What we do



Follow ODI

What Happens when Public Expenditure is Scaled Up? An Enquiry into the Costs and Costeffectiveness of Expenditure in Phases of Expansion

Working papers

This paper asks whether the costs of providing public services in low-income developing countries tend to rise when expenditure on these services increases - as has been found to be the case in developed countries. It offers elements of an answer to this question by reference to evidence from country case studies relating to episodes of expenditure growth, mainly in the 1990s, in the fields of primary education, road maintenance and child immunisation.

The hypothesis is enunciated, based on the standard microeconomics of the firm, and

assuming cost-curve convexity, that average (unit) costs are likely to rise in the long run for logistical reasons: even when services are produced and delivered efficiently, the resources required to provide services to remote locations and inaccessible clients are greater than those needed to reach more accessible areas and clients. Moreover, as short-run marginal costs are likely to be in excess of long-run marginal costs, because of immediate supply rigidities and unresolved technical inefficiencies, average costs may rise steeply when expenditures grow rapidly. If programmes require an initial phase of physical capacity construction before services can be expanded, the impression of rising average costs in episodes of programme expansion may be accentuated.

Much of the readily available evidence from OECD countries on costs in programmes on

which public expenditure is raised relates to education. Here, there is well-attested evidence of stepped-up expenditure being translated primarily into higher costs per pupil, without an attendant increase in enrolments or educational attainments. However, recent analysis of the productivity of public expenditure in the UK since public expenditure started to rise in 2000, following a period of expenditure restraint, shows a pattern of falling productivity in several sectors which enjoyed rising expenditure allocations.

These considerations suggest a methodology for empirical enquiry into the experience of low-income developing countries which first quantifies the unit and incremental costs of service provision, and then seeks to explain developments observed in terms of: initial conditions, cost structures, labour rates and input prices, and changes in technical efficiency, i.e. the ratio of output produced to inputs used. Measured unit costs tend to vary from year to year, partly due to lumpiness in periodic (capital) expenditures and partly because of measurement and recording errors. However, if the incremental cost over a period of time exceeds the range of annual unit costs, programme costs can unambiguously be said to be rising - and vice versa. If the incremental cost approximates to the average of unit costs, the inference is that costs are broadly constant.

To conduct the analysis, some measure of the output of each service examined is required. Services were chosen for the analysis because relatively homogeneous output measures for them are available on which data are usually recorded. In education, the measure adopted is the number of children enrolled in primary schools. For road maintenance, the denominator used is kilometres of road receiving routine and periodic maintenance. For immunisation, the true measure to use is the cost per fully immunised child; however, for practical reasons, the number of children receiving a full course of DPT3 injections is used as a proxy for the number of fully immunised children.

Country-level data are more abundant and readily available in the case of education than for road maintenance. Eight countries’ experiences of education costs are examined; historic evidence on road maintenance and immunisation is considered for only three and two countries respectively.

In education, six out of the eight countries exhibit rising costs in phases of expansion, in which the prime causal factor is rising real teachers’ salaries, the predominant cost

component. The reason for salary increases, however, has not in fact been a short-run

inelasticity in the supply of teaching staff. Government pay-structure rigidities in most

countries mean that the market for government jobs clears through quantities rather than via wage rates. The reason is rather that the qualifications, quality and motivations of teachers had previously been allowed to decline during years of fiscal austerity and inflation, and that it is necessary to raise salaries in order to raise teaching standards, without which enrolments would not grow to the intended extent. There have been concomitant but, in the absolute much smaller, increases in the costs per pupil of school facilities, learning materials and teacher training. One-off productivity increases are possible in education, and have to an extent been applied, but these are insufficient to counterbalance the effect on total costs of increased staff costs.

In road maintenance, the predominant cost element is the costs associated with the use of vehicles and equipment. Labour and materials are individually of lesser importance. In the three (African) cases examined expenditure increases have been occasioned by the establishment of ‘road funds’, and institutional reform has occurred with the outsourcing of maintenance works to contractors. Unit costs have been constant or have declined in episodes of expenditure growth because input and labour supply has been elastic, and because dynamic efficiency gains arising from contracting out, increasing competition between contractors and contractual innovations have counterbalanced cost increases caused by the growing inaccessibility of sites.

In immunisation, few countries have kept expenditure records, though output records are abundant. Of the two low-income countries with sufficient time series information to analyse, one - Bangladesh - reveals a broad constancy in unit costs over the years, while in the other - Nepal - the picture is inconclusive. Real prices of inputs used in the delivery of vaccines against the six diseases targeted in the Expanded Programme of Immunisation - representing nearly half of programme-specific costs - have fallen. Delivery has become more efficient as most vaccination now takes place at fixed locations, without the use of mobile clinics. These cost-saving factors have counterbalanced any higher costs associated with delivery to remote areas and arising from real staff cost increases. New vaccines and new routines whose use has now begun are increasing costs sharply.

The paper reaches the conclusion that, contrary to the hypothesis, there is no general evidence for rising costs in expanding public expenditure programmes in low-income countries. There is, however, a strong presumption of rising costs in education where staff costs are predominant, when achieving output and outcome targets requires continuing expenditure on raising teaching standards and staff incentives. In other sectors, there is greater scope for continuing efficiency gains sufficient in many cases to counterbalance the rising costs associated with extending outreach. When shortages in domestic inputs, including labour, occur, supply prices are unlikely to rise where the government has a high degree of monopsony. Quantity adjustment is likely to happen instead, with potentially adverse effects on the rate of feasible programme expansion.

For public expenditure planning purposes, however, there can be no rule-of-thumb

substitute for case-by-case analysis of the factors affecting cost, and of the scope and outlook for variations in prices, wages and factor productivity. At the same time, there could be useful cross-country research, as has been done for education in richer countries, to establish how far cost characteristics at the sector level are generic, and how far a function of local circumstances.

John Roberts