مسائل اصلی در بودجه بندی - The Basic Budgeting Problem

Approaches to Resource Allocation in the Public Sector and their Implications for Pro-Poor Budgeting Adrian Fozzard

 

. Introduction

Sixty years ago V. O. Key laid down a challenge for economists to resolve the ‘basic budgeting problem’ namely, faced with limited resources, ‘On what basis shall it be decided to allocate x dollars to activity A instead of activity B?’ (Key, 1940: 1138). He went on to suggest that solutions to this problem might be found through the application of economic theory. He warned, however, that a budgeter’s holy grail – an all-embracing theory of resource allocation that could be applied in practice – would probably prove to be a chimera since the problem of reconciling competing demands between different policy goals and interests was essentially one of political philosophy (Key, 1940: 1143). If that line of inquiry failed, Key proposed that solutions might be found through an improved understanding of the institutional arrangements by which resource allocation decisions are made, which would entail a ‘careful and comprehensive analysis of budget process’ (Key, 1940: 1144). Over the past sixty years, attempts to resolve the basic budgeting problem have been made from both these starting points. This has entailed a subtle reformulation of Key’s question. Initially, attention focused on the application of economics in the design of methods which could guide policy makers by defining the basis – the guiding principles and criteria – for allocation decisions . Subsequently, attempts were made to arrive at a better understanding of budgeting behaviour and institutional dynamics, identifying how – the process by which – resource allocation decisions are and should be made (Chapter 3). At the same time, the analytical framework for analysis of the basic budgeting problem has broadened. It is now recognised, following Musgrave (1959), that solutions to resource allocation cannot be abstracted from other functions of the public expenditure management system, namely the pursuit of macro-economic stability and efficiency in the use of public funds. From the 1970s the problem of macro-economic stabilisation dominates the literature and resource allocation is, for the most part, treated as a secondary issue. Similarly, it is no longer assumed that budgetary allocation decisions are automatically transformed into budgetary outcomes. Resource allocation in the public sector is determined by both the criteria and process of decision making and the process of budget execution. Inevitably, this has widened the institutional scope of the basic budgeting problem. Whereas attention once focused exclusively on core policy institutions – the legislature, Ministries of Finance and spending agencies – it is now clear that departments within spending agencies, right down to the field level service delivery units, also have a role to play. Changing approaches to an old problem are not merely of academic interest. All of the approaches to the basic budgeting problem – whether normative or positivist in intent – have influenced the design of budget institutions, procedures and analytical methods. Changes in budget practice have, moreover, tended to proceed incrementally and cumulatively, so that many of the innovations introduced in early reforms are still in place today. Thus, today’s budget governance structures are essentially the same as t0hose introduced in the late 19th and early 20th centuries when modern budgeting systems were first established. Similarly, the analytical methods and process proposed by rationalists in the 1960s continue to be used today. Indeed, the rationalist approach is still the prevailing paradigm for policy makers. Consequently, an understanding of the various approaches to the budget problem continues to be relevant today, even where the validity of these approaches has subsequently been questioned, and research on these approaches is still ongoing. 2. The Basis of Resource Allocation This chapter provides an overview of the guiding principles that have been proposed as the basis for resource allocation decision making in the public sector and techniques developed to facilitate their application. None of these principles can provide an all-embracing theory of budgeting since the basic budgeting problem is multi-dimensional and has to be tackled simultaneously from various perspectives. One approach focuses on the comparative advantage of the state in the economy, identifying the underlying rationale for public interventions through an analysis of the conditions of supply and demand for public and private goods (see Section 2.1). Another seeks to prioritise alternative applications of public funds by applying the principle of marginal utility using measures of cost-effectiveness (Section 2.2). This principle can be extended to embrace the maximisation of utility through an assessment of the net social benefits of public spending using cost benefit analysis (see Section 2.3). An alternative approach recognises the primacy of citizens’ expenditure preferences and seeks to develop mechanisms of collective decision making so that these can be communicated to decision-makers (see Section 2.4). Lastly, the basic budgeting problem can be seen as a problem of resource redistribution in order to address social equity and poverty concerns (see Section 2.5). These principles and the analytical techniques which they have generated are complementary and a technically sound process of resource allocation decisionmaking would apply them all. Nonetheless these techniques can only provide imperfect technical solutions. Ultimately, resource allocation entails a political process in which economic principles and technical methods may play a small part in determining the outcome. This decision making process is the subject of Chapter 3. 2.1 Public goods and the rationale for public intervention In a perfect market, an efficient allocation of resources will be achieved by the forces of supply and demand, through the price mechanism, without the need for public intervention. However, public intervention may be justified in cases of market failure, where the price mechanism results in an allocation of resources that diverges from the social optimum. This may occur for a number of reasons: in the case of public goods, externalities, natural monopolies or asymmetrical information. The appropriate public sector response – distinguishing public provision, financing or regulation – and level of public spending will depend on the type and degree of market failure that the public sector seeks to correct. Public goods, club goods and mixed goods For Samuleson (1954; 1969), the distinction between private and public goods provides the underlying rationale for public expenditure. Public goods are non-rival (consumption by one person does not reduce the supply available for others) and non-excludable (users cannot be prevented from consuming the good). These characteristics prevent the provider of public goods from charging consumers for their consumption and so, if they are to be provided at all, they must be provided by the public sector. Defence and rural roads are often cited as examples of pure public goods. Most goods and services do not fully satisfy Samuelson’s criteria of non-rivalry and nonexcludability. Club goods, for instance, are non-rival up to a point of congestion and excludable. For such goods, the socially efficient level of provision may not correspond to the efficient level of provision for users beyond the point of congestion. This provides an opportunity and incentive for market provision through consumption sharing agreements, so that club members can maximise their benefits by excluding non-members. As a result, the level of provision is likely to be lower than would be socially desirable (Buchanan, 1965). Schools and other public services in which access can be restricted share these characteristics (Khumalo and Wright, 1997). This can give rise to a situation where public schools levy supplementary charges or use non-market methods to restrict access. Obviously, this has implications not only for the social efficiency of public provision but also for the social distribution of benefits, since wealthier individuals are likely to benefit disportionately from the more exclusive services. Mixed goods share the characteristics of both private and public goods, as is the case where private goods generate positive externalities. Positive externalities arise where the benefits of a particular good or service are enjoyed by both the purchaser and other individuals who do not contribute to the cost of purchase. Education generates both private and public benefits, the former through enhanced earning potential, the latter by creating