Tuesday, October 14, 2014

The Ergas retort that wasn't

It is always flattering to be noticed. So it was with some glee that I saw that Henry Ergas in a letter to the editor of the AFR responded to some comments I made about the Vertigan review in an opinion piece earlier in the week.

The reply was brief so I'll repeat it in full here (including the title, which I thought was the best bit).

What cost NBN zeal, David Havyatt?
Where two bridges have been built, David Havyatt would destroy one to align reality with the theory of natural monopoly.
Such zeal in pursuit of the NBN caliphate would be commendable were it not to be realised at the expense of taxpayers and consumers. The fact is that the copper network and the HFC now exist and can both be upgraded at relatively low cost. Allowing them to compete might entail some duplication of costs but even if it increased them by an implausibly large 20 per cent, a one-year acceleration in broadband deployment and a 2.5 per cent increase in the rate of productivity growth would more than outweigh that impact.
Effects on this scale are well within the observed range of benefits from greater competition.
I am flattered by Mr Havyatt’s interest in my oeuvre.
Rather than selective quotation from submissions dealing with other matters, readers who wish to explore my views on natural monopoly might do better to consult my book Wrong Number.

Mr Ergas is particularly fond of the argument by analogy. I have previously been subjected to ones based on bridges in relation to vertical integration. My response to this one is to simply note that the correct analogy is that there exist two bridges that today combined cannot handle the combined demand of the current and future traffic estimates. The technical solution that constitutes the Multi-Technology Mix is a solution that bolts an extra lane onto each bridge. Labor's NBN proposed building a brand new bridge that can handle all that traffic and more. 

Mr Ergas and I would disagree on the wisdom of those two approaches. And that hinges on our different assessment of how certain we are about future demand. He and his colleagues in preparing the CBA part of their report place great store on the idea it may never be required to expand it to that degree. 

It is interesting to note that former Minister Stephen Conroy also liked to use a bridge analogy and talk about the Sydney Harbour Bridge and what it would be like if it was only one lane each way. Today we are struggling with the need for a third crossing.

The next part of Mr Ergas response is a hypothetical. It is a hypothetical based on the economist's approach of dealing only with the future costs - all historic costs being sunk and hence not part of the decision making framework. This is part of the philosophical underpinning of economics that is not part of my fundamental argument here - which is being conducted within the framework of the neoclassical norm. It is discussed in the footnote below.

The analysis he offers is based on an increase of costs from duplication - somehow limited to a cost increase of 20%. But to get genuine and complete facilities based competition the footprint for HFC would need to expand by a factor of 2 and the deployment of FTTN would need to increase by 33% above the costs incurred in the Strategic Review model. That is a lot more than a 20% increase in costs. 

The next part is pure faith. The incentive from competition would supposedly on its own result in a one year acceleration in the roll-out and a 2.5 per cent productivity increase. These, we are told, are effects that are within the scale that has been observed from competition. This I simply don't understand, for two reasons. The only competitive network deployment I've ever observed was the deployment of competing HFC networks. These were indeed rapid - but they suddenly ended in stand-off. That's why HFC only covers some 35% of premises. And competition dramatically increases revenue risk, and that drives up financing costs of both debt and equity. So competition's first effect is to drive productivity the other way.

Mr Ergas says he is flattered that I showed an interest in his wider body of work. I could delve into it in much more depth but suffice to say at least I'm well aware of his little book Wrong Number. Indeed, I wrote a review for the Australian Journal of Telecommunications

Let's interrogate what Mr Ergas has to say there about natural monopoly. At page 34 he writes:

Where the access provider's facilities are genuinely a natural monopoly - that is, a service whose costs are minimised if it is provided by a single firm - duplication may still be desirable (because the
allocative and dynamic efficiency benefits being brought by competition might outweigh the cost savings in production by monopoly.) 

This is a very big call. 

Allocative efficiency is the efficiency gained by society's resources being employed to make output match the preferences of society. If the service is a family of services there might well be an argument that the process of competition will result in firms adjusting their prices to match the preferences. The practical reality has been that competition has resulted in reduction in efficient price discrimination. In long distance telephony competition saw the elimination of cost reflective charges based on distance and on time of day pricing. This is because, at the margin, a competitive firm can always increase profit by slightly widening the off-peak "window" or increasing the distances in charge bands.

In the case of the NBN issues, become even simpler because the industry design has been to limit the monopolist to only those services absolutely necessary to be in the one firm. This is the motivation for both the operation at Layer 2 and the choice of 121 points of interconnect. As a fundamentally single product firm there is little allocative efficiency to be gained within the provision of the services.

More specifically the allocative efficiency loss is the presumption that a profit maximising firm will reduce output to below the efficient level. This is the role of regulation. However, as will also be discussed below this includes some specific assumptions about the incentives of managers.

