The basic argument for centralising decisions over public investment, for example the UK's £300bn plus National Infrastructure Plan, is that projects across the country can be assessed, and scarce funds then directed, on the basis of which are most valuable compared across the whole portfolio of potential projects.
A further advantage for centralised decision making is that interregional spillovers can be consider. For example, a project that should be high on Scottish priorities list is upgrading the A1 trunk road in Northumberland north of Newcastle, but it would be hard for Scotland to act on this preference if it only controlled spending in Scotland.
Centralised spending decisions are current practice in the UK. But what are the arguments against this centralisation, and instead having both a regionally balanced portfolio of public investment (which must, at the margin, lead to lower expected return projects being invested in at the expense of higher expected return projects), and local decision makers?
If factors are complementarity but there are costs of congestion, then a project can easily have the highest expected return (measured by willingness of the beneficiaries to pay for this improvement divided by its cost) but not be the efficient choice. Congestion (of fixed non-reproducible factors, typically land; and also of hard-to-adjust-in-the-short-term factors like space on a road, or bandwidth on a network) causes real losses in economic output, and a project that in the absence of congestion would be highly valuable (because it is complementary with all the other factors located in the congested region) may not yield greater real output than a similar cost project in a less congested region which is less valuable in terms of the willingness of the beneficiaries to pay.
A recent article in the FT, 'London’s $8.5bn traffic jam slows down growth' viewed further investment in the South as the solution to congestion in the South. It cites Edmund King, president of the AA, claiming support for this from those furth of the region: "“Anyone who regularly commutes on the M25 ..., knows it’s an absolute nightmare, ... But it’s a key link to airports and ports – and not just for drivers in the south. If you ask Scottish hauliers what’s the most important road in the UK it’s the M25.”" And obviously, once the airports and the ports are in the South, the value of the road network in the South is increased. But instead of constantly increasing capacity in the South to deal with the use of this network by those not from the South (and to deal with the population growth in the South induced by everyone crowding round these assets), it may be more efficient to invest in capacity elsewhere. Project evaluation should not always take demand as given, but instead should countenance that supply lead demand. And project evaluation should not always be conducted on a marginal basis: investing in infrastructure in the North may lead to increased values on future investment in the North, and perhaps some of this additional value should be recognised when evaluating the first project.
If congestion's economic impact is high enough then this could strengthen the case for targeting public infrastructure investments in highly productive and congested areas, precisely to limit congestion problems. However - this simply may not be possible in the medium and long run. If you alleviate the congestion through public investment London then, given the income differentials across the UK, London's population will grow until the congestion is recreated. If the costs of fighting congestion at a given location are convex (this may or may not be true, but it's certainly true that the level of congestion given fixed infrastructure is convex in the number of users), then this is, eventually, an inefficient use of public capital.
# Network effects and lock in:
This complementarity between factors of production can lead to "lock-in" when the situation changes (technological changes, changes in the terms of trade, etc). A great example is Michaels & Rauch (2013) (described in VoxEU) which discusses the positions of medieval towns in England and France. Three facts underpin their analysis: (1) Both England and France had urban networks created by the Romans, based on complementarity with the Roman road network; (2) Roman civilisation collapsed much more strongly in England than in France; (3) The dominant transportation technology of the middle ages was by boat. They find that medieval urban locations in England were sited much more appropriately with the prevailing technology than those in France were. This suggests that France suffered from lock-in: it was never worth it to rebuild all their infrastructure at a more appropriate site because every new investment must be optimal conditional on the locations of all existing assets with which they are complementary. England benefited, in the sense that it could re-optimise, from having its slate wiped clean.
Perhaps a diversity of locations should be maintained, and their varying idiosyncratic characteristics can provide insurance against secular changes. By having multiple centres that each have a full suite of infrastructures and assets, the costs of switching focus from one to another as external changes occur are minimised and so we do not become locked in to using a location that is eventually sub-optimal.
The power of markets to aggregate widely dispersed knowledge is justifiably lauded. Given this ideal however, we should not want the projects assessed by a single body with monopoly power to decide which get funded or not. This single decision making body will have a particular set of preferences, a particular information set, a particular expertise, and a particular method of analysis. As Chris Dillow says "Let's suppose that we want to find the best possible policy, according to some objective criteria ... Suppose too that there is bounded rationality and limited knowledge and that each individual selects the best option using his own information set and decision rule. In these conditions, each individual, if s/he is moderately competent, will find a local maxmimum - the best option, given his/her information and decision rule. But local maxima aren't necessarily global maxima. ... experts might well not find that global maximum because their decision rules and information sets might not be wide enough to encompass the best option: this might be because of deformation professionnelle, or groupthink or simply because their Bayesian priors limit the number of options they search for. Instead, widening the population of searchers increases our chances of finding that global maximum, because doing so brings more decision rules and information sets to bear on the problem. Cognitive diversity - in the sense of different ways of thinking - can therefore beat experts. It increases our chances of finding the best option."
In this context, multiple decision making bodies, each finding their own constrained local maxima, may bring us closer to the global maximum than a single decision maker (who does admittedly operate with fewer constraints).
# Fiscal transfers and stability
Presumably public investment opportunities in the highest yielding projects are lumpy over time. This could represent a big swing in fiscal transfers, and there is always a balance of payments (even if no-one is measuring it regionally). Any sudden positive (negative) impact on fiscal transfers would need compensating outflows (inflows) of private capital, increases in net imports (exports), appreciation (depreciation) of land values & other fixed assets, or labour in(out)-migration - all of which can be destabilising with associated losses. Conversely, if public investment were stable regionally then it would contribute to the automatic stabiliser effect of public spending and countercyclical policy.
# Distribution of gains
Even ignoring all the points above and assuming that a single decision maker can accurately select the highest yielding projects, everyone agrees on these, and there are no negative effects from congestion, then it is still not the case that we should automatically use public funds, raised from general taxation, to invest in a regionally unbalanced portfolio of projects. This is because the returns are not necessarily equitably shared. The obvious case is that of land-owners: those situated around the newly created public capital likely capture some of the value; but this is also true of workers. It is clearly the case that public funds can be used to systematically and predictably favour citizens in one part of the country over another. That public funds should be so used must be a matter for democratic debate, not necessarily the outcome of a spreadsheet calculation.
And the sums involved are not small: Wings Over Scotland estimates that investment in Scotland under the NIP will be £1bn from a publicly funded total of £136bn, to which Scotland will have contributed approximately 1/11th from tax revenues. If the populace as a whole agrees with these priorities and recognises these regional transfers, then fine. However, dressing this decision in the language of cost-benefit analysis overestimates how sure we can be that the selected projects are the optimal choice, and obscures the important distributional decisions that have to be more democratic than technocratic.
Given these points, there is clearly a case to be made in favour of some local decision making, and for reasonably regionally balanced public investment, on efficiency grounds. I believe that the balance in the UK at the moment is far too far away from regional balance, and towards centralised decision making. There is scope for benefits from a presumption in favour of a regionally balanced public investment portfolio. A centralised decision making body which has such a regionally balanced portfolio objective, would deal with all the headings other than "Information". Given this information point, there is definitely space to enhance local control over some public capital spending. Regional balance should not be perfect, since there is some value to picking high marginal value projects, and local control of investment budgets should not reach 100% since there is value to considering interregional spillovers.