a literate population which will benefit employers and promote social development. Since individuals will be prepared to pay for the benefits that accrue to them directly, but not for the benefits that accrue to other individuals, they will consume less of these goods than would be socially desirable. In these circumstances there is a rationale for public expenditure to subsidise consumption or for direct public provision so that a socially optimum level of provision and consumption is achieved. Determining how much the public sector should pay As a rule, the level of public spending on a particular intervention should correspond to the cost of the public goods it generates: since users will only pay up to the value of the private benefits they receive, the additional costs of public benefits will have to be met by the state (Musgrave, 1969). However, private benefits are likely to vary amongst individuals owing, for instance, to the ability of different social groups to transform the benefits of public services into meaningful improvements in their quality of life, such as higher income. In principle, willingness to pay provides a measure of these private benefits. In practice, perfectly discriminating price structures are impossible to design and administer and so it is usually easier to set user charges for a given level of services at a flat rate on the basis of marginal costs. In this way at least part of the cost of providing private benefits can be recuperated by the public sector. Some selectivity is needed to address equity concerns, through targeted subsidies (see Section 2.5) and to address differences in the quality of services arising from the rationing of access. Khumalo and Wright (1999) have suggested, for example, that transfers to South African schools should be provided as grants per pupil so that parents bear a larger proportion of costs in the ‘more exclusive clubs’ where pupil teacher ratios are lower. Selecting appropriate interventions and mechanisms for intervention For Pradhan – mimicking Keynes – the implications for public sector resource allocation derived from this analysis are clear: ‘public expenditures should be concentrated first on goods and services that the private market will not provide or will provide too little, rather than merely substituting for or even marginally improving upon the private market outcome’ (1996: 4). The role of the public sector in the finance and provision of goods and services is, therefore, residual. Goods and services should be provided by the private sector through market mechanisms where possible, since this will tend to be more efficient. Indeed, unless strictly necessary, public sector provision should be discouraged since it may crowd out more efficient private sector providers. On these grounds, the public finance approach will tend to favour market solutions and curtail the scope and scale of the public sector. Following this rationale, assessments of the scope for public intervention should be based on an analysis of the prevailing demand and supply characteristics of goods and services. For Pradhan this market analysis is the ‘principal, initial criterion in screening public expenditure allocations’ (1996: 31). The approach is necessarily reductionist, since the market characteristics of goods and services can only be assessed on a case-by-case basis. Using market analysis, public sector involvement in production, for example, can often be demonstrated as lacking justification on the grounds that conditions exist for private sector investment and market failures, such as those arising from natural monopolies, can be overcome with adequate regulation. Market analysis also allows policy makers to identify opportunities for private sector provision of services that have traditionally been considered as the public sector domain, such as agricultural extension, tertiary health care and higher education. In the case of Kenya’s agricultural policy for instance, SuthiwartNarueput (1998) demonstrates that a large proportion of public expenditures is allocated in the provision of private goods that could, and by implication should, be provided and financed by the private sector. Such analysis is contextual, since the conditions of demand and supply will be unique to each economy and can be expected to change over time. The demand for education, for example, is likely to be lower in rural communities where child labour makes a significant contribution to household income and there are limited opportunities for salaried employment than would be the case in an urban environment with a growing market for an educated workforce. These differences in market conditions will have implications for the private provision of education. Where market characteristics are not, at present, favourable to private sector provision and financing, the public sector can tailor its interventions so as to promote and facilitate market solutions. Pradhan, for instance, argues that regulation of the health insurance market so as to redress the problems of informational asymmetries may be more cost-effective than direct public provision and provide opportunities for private sector provision (1996: 52). The introduction of cost recovery mechanisms in the public sector can also be justified on these grounds since it creates an opportunity for alternative, more efficient private sector providers who are crowded out where the public sector provides services free of charge. However, in practice, transactions costs, the concern for equity and targeting difficulties (see Section 2.4), make it difficult to realise these efficiency gains for low-cost services, particularly in poor rural communities, where the imposition of user fees can lead to the exclusion of the poor. Prioritising between interventions While there is no doubt that analysis of market conditions helps decision makers identify appropriate and inappropriate public sector interventions, it does not inform policy makers how they should prioritise between interventions. Some guidance may be provided by the nature of the market failures that government identifies. If the goal of public policy is to ensure an efficient allocation of resources, the public sector should prioritise on the basis of its comparative advantage, obeying Pradhan’s exhortation to produce those goods that would not be produced by the market before those that would be produced too little. Following this logic, priority should be given to the provision of pure public goods before mixed goods and to mixed goods that generate substantial externalities – such as public health services – before those that generate substantial private benefits – such as tertiary health care. Prioritisation should also take into account the public sector’s capability, in terms of both the financial and human resources at is disposal. On this basis the 1997 World Development Report outlines a hierarchy of State functions, distinguishing: minimal functions, covering the provision of pure public goods and a safety net for the poor; intermediate functions, which include addressing externalities and other market failures; and activist functions, aimed at co-ordinating private sector activities and redistributing assets. As State financial and managerial capability increases, it may progress up this hierarchy of functions, so that all governments would provide pure public goods – law and order, public health, rural roads – and would accumulate and expand other functions – education, agricultural research and extension – as their financial and managerial capacity improves (World Bank, 1997: 26-27). The hierarchy is more as a useful guide to prioritisation at the lower end of the spectrum of state capacity, where the state is debilitated by war or extremely low levels of resource mobilisation, than it is for the majority of states, which fall into the intermediate category or aspire to an activist role. Since the vast majority of state functions fall into the intermediate category, the hierarchy is too broad a category to usefully guide the prioritisation of public expenditures. Where market failures are identified, political imperatives may compel Governments to intervene whether or not they have the means to do so effectively. This may lead the situation described by Tanzi (1995) where Governments adopt cheap but inefficient policies – in the sense of Tinbergen efficiency, where a modest change in policy leads to a significant change in outcome – rather than do nothing. Governments may, for instance, use regulatory controls rather than subsidies or direct provision, even though public spending on direct provision might lead to a more efficient allocation of resources. More often, Governments continue to ‘provide’ services even though it is patently obvious to service users that they lack the means to do so and, as a result, coverage is patchy and the quality of services poor. This over-extension of human and financial resources is one of the root causes of government failure in developing countries. Conclusion Ultimately, analysis of demand and supply conditions allows policy makers to distinguish appropriate and inappropriate public sector interventions on the basis of the comparative advantage of the State. It also provides some guidance regarding the relative priorities between interventions and the structure of cost sharing between the public private sectors. However, the approach is reductionist and, consequently, does not provide a basis for determining the appropriate allocation of public resources across the public sector. Nor does it indicate the appropriate level of public spending on individual interventions. Solutions to these problem are derived, in part, from the principles of marginal utility and an assessment of the net social benefits arising from public spending. 