Dynamic efficiency is a far more problematic concept. This is an attempt by economists who otherwise deal in static equilibrium models to address the fact that efficiency changes over time, and that importantly investment decisions made today will affect efficiency in the future.

It is interesting that Ergas should pursue this line because in all other work he has been a promoter of real options theory. Most particularly, this is usually described as the value to an incumbent firm of the value of delaying an investment decision rather than of making it. The Vertigan panel did not include a quantified option value in its Cost Benefit Analysis (because you can't quantify it), but did refer to the option value of delay in making its case for the Multi-Technology Mix.

The practical examples of a monpolist attempting to delay but competition forcing action date as far back to the 1970s when Telecom Australia was slow to introduce fixed point to pint data services which promoted much of the initial deregulatory thrust. More recently it was Telstra's competitors who first invested in ADSL2+.

It is hard to accept that the dynamic efficiency benefits from competition are worth pursuing when the proponent puts so much store in the value of delay. 

Ergas reveals that his real issue is with the comparison between regulated monopoly and competition when he writes (at page 98):

Productive and/or dynamic inefficiency is more likely to arise from the regulation of monopoly than from the fact of monopoly per se.

And on this Ergas is absolutely right. The critical issue is the form of regulation. 

One of the conclusions that those who developed the initial NBN policy reached was that there was no form of regulation that managed efficiently regulation of access to the "bottleneck" elements of a vertically integrated telco incumbent. It comes down to the fact that if it is possible to grow the total market, the incumbent will make that decision based on marginal costs. But the competitive firm under access pricing will always face a version of an average price.

That is, structural reform is a key aspect of changing the regulation.

Ergas addresses the structural question at some length at pages 164-7. He notes that "vertical externalities" can result in significant inefficiencies in price, service quality, investment and ongoing adaption to change. These arise because the upstream and downstream firms are both making investments but if they are not aligned they are inefficient.

The same thing can, and does, happen inside single firms. If the factory only makes green widgets and the sales department only sells red widgets you don't get much sold. You need to design your incentive scheme correctly to avoid these issues.

Normally we rely upon the market and price signals to transit this information. The problem arises if there is significant market power in both the upstream and downstream markets where monopoly pricing trumps market pricing. The consequent problem is called double marginalisation. But the issue disappears if the upstream firm is a regulated monopoly and the downstream firms are in a highly competitive market - the planned design under the NBN. 

The other three elements listed all come down to information flow. Experience with vertical structural separation has shown that the firms do struggle with the co-ordination issues. However, these are not unsolvable problems. For example, structuring prices for the monopoly elements can be done using a take-or pay contract structure following an iterative demand bidding process. This mimics for the industry the process of market price signals as opposed to regulatory determination.

Marginal economics, atomistic agents and the real world

My critique above has been framed within the paradigm of neoclassical economics. However, there are very good reasons why this dominant theme itself should be questioned.

Specifically this school is also known as the marginalist school, because all its theoretical insights emerge from what economic agents will do "at the margin." Questions of the type "Will I trade one pound of butter for one new gun?" underpin the construction.

From this analysis two big concepts are developed - demand and supply. In the ontology of economics these are actually imbued with an existence, they are not just analytical tools (the latter was the position Milton Friedman argued for in his piece on Positive Economics).

As a consequence of imbuing demand with existence economists will then undertake exercises such as a Cost Benefit Analysis wherein the net benefit to consumers is defined to be the whole area under this demand curve up to the point of quantity actually consumed. But in almost all cases the concept of demand is only well-defined in the immediate vicinity of the current market.

Consumers and suppliers both make decisions based on heuristics, or rules of thumb. It is technically impossible to access and process all the information to make a "fully informed rational decision." We are hard wired for this, in nature we often need to make decisions before being able to access additional data. And so preferences are formed by current and recent experience.

One of the things that informs us is our experience and expectation of price movements, both of the commodity under investigation and all others. As a consequence expected demand today winds up being a function of actual demand today. The system as a whole has all the properties of a non-linear dynamic system. The conclusion of that is that "demand" can exhibit rapid unexpected and dramatic shifts.

This is the mathematics of a bubble in asset prices. When house prices start being determined more by the expected value from a subsequent sale than from the "real" value of either potential rental income or rental outgoings foregone, then they rise rapidly. At some, unpredictable point, the gap between the two values increases to the point where profit takers halt the rise. If prices ease of as a consequence of this mild change in demand, the overall "sentiment" can change so that pricing is almost exclusively based on "real" value. And hence the bubble is burst.

Marginalism also infects the approach to assets. As in the case of telecommunications assets, once the asset is purchased and if it has no scrap value - it is sunk - the marginalist regards it as no longer being part of the analysis. 

