Can we assess the distributional impact of cuts to spending on public services?

Date: 08 October 2010
Publisher: IFS

The Government is planning a fiscal correction that by 2014-15 comprises 26% tax increases, 10% welfare cuts, and 64% cuts to spending on public services. A comprehensive analysis of how the pain from this fiscal tightening will be shared across groups requires distributional analyses of the tax rises, the cuts to welfare benefits, and the cuts in the provision of public services which it will entail. It is difficult enough to quantify the impact of the first two of these items, as IFS researchers did most recently here. But quantifying the third is much harder. In this observation, we explain why this is, and conclude that the existing studies which have attempted to evaluate the progressivity or regressivity of spending cuts must be interpreted very cautiously. A fuller analysis of the issues discussed in this observation can be found here.

To estimate the distributional impact of public services, we need to express in cash terms the extent to which households value public services, and how that valuation varies according to household income. A number of recent studies (Barnard 2009, Volterra 2009, Horton & Reed 2010) have attempted to do just this. In estimating the distributional impact of public services, the key approach taken by these studies is to assume that the value of a public service to a household is equal to the cost of providing it. The studies then allocate spending on public services - and hence the benefit of that spending - to households in proportion to the size of the household, correcting, where possible, for the fact that some public services (such as health, education, transport and housing) are used more heavily by some types of household than others. The general conclusion of these studies is that there is no reason to doubt the redistributive influence of public spending in general, and transfers in kind in particular.

This approach can be very useful, but there are some important caveats that must be borne in mind in interpreting any results. In particular, the cost of a particular service will not typically be the same as its value to the user. Supplying a household with a publicly-provided service that costs £100 is not the same as giving them £100 since purchasing that service is not necessarily how they would have chosen or have been able to spend the money themselves. This key difference between the cost of public services and their value to households means that different households might place very different values on a public service even if they made the same use of it. For example, while one family with children might be quite content with provision of, for example, an improvement to a local school in return for the £100 it cost to provide, another family with a child at the same school might have preferred half that amount spent, and a third might have preferred twice that amount. The second of these certainly values the service provided less and whether or not the third does depends on what options are available for further private spending on schooling. This variation in value is really important when analysing how the benefits of public services vary over the income distribution because one of the main things that affects valuations is precisely people's incomes.

The challenge of quantifying the benefits of spending on public services is even greater for spending on goods that are consumed collectively (known as pure public goods). Leading examples of these include national defence and environmental protection. Two recent publications (Volterra 2009, Horton & Reed 2010) allocate the spending on such items in proportion to the size of the households (and then assume that the value of the service equals its cost, as we explained above). While it is tempting to assume that the service provided by, the Ministry of Defence for example, is equally beneficial to all, this is far too simple. The services provided by Ministry of Defence spending are complex, involving protection of borders but also the ability to project military power abroad in pursuit of foreign policy interests and a means to defend in extreme circumstances the established political order. Not only are those with lower incomes probably less willing to afford these benefits, but the benefits themselves are not likely to be felt with equal keenness across the distribution of income. If military spending tends to benefit the well-off better than the less well-off, then allocating the benefit equally in cash terms will exaggerate the progressivity of military spending (and therefore the regressivity of cuts to military spending).

Even if all households feel similarly enthusiastic about the services provided by the Ministry of Defence, allocating an identical valuation in cash terms can still overstate the progressivity of military expenditure. To see this, consider an expansion in spending on defence. If the cash benefit is shared equally across all households the resulting quantity represents a greater proportion of the total income of the poor than it does the rich - implying that the expansion in military expenditure is a progressive move. Indeed, the expansion in spending on any pure public good will appear progressive, if the benefit is assumed to be equal to the cost and is spread evenly over households. This uneasy implication of the method of allocating the benefit across households highlights the fact that the underlying assumption is open to question. If an attempt is to be made to quantify the value of spending on public services, it is important to recognise that the cash valuation is likely to vary according to the income of the household in question. This discussion, abstract though it may seem, is a vital reminder of both the complexity of converting the benefit of public expenditure into a cash figure, and the potentially misleading implications of apparently innocuous assumptions.

It is worth emphasising one further point on the how the results from the papers listed above can - or cannot - be used to evaluate the distributional impacts of changes in public expenditure. It may be possible to determine a reasonable estimate of the distributional impacts of aggregate spending on a particular public service. However, the distribution of the change in spending on that public service might look very different. Simply because health spending tends to be progressive does not mean that every conceivable reduction in health spending is regressive; the fact that spending on higher education arguably benefits the rich more than the poor does not mean that every cut to higher education spending would be progressive. The devil will be in the detail; the precise composition and manner of implementation of the package of impending spending cuts will crucially determine the extent to which they are progressive or regressive.

Given the difficulties, conceptual and empirical, involved in the valuation of the benefits of public services, producing a single set of numbers that represent the distributional impact of the entire package of cuts to spending on public services necessarily involves making a series of contentious assumptions which may crucially impact the outcome; an alternative and no less plausible set of assumptions could yield quite a different outcome. For this reason, we suggest such exercises are interpreted with caution. The best approach is perhaps to put bounds on what constitutes a reasonable opinion about what the distributional effects may be rather than trying to tie things down to a single answer.