2.2 Marginal utility and cost effectiveness It is one of the tenets of classical economics that individuals will seek to equalise the marginal utility that they gain from each unit of spending across the range of goods and services they consume. In principle, governments should allocate resources on the same basis: ‘just as an individual will get more satisfaction out of his income by maintaining a certain balance between different sorts of expenditure, so will a community though its government. The principle of balance in both cases is provided by the postulate that resources should be so distributed among different uses that the marginal rates of satisfaction is the same for all of them … Expenditure should be distributed between battleships and poor relief in such wise that the last shilling devoted to each of them yields the same real return’ (Pigou in Key, 1940: 1139). Practical applications of this principle in the public sector presents a number of difficulties. Firstly, governments represent diverse interests, each with different utility functions, so that for some additional spending on battleships has a higher marginal utility than additional spending on poor relief, while for others the reverse may be true. Consequently, a perfect balance of marginal utility ‘may be possible only when a community is literally a unitary being, with the government as its brain’ (Premchand 1983: 44). Secondly, even if one accepts the notion of a unitary community, government is still left with the problem of constructing a utility function encompassing all the goods and services that it might be called upon to provide in order to derive the marginal utilities at various levels of expenditure. Informational constraints would render this global analysis impossible. Lastly – as Key (1940: 1137) pointed out – practical applications of the principle of marginal utility are hampered by the lack of a common measure of utility which would allow comparison of the utility derived from alternative applications of public funds. Relative effectiveness of public interventions One of the first attempts to apply the principle of marginal utility in a ‘theory of budgeting’ was made by Verne Lewis. Lewis argues that analysts should focus on increments of public expenditure, at the margin, since ‘this is the point of balance at which an additional expenditure or any purpose would yield the same return’. The relative value of these increments can then be assessed in terms of their ‘relative effectiveness in achieving a common objective’ (1952: 42). It is the task of politicians to determine this common objective and assess the relative effectiveness of alternative applications of public expenditure in achieving this goal. Budgeters can assist decisionmakers by presenting alternative proposals at varying levels of expenditure for each programme. In this way the trade-offs between alternative applications of additional funding can be revealed. Lewis argues that the concept of ‘relative effectiveness’ with regard to a ‘common objective’ effectively circumvents the problem presented by the lack of a common measure of utility. However, his solution has a number of shortcomings. Firstly, he fails to identify on what basis ‘relative effectiveness’ may be assessed, though, by seeking an explicit link between programme costs and outputs, he points the way to a solution. Secondly, it is unlikely that government policy can be reduced to a ‘common objective’, rather the public sector may address diverse policy goals. Although attempts have been made to develop broad inter-sectoral applications of the principle of ‘relative effectiveness’, applications have been more successful where they are restricted to the appraisal of alternative interventions in support of a single policy objective. Measuring cost effectiveness One of the most common applications of this principle is in measures of cost-effectiveness. These relate expenditures to the achievement of a particular policy outcome. The WDR 1993, for instance, uses Disability-Adjusted Life Years (DALYs), a measure of the number life years saved, adjusted to take into account of the suffering of disabilities such as blindness or chronic illness, to assess alternative interventions. Alternative interventions can then be compared and ranked on the basis of the cost per DALY saved. On this basis the WDR advocates that a larger proportion of public funds should be allocated to public health and a minimum package of essential clinic services than is currently the case in most developing countries and fewer resources should be allocated discretionary clinical services delivered at the tertiary level (WDR, 1993). Measures of cost effectiveness can be constructed for most government interventions, expressed either as the unit cost of the output of a programme (number of primary school graduates) or the unit cost of achieving of a particular outcome (level of literacy in a particular age group). Generally, measures of cost-effectiveness per unit of output are to be preferred, since the relationship with expenditures is likely to be direct, timely and more easily quantifiable, whereas the intervention of exogenous factors and time lag effects are likely to obscure the impact of additional expenditure on measures of outcomes. From a theoretical standpoint, the principal weakness of measures of cost effectiveness is that they do not reflect a ‘social’ utility function, reconciling the relative utility for different members of society. It is merely assumed that the outcome measured by the indicator is desirable and equally desirable to all. Furthermore, measures of cost-effectiveness focus on only one of the benefits arising from public interventions, ignoring other aspects of provision. DALYs, for instance, will only measure the health outcome of interventions and fail to take into account the extent to which the intervention treats clients with dignity. Nor will measures of cost effectiveness capture the externalities generated by the intervention – the secondary benefits accruing to other individuals. Consequently, measures of cost-effectiveness provide an incomplete basis for comparison between alternative interventions. From a practical standpoint, the most significant limitation of measures of cost-effectiveness is that they can only be applied narrowly to interventions in support of a particular policy goal: interventions intended to reduce morbidity and mortality cannot be compared against interventions intended to improve literacy. The problem of how to weigh-up up the relative costs and benefits of interventions contributing to different policy goals – such as arise in making inter-sectoral allocation decisions – is left unresolved since there is no common basis for comparison of the relative worth of alternative goals. Cross-sector applications Attempts have been made to establish a common denominator for the comparison of the relative effectiveness of alternative sectoral resource allocations. Ferroni and Kanbur, for instance, have sought to construct a decision-making tool that ‘permits the estimation of the opportunity cost in terms of poverty alleviation of allocating a marginal dollar to a particular sector or spending programme’ (1990: 1). This entails three stages of analysis: firstly, quantification of the impact each dimension of the standard of living (such as health, literacy or income) on the social valuation of the standard of living; secondly, quantification of the link between public expenditures and dimension of standard of living, essentially a measure of the cost effectiveness for incremental changes in outcome, expressed as a single measure of the standard of living; and thirdly an assessment of the proportion of public expenditure that reaches the poor. Using this