers turned out to be produced largely by incompatibilities between generations of IBM’s own product line, not by any procurement bias against it. Where a compatible system was already installed, agencies stayed. That is adaptation cost driving market outcomes, measured in the field.
The fix works, but only if the money goes to the right place
If the rule-setter’s private gain is what drives excessive revision, the obvious remedy is to charge for it — make the rule-setter internalise the damage its change causes.
That works. But there is a subtlety that turns out to be the whole ballgame, and I got it wrong the first time I thought it through.
Suppose you levy a charge on the rule-setter equal to the damage its revision causes, and the proceeds go to a general fund. The rule-setter now revises at the socially correct rate. Good. But every firm downstream still faces the prospect of losing position, so every firm still buys protection. The defensive expenditure is untouched. Worse, removing the misalignment removes the one justification defensive spending had — that it reduced a revision probability which was itself too high. So protection becomes unambiguously excessive, and you would need a second instrument to correct it.
Now suppose instead that the displacement component of the charge is paid directly to the affected firms, as compensation for the position they lose.
Everything changes. A firm that will be made whole for lost position has no reason to spend on protecting that position. The defensive expenditure collapses to what is justified by real adaptation cost alone. The rule-setter still faces the full social cost of its decision, because it pays the compensation. And the marginal firm’s investment decision is now the socially correct one, because the exposure it faces is real cost rather than a transfer it can be insured against.
One payment. Three margins aligned: the decision to revise, the decision to protect, and the decision to invest in the first place.
The allocation of the proceeds is not an accounting detail. It is the entire mechanism. A charge that funds the treasury fixes one margin and worsens another; the same charge paid to those who bear the loss fixes all three.
There is a further wrinkle about what has to be observable. Compensation aligns defensive spending only if a firm that protects more receives less — that is, only if the payment tracks what the firm actually lost rather than what it was expected to lose. A payment fixed in advance on categories like firm size or membership class is much easier to administer, and it will bring the right firms into the market. But it leaves the marginal incentive to over-protect completely intact, because a firm that spends more on standing still keeps the same fixed payment. The administratively easy scheme is precisely the one that fails on the margin that motivated the exercise.
Committees make it worse, and compensation cannot fix them
Everything above assumes the rule-setter is external to the firms it governs — a platform, a licensor, a certification authority. Standards committees are not like that. Their members are the firms.
That difference is usually waved at as “a different setting.” It is worse than that, and the direction is signable.
In a member-owned body, the gains from a rule change are distributed among the participants, and participation is what determines the share. So representation now has two returns rather than one. It protects your position, as before. And it increases your claim on what the change yields. Standing has become an investment with a return, not just insurance.
The first consequence is that firms spend more on representation than they would facing an external rule-setter, at every size, and more so where the distribution of proceeds is more tightly linked to how actively you participate.
The second consequence is the one that matters for policy. Compensation cannot fix this. Compensation works against displacement because it is indexed to the loss: protect more, get compensated less, and the marginal incentive to over-protect disappears. A share of the proceeds has the opposite structure. It rewards participation regardless of what you lost. No payment indexed to displacement can neutralise it, at any rate of compensation.
Which means the constitutional question — does participation in the body determine your share of what the body’s decisions produce? — is prior to any payment scheme. If the answer is yes, you can compensate displacement perfectly and still have an arms race in representation, because the arms race is being driven by the distribution rule and not by the exposure.
That is a hard result to like if you believe in participatory governance, and I do not think it argues against participation. It argues that participation and profit-sharing should be separated. A body whose members participate in decisions but whose gains do not scale with participation intensity avoids the wedge. A body where showing up more gets you more does not.
What this does and does not show
I want to be careful about the limits, because this is a model and models are narrow.
It concerns a single external rule-setter with a commercial position, or a member-owned body with an explicit distribution rule. It treats the transfer of market position as exactly zero-sum, which is a benchmark and not a general fact: a rule change that alters output, quality or consumer surplus will not have the transfer cancel in welfare, and the accounting has to be redone. It has one revision opportunity, so it says nothing about sequences of small changes that erode a position over time, or about reputation.
And nothing here is measured. The framework says what to measure — adaptation expenditure, displacement, standing expenditure, as three distinct things — and that separation is exactly what the existing empirical work does not do. Greenstein measured the consequences of incompatibility without separating them, and said explicitly that the degree of lock-in can never be identified from his data. That is still true.
What I think it does show is that a set of institutional questions that are usually argued in technical terms are not technical questions at all.
Whether a rule should be changeable is not really about migration paths. Whether representation should be open is not really about inclusiveness. What determines the burden on the firms building against a common rule is what the rule-setter gains from changing it, what changing it costs, and where the proceeds go. Those are constitutional choices, and they are made — usually implicitly, usually without anyone writing them down — when the institution is set up.
The firm that stops building and starts attending the working group has read the situation correctly. That is the tragedy of it.
References
Anton, J. J., & Yao, D. A. (2004). Little patents and big secrets: Managing intellectual property. The RAND Journal of Economics, 35(1), 1–22.
Chen, J., Doraszelski, U., & Harrington, J. E., Jr. (2009). Avoiding market dominance: Product compatibility in markets with network effects. The RAND Journal of Economics, 40(3), 455–485.
Farrell, J., & Saloner, G. (1985). Standardization, compatibility, and innovation. The RAND Journal of Economics, 16(1), 70–83.
Farrell, J., & Saloner, G. (1988). Coordination through committees and markets. The RAND Journal of Economics, 19(2), 235–252.
Farrell, J., & Shapiro, C. (1988). Dynamic competition with switching costs. The RAND Journal of Economics, 19(1), 123–137.
Greenstein, S. M. (1993). Did installed base give an incumbent any (measurable) advantages in federal computer procurement? The RAND Journal of Economics, 24(1), 19–39.
Krueger, A. O. (1974). The political economy of the rent-seeking society. The American Economic Review, 64(3), 291–303.
Smirnov, V., & Wait, A. (2004). Hold-up and sequential specific investments. The RAND Journal of Economics, 35(2), 386–400.