In doing so the marginalist commits a variant of Zeno's paradox of Achilles and the tortoise In a race, the quickest runner can never overtake the slowest, since the pursuer must first reach the point whence the pursued started, so that the slower must always hold a lead. (as recounted by Aristotle, Physics VI:9, 239b15)

Because by the time the next decision is made the expenditure thus far is sunk the marginal upgrading of an existing asset will always win out over a decision to replace the asset.

A related problem for the marginalist school is its approach to economic actors as if they are all single atomistic agents. The particular problem this raises was first identified by Berle and Means who identified that the managers of firms - the ones who make the actual decisions - were not necessarily motivated by profit maximising. Growing their own status was one specific alternative goal - a goal to which was attributed acquisitions that grew the company and hence the CEO's importance without growing returns. 

The principal-agent model - in which a principle (shareholders) seek to ensure agents (managers) act in the principles interests rather than their own - has had its major impact in creating remuneration schemes that tie senior executive pay to company performance. In practice this is an attempt to ensure that the neoclassical economists assumptions about the behaviour of firms might be reflected in practice.

There have been many consequences of this, the most notable being a shift in the focus of firms to short term performance over long term performance - despite the fact that the bulk of investors are mostly concerned with long term returns.

But it is the implication for policy makers that is of interest here. The assumption of privatisation has been that the "profit motive" drives efficiency - despite the existence of the principal-agent problem. The issue becomes far more complex when the subject is also a monopoly that will be subject to regulation - because now there is a second principal to whom the management team are accountable. "Incentive regulation" was a device designed to create a pathway for managers as agents to meet the requirements of both principals.

It does seem, however, that the incentive issue is far simpler if there is merely one relationship - formed by public ownership.

In a piece for the Economic Society Mr Ergas did a nice job of describing the different dimension of the principal-agent problem as it applies to public policy:

“The core of public finance”, as Jurgen von Hagen has succinctly put it, “is that some people spend other people’s money”. This separation between spenders and payers gives rise to a wide range of problems of accountability and control (which economists typically analyse under the rubric of ‘principal-agent’ problems), reflecting divergences of interest between these parties and the inability of voters and taxpayers to costlessly and perfectly discipline the behaviour of those who spend money on their behalf. These principal-agent problems are aggravated by the fact that the spenders themselves are not a monolithic entity. Even if spenders as a whole face the collective consequences of their decisions, each individual spending unit (such as a Minister, a Department or a territorial level of government) may view the stock of available public funds as a ‘common pool’ (like an open seas fishery), which it can draw on at a fraction of the resulting opportunity cost while still garnering for itself all or the bulk of the political benefit. The scope to transfer the costs of wasteful projects to future generations, which have little or no voice in the political process, as well as to future governments (which will bear the political consequences of ‘pulling the plug’ on failed ventures), then makes the risks of inefficient outcomes all the greater.

From this Ergas made the case for formal project appraisal as part of the control mechanisms to manage these risks. This year's Nobel Prize Winner in Economics Jean Tirole made similar observations in his analysis of privatisation (in Incentives for Procurement and Regulation with Jean-Jacques Laffont). There the concern is the inability of the principal in the case of public owners to make long term commitment to objectives for the public enterprise.

The issue here though is the treatment of principal-agent issues as add-ons or simple tools for critique of other outcomes. Principal-agent issues need to be dealt with in the core of the economic analysis. Market design needs to not only include design of the institutions but also the incentives to apply. 

In the specific context of the NBN relying on the incomplete contracts (in an economic sense) devised by lawyers without effective incentive regimes is a major flaw in the policy development to date.

This is a convenient point to wrap up this much wider discussion.

However I do want to touch on one recommendation made by both the Harper inquiry into competition and the Vertigan panel - that access regulation for all regulated industries be moved to one regulator focussed only on that task. This is nowhere near as new and novel as suggested - it was the original intention behind taking access issues from AUSTEL to the ACCC. But it is interesting in the light of one of the key items noted in the citation for Tirole's Nobel Prize.

The progress in these areas largely reflects two methodological breakthroughs: game theory and the theory of mechanism design.2 By the end of the 1970s, the time was ripe for applying these tools to the major issues of imperfect competition, regulation, and competition policy. Over the next decade, many economists were drawn into these fertile fields. The analytical revolution was to a large extent a collective effort but, among many contributors, Jean Tirole stands out. No other scholar has done more to enhance our understanding of IO in general, and of optimal policy interventions in particular.
Although general theories can be of great value, in the end all regulation must be industry-specific. This point is illustrated by example in Laffont and Tirole (2000), where they consider the regulation of the telecommunications industry, as well as in Tirole’s studies of other industries, ranging from banking to electricity. The research on the regulation of specific industries illustrates Tirole’s exceptional ability to grasp the central features of an economic environment, to formulate these features mathematically, to analyze the resulting model, and to produce normative conclusions of great practical significance.

(Emphasis added) 

(My own copy of Laffont and Tirole's Competition in Telecommunication (2000) is autographed by both!)


No comments:

Post a Comment