approach, they suggest, it would be possible to assess the relative cost-effectiveness of alternative allocations of public expenditure in improving the standard of living of the poor. The fundamental problem of this approach lies in the construction of the common denominator: the social valuation of the standard of living. The weighting of the various dimensions of the standard of living is subjective, though some account can be taken of individuals’ relative preferences through direct consultation (see Section 3.4). Moreover, the approach requires analysts to assume that a single weighting can be constructed, thereby ignoring the variations in the relative preferences of different social groups. As such, the approach is normative in intent. Problems also arise in modelling the relationship between expenditures and outcomes. As Ferroni and Kanbur point out, one component of public expenditure can have an impact on several dimensions of standard of living simultaneously, the various dimensions of the standard of living can affect each other, and – as noted above – the relationship between expenditures and outcomes is likely to be complicated by a wide range of exogenous factors and timelag effects. Given the difficulties in constructing a common denominator of relative effectiveness across the public sector, practical applications are likely to be much easier where analysts recognise the diversity of policy objectives rather than seeking to consolidate all appraisal criteria within a single measure. Pradhan, for instance, proposes that analysts should assess the relationship between public expenditure and the outcomes of public interventions (such as the quality and quantity of education and levels of morbidity), so that the impact of an increment of additional spending on each intervention can be assessed, and then present the results in such a manner that decision-makers can ‘evaluate the trade-offs between alternative combinations of expenditure-outcome combinations’ (1996: 97). Although Pradhan suggests that attempts can be made to establish a common denominator using the principles of cost-benefit analysis (see Section 2.3) or by imposing valuations for specific outcomes, he argues that the presentation of simple, easily interpreted measures of the marginal cost per unit of outcome would, of itself, provide a useful input to resource allocation decision-making in government and the legislature and, ultimately, a guide for the individuals’ voting behaviour. Policy consistency and multiple objectives In practice, analysts do not assess the relative cost-effectiveness of alternative inter-sectoral expenditure allocations in the manner proposed by Pradhan. Instead, they apply the concept of relative effectiveness in a much degraded form by assessing whether interventions are consistent with government policy objectives. Indeed, consistency with policy objectives is often seen as the principal appraisal criterion for projects and programmes (Schick, 1998: 2). Obviously, in the absence of quantitative measures of effectiveness it is impossible to assess the relative effectiveness of alternative interventions other than on purely subjective criteria. Furthermore, without clear exclusion criteria, appraisals of policy consistency can embrace virtually all initiatives. Nevertheless, the criterion is sufficiently flexible and undemanding to have gained broad acceptance. Techniques have been developed to assist decision-makers assess the trade-offs between programmes where the government pursues multiple objectives. Multi-criteria appraisal, for instance, provides a framework for the assessment of trade-offs between programmes, accommodating the perspectives of different stakeholders. This creates opportunities for public decision making (Floc’hay and Plattu, 1998: 268). However, these techniques give no guidance regarding the basis of appraisal, assuming that ‘the analysis will arrive at good estimates of the impact of each alternative on all outcomes by the most reasonable means’ (Mohr, 1988: 199). Ultimately, therefore, the responsibility for determining allocation priorities is returned to the political arena with little guidance from technicians. Conclusions While the principle of marginal utility is a sound basis for resource allocation decision-making in the public sector, problems are encountered in the transformation of this principle into an operational tool. One approach is an assessment of the relative cost-effectiveness of alternative resource allocations in achieving a common objective. Measures of cost-effectiveness are widely applied, though these are restricted to the appraisal of alternative applications of funds intended to achieve a particular policy goal, usually within a single sector or programme. Although measures of relative cost-effectiveness can be constructed across sectors, their basis is normative and informational constraints have discouraged their use. Instead, decision-makers tend to use simple, subjective criteria in assessing the consistency of interventions with policy objectives, leaving it to the political process to determine the relative merits of alternative uses of public funds. 2.3 Allocative efficiency and cost benefit analysis If the concept of social utility or welfare is to be used as a guide to resource allocations across government two key elements must be in place: firstly, criteria and mechanisms for the reconciliation of differences in individuals’ relative utilities for different combinations of goods so that a comprehensive social utility function can be described; and secondly, a common denominator of utility as a basis for comparison of alternative uses of public funds. Measures of cost- effectiveness cannot provide the first of these elements, and offer only a partial – sector or programme specific – solution to the second. Comprehensive solutions have been found in the concept of allocative efficiency and the monetary valuation of costs and benefits, both of which are applied in cost-benefit analysis techniques. Allocative efficiency and the valuation of costs and benefits The problem of divergence in individuals’ utility functions can be resolved by applying the normative principle of Pareto optimality. Following this principle, public interventions can be said to demonstrate optimality, or allocative efficiency, where at least one individual is made better off and no individual is made worse off: there are only winners. Rigorous application of this criterion is impractical since it would be impossible to identify all winners and losers, losers would have an incentive to overstate their losses and the scope for public intervention would be severely restricted. Consequently, a potential Pareto optimum – the Hicks-Kaldor criterion – is generally applied by which an intervention is considered acceptable if the amount by which some individuals gain is greater than the amount that others lose, leading to a net-benefit, so that, in principle, winners could compensate losers for their costs. No actual cash transfer is required. An intervention may therefore be considered efficient even if some individuals lose, as long it generates net benefits (Boardman et al, 1996: 29-34). In this way, the principle of allocative efficiency is underpinned by an assumption that social welfare may be enhanced by the redistribution of resources within society, even where this entails redistribution from the poor to the rich. Monetary value can be used as a common denominator for the assessment of the relative merits of public interventions, taking into account their costs and benefits to society. The benefits of public interventions in the productive sectors, such as in agriculture and industry, can be determined with reference to increased production and valued on the basis of market prices, where an efficient market pricing mechanism is in place. Where government policy or local market conditions result in price distortions – such as those arising from monopolies, taxes or administered prices – equivalent border prices can be used. The problem lies with interventions, such as those in the education, health and defence sectors, that generate outputs and outcomes for which there is no corresponding market valuation. How does one, for instance, measure and then value the benefits arising from a reduction in mortality? or the cost of environmental damage? For economists, solutions to these problems can be found in the use of shadow prices: surrogate prices that represent the opportunity cost of a particular good. This may entail, for instance, the valuation of loss of human life on the basis of foregone income. Alternatively, value can be estimated based on measures of willingness to pay, using for instance the compensating wage differentials for employment with different levels of risk as the basis for an assessment of the value of life, or by using values revealed in questionnaire surveys (Cullis and Jones, 1998: 137-142). Obviously, all these methods have their shortcomings. Ultimately, the valuation of intangible costs and benefits – such as human life, human rights or environmental diversity – rests on subjective criteria. Cost benefit analysis Cost Benefit Analysis combines both potential-Pareto criteria of allocative efficiency and the monetary valuation of social benefits so as to generate a single measure of the net-social benefit (utility) generated by a particular intervention. In principle, the benefits and costs to all those affected by a particular intervention should be identified and valued, including those affected indirectly, as, for example, in the employment generated in producing materials for a road building programme (Boardman, 1996: 2-46). In practice, some limitation must be made on the scope of impact analysis, sometimes excluding indirect impacts such as externalities. The discounted net social benefits, or social rate of return, is then assessed, taking into account the benefits and costs generated and their distribution over time. The discount rate chosen is critical in determining the viability of the project: a high rate will reduce the stream of benefits, may penalise large projects with substantial start-up costs and will tend to penalise deferred benefits, such as those from environmental protection. Ideally, the discount rate should reflect the social-time preference, as a measure of deferred or inter-generational costs and benefits. More often the discount rate is based on the forecasted long-term rate of interest, adjusted to take account of risks and the opportunity cost of private investment. The discounted social rate of return is then compared with the status quo, alternatives and a threshold rate of return, in order to determine whether the intervention is viable and superior to alternatives. In theory, the calculation of a social rate of return would allow decision makers to bypass allocation decisions, since the investor could proceed with all feasible expenditures that have a higher rate of return than the long-term rate of interest for borrowing. In practice, since the number of interventions underway at any one time is limited by absorptive capacity and the ability to mobilise financing, choices still have to be made. However, the method does provide the basis for these choices: priority should be given to those interventions that generate the highest net present value. Assessing distributional impact Difficulties arise in assessing the distributional impact of interventions. The returns computed through Cost Benefit Analysis are social returns, to the economy and society as a whole: the method does not distinguish between private and public costs and benefits, nor does it take into account the distribution of costs and benefits among social groups. Thus cost benefit analysis may indicate the viability of a particular intervention on the basis of its social return, but fail to identify those benefits provided in the form of private goods (which should be financed by individuals) and public goods (which should be financed by the public sector). Private returns to education, for instance, particularly tertiary education, may be considerably greater than the social returns, suggesting a reduced role for public subsidies of this service (Appleton et al, 1996). If these distinctions are to be made, the technique is most usefully applied following an analysis of the underlying rationale for public intervention (see Section 2.1). The net-social value may also hide an inequitable distribution of costs and benefits: a small number may enjoy substantial benefits while the direct costs are shared by a large number, not just in terms of the costs of financing the project through general taxation but also owing to costs imposed through, for instance, displacement to make way for a dam or irrigation scheme. Moreover, the method, as usually applied, is either neutral as regards the social distribution of benefits or, where shadow prices are based on income, will favour interventions that benefit higher income groups disproportionately. These problems can be corrected by the use of distributional weights in favour of poorer social groups, by for instance, weighting costs and benefits accruing to a particular groups in proportion to their share of national employment or income (McGuire and Garn, 1969). However, such weights are ultimately set arbitrarily and will undermine the validity of the results as an indicator of efficiency (Boardman et al, 1996). Scope of application Cost Benefit Analysis is widely applied in the analysis of individual public expenditure decisions, particularly in the appraisal of investment projects, and especially where those projects will generate a stream of income. However, even at the level of the project, the validity of results may be compromised by the incomplete coverage of project impacts and the subjective nature of valuations: it is not uncommon for different analysts to arrive at different valuations of the same project. Given the wide margins of uncertainty regarding valuations, the method is most useful in distinguishing between those interventions that are clearly viable and those that are not, or in ranking alternative projects with similar characteristics – such as alternative power plants, irrigation schemes or dams – within the same sector. The method is impractical as a tool for inter-sectoral and inter-programme resource allocation decision-making. For Pradhan (1996: 96) the principal problem encountered in macro-level applications is one of valuation, particularly for those costs and benefits that do not have market prices or where market prices do not accurately reflect the social cost, since differences in the valuation of distinct outcomes – such as reductions in mortality and increases in literacy – will have a significant impact on the relative returns on these interventions. Problems also arise in scaling-up the method to address broad policy issues, such as the comparison of alternative crime reduction strategies or educational policies, or comparisons between sectors, where the wide range of options with complex impacts precludes systematic and comprehensive analysis. This restricts the effective application of the technique to ‘lower-level problems’, such as the viability of a particular investment project, where the impact is more easily identified, rather than ‘higher-level’ policy issues (Lindblom, 1959: 80). Owing to these analytical and information constraints, application of cost-benefit analysis in resource allocation decision tends to be reductionist and bottom-up. Resource allocations are determined through the appraisal of individual measures in isolation, rather than based on a comprehensive assessment of the opportunity costs of all alternative solutions. Conclusion Cost benefit analysis continues to be applied in the appraisal of large-scale investment projects, particularly those in the productive sectors in which costs can be assessed against a stream of income. Notwithstanding the sensitivity of the results to the scope of impact analysis, cost and benefit valuations, discount rates and the distributional weightings applied, the technique does provide a rigorous basis for decision-making at this level. Unfortunately application of the technique to higher-level, inter-sectoral and inter-programme allocation decisions is impractical owing to information constraints. Although the general principle of benefit valuation can be applied, this can only be considered an ‘approximate cost-benefit measurement’ based on the valuation of a narrow range of direct impacts (Pradhan, 1996: 96). 2.4 Citizens’ preferences and collective decision making Cost-benefit analysis relies on a technically constructed measure of utility as a guide to the optimum allocation of resources. An alternative approach would be to allocate resources according to citizens’ revealed preferences. In principle, the preferences of economically rational citizens will coincide with their utility functions. All that is needed is a mechanism by which these preferences can be revealed and aggregated. In the market, this achieved through the price mechanism, as determined by the forces of supply and demand. This solution is not generally applicable in the public sector (see Box 1), where citizens’ preferences are revealed through direct consultations and voting arrangements. Direct consultion Decision makers may solicit individuals’ preferences regarding resource allocations through surveys. From a theoretical standpoint, this kind of direct consultation is likely to provide a more accurate reflection of citizens’ preferences where the number of beneficiaries of public goods is small, since the cost-share borne by the citizen will be relatively large and failure to reveal demand is likely to reduce supply. In contrast, where the number of beneficiaries is large – as is the case for many public goods – the individuals cost share is minimal and failure to reveal will have little impact on supply; consequently the individual will try to free ride so as to benefit from the good without contributing to its cost (Cullis and Jones, 1998: 66). From a practical point of view, direct consultations can, at best, provide a partial view of individuals’ preferences for a limited range of alternatives. This suggests that direct consultation is unlikely to be an effective means of revealing citizens’ true preferences regarding broad expenditure allocations. More importantly, direct consultation does not provide a mechanism for aggregating individuals’ preferences across the range of spending options. Nevertheless, evidence from the USA suggests that citizens do have internally consistent policy preferences which could, in principle, provide some guidance to policy makers (Hansen, 1998; Jacoby, 1994). Survey information can also provide disaggregated information regarding preferences which can be related to income groups, providing the basis of a demand curve for public services which would suggest the appropriate level of expenditure (Preston and Ridge, 1995). Politicians in the United States – and increasingly in the United Kingdom too – do in fact make widespread use of polls, attitudinal surveys and focus groups as a means of gathering information on citizens’ preferences regarding public expenditures (Lee and Johnson, 1989: 97). Participatory Poverty Assessments can serve a similar function in developing countries, providing insights regarding citizens’ priorities for public expenditure. In Uganda, for instance, evidence that lack of water was a high priority for rural women gathered through PPAs led to the government increasing resource allocations for rural water supply (Robb, 1999; Norton et al, 2001). Box 1: Revenue earmarking and choice For Buchanan, the earmarking of specific revenues to a particular service allows the taxpayer to ‘sense or be conscious of a more direct relationship between his own tax payment and the benefits he expects to receive’ (1967: 22). This provides for more efficient choice since the beneficiaries of services can assess their relative worth and determine the level of financing accordingly. In this way earmarking reveals taxpayers’ preferences for public services and sends a demand signal to the public sector about how much of the public service to supply in much the same way as the market would. However, the validity of earmarked revenues as a demand signal only holds where the goods are excludable, so that no one receives a service without paying or pays without receiving, and so raises questions as to why the service is provided by the public sector at all. Furthermore, earmarking has serious limitations as a general principle for allocational decision making. Services which generate externalities are likely to be undervalued, since individuals will vote for expenditures or choose to pay for services up to the point of their direct benefits but not beyond. Earmarking can also lead to inflexibility in resource allocation causing inefficiencies where, for instance, earmarked funds have to be applied in road construction and maintenance even though this has a lower marginal rate of return than alternative applications of public funds (Gwillian and Shalizi, 1999). Notwithstanding these limitations, Musgrave (1986) has argued that the earmarking of revenues to broad areas of expenditure – so as to ensure adequate fungibility – would make the relationship between tax and public benefits more explicit and thereby facilitate an analysis of trade-offs between them. On the one hand this would make increases in taxation to fund popular expenditures more palatable to politicians and taxpayers, on the other hand it would help cap less popular expenditures. In practice, earmarking may not actually work in this way since increases in earmarked revenues to a particular activity may be largely offset by reductions in expenditures financed from direct taxation (Dye and McGuire, 1992). Collective decision-making For Musgrave, the problem of preference revelation and aggregation is best resolved by constitutional, voting arrangements. Where allocation decisions are subject to voting, taxpayers ‘are induced to reveal their true preferences … [because] knowing that the outcome of the vote will be mandatory, taxpayers will find it in their interest so as to have the outcome conform to their desires’ (1986: 77). Moreover, in contrast to direct consultation, voting provides a mechanism for aggregating preferences. The difficulty lies in the design of appropriate voting mechanism. Following Arrow’s ‘Impossibility Theorem’, it is accepted that all voting mechanisms – including majority voting – are imperfect in the sense that they are unable to satisfy all reasonable conditions, including rationality and equality (see Cullis and Jones, 1998: 76-77). Majority voting has the advantage that it is broadly accepted as the model in Western democracies. Unfortunately, as Mueller (1979) demonstrates, the majority voting procedure may not generate the Pareto optimum or potential Pareto optimum outcomes because the majority has an incentive to approve policies that redistribute resources from the minority regardless of the impact on allocational efficiency. This ‘tyranny of the majority’ may lead not only to inequitable and inefficient resource allocations, but also to higher than desirable levels of expenditure on public services, since the costs of excess provision benefiting the majority can be passed on to the minority. The only way that this can be avoided is to operate within ‘a constitutional framework that protects individual rights’ (Tideman, 1997). One means by which this could be achieved would be to allow all citizens a veto over policy decisions. However, this would require multiple voting to arrive at an agreed outcomes and so renders the method impractical as decision making tool. Again, the role of politicians in protecting the interests of the minority is crucial to the achievement of desirable outcomes. From a more practical perspective, it is unclear to what extent majority voting systems in representative democracies actually permit citizens to express their preferences. Voters must choose between candidates presenting manifestos which cover a wide ranging policy package that is unlikely to match their preferences on each allocational decision. This may lead to a situation whereby individuals are forced to vote for a policy which they dislike in order to secure an outcome for which they have a particular preference. In these circumstances, the ranked order of policy options may not correspond to individuals’ intensity of preference, so that the outcome of the vote may differ from the welfare optimum. The role of decision-makers and merit goods Efficient collective decision-making mechanisms would reduce politicians and bureaucrats to the role of passive executors of the collective will. However, even if politicians and bureaucrats accept the principle that they should comply with citizens’ revealed preferences, their intervention may be justified on the grounds that individuals are not always aware of and will not necessarily act in their best interests. In these circumstances, the public sector must intervene so that some goods and services are provided if even citizens do not considerable them necessary or desirable. The need for such merit goods – goods that are mandated regardless of individual and collective preferences – has been variously explained. Individuals might undervalue certain goods, such as primary education, because they are unaware of or are unable to assess its long-term benefits. Compulsory education may, therefore, be necessary to ensure that all children attend school. Citizens might also be considered as voluntarily delegating certain decisions to government on the grounds that they lack the competence to make these decisions for themselves. Alternatively, Musgrave and Musgrave (1989: 57) suggest that society ‘may give rise to common wants … [and] these obligations may be accepted as falling outside the freedom of individual choice that ordinarily applies’. Whatever the underlying rationale, the notion that the public sector should intervene contrary to individual and collective preferences smacks of paternalism and runs counter to the principles of citizen sovereignty and rationality that underlie economics. Nevertheless, it does seem consistent with many citizens’ experience of the public sector. Conclusion While few would question the principle that expenditure allocations should reflect the citizens’ preferences, the transformation of this principle into a practical tool for decision-making presents numerous problems. Most democracies have settled for representative systems in which decisions are made by majority voting. These systems will not necessarily generate socially efficient outcomes. Consequently, while direct consultation and collective decision-making may provide a guide to decision makers, some political and bureaucratic intervention, acting the broad public interest, is generally considered necessary to achieve the desired allocational outcomes. 2.5 Equity, incidence and targeting While there is a tendency for economists to regard the existence of market failure as the fundamental rationale for public expenditure, it is now accepted that the reduction of social inequalities and poverty is also a legitimate concern of Government and goal of economic policy (Tanzi et al, 1999). Analytical methods have focused on the redistribution of income, as measured by the net-impact of taxation and expenditure on household income and consumption amongst different social groups. This approach is consistent with a conceptual framework in which equality and poverty are defined in terms of income alone. Although this unidimensional characterisation of equity and poverty is now regarded as inadequate – since equity and poverty are regarded as multidimensional phenomena in which income is but one, and not necessarily the most important, facet (see Box 2) – the distributional impact of public spending remains an important criteria in the assessment and design of expenditure policy. Assessing the distributional impact of public spending The extent to which public interventions redistribute resources can be assessed by analysing the social distribution of the costs (taxation) and benefits (expenditures) of public interventions using benefit incidence analysis. Studies have tended to focus on the distributional impact of public spending. In the case of direct transfers, the impact can be assessed by comparing household income or consumption levels and the amount of transfer received by each household. On this basis it is possible to identify the proportion of the transfer received by different income groups and the contribution of the transfer to the household income of these groups (Jarvis and Micklewright, 1995). Spending on public services can be analysed in a similar manner (Selowsky, 1979; CastroLeal et al, 1999). The public service, such as education or health care, is treated as a subsidy, valued at either the average or marginal unit cost of service provision, ideally taking into account capital and recurrent costs. The distribution of this subsidy is then assessed on the basis of the usage of the service by income group, usually based on survey results. This subsidy can be considered as a transfer to household income, allowing an assessment of the extent to which the subsidy contributes to lifting households above a threshold level of poverty. More often, analysts focus on the distribution of aggregate public spending on a particular service by income group. Where information on service costs is incomplete or poor quality, data on service usage by income group allows an assessment of the extent to which the poor benefit in proportion to other income groups. However, cost information is needed if it is intended to compare and aggregate the impact of spending on different types of services (such as primary, secondary and tertiary education) by income group. Studies using benefit-incidence techniques have often revealed that the distribution of public spending on key services such as health and education is regressive. This is because a large proportion of public spending is allocated to secondary and tertiary level services which are used disproportionately by higher income groups (Castro Leal et al, 1998: Lloyd-Sherlock, 2000). The policy implications of this analysis are clear: spending on primary services should be increased relative to secondary and tertiary services if the intention is to benefit the poor. The advantage of benefit incidence analysis is that it is easily understood and has modest informational requirements. However, the method does have limitations (see van der Walle, 1998). The technique cannot be applied where there is no data on service use by social groups, as is usually the case of roads or police services. Reliance on aggregate cost information fails to take into account possible variations in unit costs or the quality of services – as reflected in class size, for instance – between service delivery units. This will tend to obscure differences in benefit Box 2: Multi-dimensional poverty Analysis of the impact of public spending on poverty and equity has usually rested on an assumption that poverty can be characterised and measured as a function of household income, or consumption as a surrogate of income. Amartya Sen – supported by the experience of the poor themselves (World Bank, 2000) – argues that this characterisation is inadequate. For Sen (1985, 1999) equity and poverty are better understood in terms of the substantive freedom and capabilities of individuals to achieve the things that they value. This leads to a recognition that equity and poverty are multi-dimensional phenomena, which embrace basic requirements such as being well nourished, sheltered and clothed and ‘such complex achievements as taking part in the life of the community, having a joyful and stimulating life or attaining self-respect and the respect of others’ (Sen, 1999a: 31). These dimensions are intrinsically important in themselves, not simply as means of increasing income. Furthermore, individuals may experience different dimensions of poverty, so that it becomes impossible to identify the poor as a single category or summarise their experience in a single, all-embracing measure of poverty such as household income. Sen’s analysis has important implications for the basis of resource allocation in the public sector. Although he recognises the importance of income and human development in improving the standard of living of the poor, he suggests that other dimensions of poverty reduction such as security, social inclusion and empowerment may be equally important. This obscures the basis for prioritising between resource allocations, since it is no longer appropriate to assess services as an income transfer. If all dimensions of poverty are intrinsically important, how should technicians choose between interventions aimed at improving household security (such as spending on law enforcement) and education? Where poverty is defined as a function of income, the answer is clear: priority should be given to those interventions that generate the greatest increase in income for the poor. Where there is no common-denominator for the various dimensions of poverty, direct comparison between their rival merits becomes impossible (see Section 2.2 and 2.3). The importance attached to empowerment and good governance also bring into question such technocratic approaches to decision-making: consultative and participatory approaches are more consistent with Sen’s poverty reduction agenda. All of these considerations point to the key role of the budget process, and its links to the political process, as the key determinants of whether or not resource allocations will actually address the poverty reduction concerns identified by the poor. incidence where public services are effectively club goods (see Section 2.1) or where the poor are not given the same treatment as higher-income groups. The cost of services is an inadequate proxy for the benefits received anyway since it fails to take into account the ability of different social groups to transform access to services into improvements in the quality of life as measured, for instance, by increased income. Furthermore, government expenditure only represents the gross transfer to households, since use of the services may imply costs such service charges – official and unofficial – travel and the opportunity cost of time lost to productive activities, all of which may impact differentially on the net transfer actually received. A similar criticism may be advanced regarding the scope of analysis. There is a tendency for studies to focus on the distributional impact of spending on particular services or transfers. This fails to capture the true impact of increased spending, since this will have to be financed either through increased taxation or an increase in the deficit. Both these financing options will have differential impact on social groups, so that the net benefits for some may be reduced. Despite these limitations, the method does provide a good indication of who benefits from spending. Additional information is needed to assess the distribution of these beneficiaries: to determine whether they are concentrated in urban areas or particular regions of the country. Additional information is also needed to understand why particular social groups get more or less than their fair share of benefits. The behavioural response to public expenditure This is perhaps the fundamental weakness of benefit-incidence analysis: the method describes but does not explain the social distribution of benefits. To do so requires an understanding of the behavioural response to public expenditures and service delivery. Insights on the behavioural response can be gained by using modelling techniques. A model of the determinants of labour supply in Sri Lanka, for instance, suggests that households receiving food subsidies will reduce the amount of contracted work they undertake, so that the net-gain in income is less than the value of the subsidy (Sahn and Alderman, 1995). Similarly, a study of interhousehold transfers in Peru suggests that state pensions reduce the amount of transfers from the young to the old, again reducing the net gain to income from a transfer mechanism (Cox and Jimenez, 1992). A similar approach can be used to assess the behavioural responses to changes in service provision, such as the impact of user fees on health seeking behaviour (Gertler et al, 1987). Although behavioural studies can provide a useful insight into the impact of public spending, including predictive models, they are much more complex and demanding in terms of data, and, consequently, are still not widely used. For practitioners, the information gathered through service delivery surveys and other methods of client consultation will generally suffice. Such consultations are cheap to implement and provide some insight into the motivations behind the service users and the factors influencing demand, such as the quality of service. Although the survey results may not be technically robust, providing an inadequate basis for prediction of behaviour in response to policy changes, they can provide useful guidance to policy makers regarding spending priorities and delivery mechanisms (see Section 3.4). Mechanisms of redistribution and targeting There is some evidence to suggest that redistributive policies tend to have the greatest impact on poverty where they have entailed the redistribution of assets (WDR 2000). This may include radical measures, such as land reform, or, more often, policies intended to generate marketable assets amongst the poor, such as investments in education, health care and appropriate agricultural technologies. There may also be strong political motives for the preference for public investments in assets rather than transfers of income as a means of reducing poverty and inequality. Direct cash transfers are likely to be unpopular amongst the higher income groups, who are required to finance redistribution through taxation. Higher income groups may also have strong views regarding the valuation of alternative applications of redistributed income, so that certain expenditures – such as education and health care – may be seen as merit goods and as such may be preferred to expenditure choices made by the poor themselves (see Section 3.4). Cash transfers are also likely to be difficult to administer, particularly where poor individuals are difficult to identify. One of the key considerations in the design of redistributive policies is the targeting mechanism to be employed. Public interventions can be either broadly or narrowly targeted. Access to broadly targeted interventions is universal: individuals may benefit irrespective of the income group to which they belong. Examples include food price subsidies or free public schooling. Narrowly targeted interventions, on the other hand, seek to exclude the non-poor. This may be achieved by, for instance, means testing potential beneficiaries or providing benefits and participation requirements that are unattractive to higher income groups and so effectively self targeting, such as subsidies on low status foodstuffs or a work requirement for transfers (van der Walle and Nead, 1995). The choice between these strategies is usually determined on the basis of their cost-effectiveness in reaching the intended beneficiaries, reflecting a trade-off between the unit cost of administration and errors of targeting. Universal programmes commonly suffer from errors of inclusion, whereby resources intended for the poor are delivered to higher income groups. In principle, the replacement of universal interventions with those that are narrowly targeted will reduce leakage to the non-poor and so improve cost-effectiveness. However, these efficiency gains must be set against the increased cost of some narrowly targeted programmes and participation costs for the poor – as in a food or cash for work scheme, or infrastructure programmes requiring labour participation – which will reduce the net transfer to the intended beneficiaries (Ravallion and Datt, 1995). Narrowly targeted programmes, on the other hand, suffer from errors of exclusion, whereby the intended beneficiaries are unable to benefit, owing to difficulties in substantiating claims or participation costs. Costs of exclusion are rarely considered in the design of targeted interventions, which generally seek to reduce leakage and, thereby, reduce the total cost to public sector regardless, of broader, social efficiency concerns (Cornia and Stewart, 1993). In the case of public services, distributional concerns will influence the share of public expenditure in meeting the costs of service provision and the way in which these subsidies are administered. Where higher income groups benefit disproportionately from a particular service there is a strong case for public subsidies to be reduced and a substantial proportion of the costs recovered from user fees. There is, of course, a risk that the application of such charges would be regressive and discourage the poor from using services, as suggested by the significant increase in school enrolment when primary school fees were abolished in Uganda. One solution lies in moving from supply-side subsidies (which cover the cost of service provision irrespective of the beneficiary) to demand-side subsidies (which can be targeted to specific service users), such as the graduated remission of user fees. Where the poor cannot be identified administratively, such schemes are best administered locally, by, for example, allowing communities or the mangers of public services to identify the families which would benefit from the fee remission or study grants. Unfortunately, these targetted subsidies have high transactions costs and are difficult to monitor. As a result they are open to abuse: often the higher-incomes groups end up benefiting anyway (Gilson, 1998). purposes of redistribution and how these resources should be distributed within society. Unfortunately, economic theory provides little guidance on these questions since there is no consensus regarding the optimum or just distribution of income or assets (Cullis and Jones, 1998: 217-223). These issues have to be resolved in the political sphere. This requires politicians to judge the extent to which taxpayers are prepared to pay the costs of redistribution. While voters may endorse redistributive policies, particularly those that seek to reduce glaring inequalities and poverty, they will also seek to benefit from the services and transfers financed through general taxation. In satisfying these contradictory goals, there is a tendency for politicians to broaden the range of beneficiaries – for example, preferring supply-side over demand-side subsidies – so that programmes drift from narrow to broad targeting, reducing their redistributive impact (see Section 2.9). Clearly, the political process plays a crucial role in determining not just the extent of redistribution, but also the means by which this redistribution can be achieved. Conclusion Ultimately, policy makers have to determine how much public spending should be allocated for the References Barro, R., Lee, J., 2001. International data on educational attainment: updates and implications. Oxford Economic Papers 53 (3), 541–563. Blanchard, O., Tirole, J., 2004. The optimal design of unemployment insurance and employment protection a first pass. NBERWorking Papers, vol. 10443. Boeri, T., Conde-Ruiz, I.J., Galasso, V., 2003. Protection against labour market risk: employment protection or unemployment benefits? C.E.P.R Discussion Papers, C.E.P.R. Discussion Papers, vol. 3990. Botero, J., Djankov, S., La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 2004. The regulation of labor. The Quarterly Journal of Economics 119, 1339–1382. Heckman, J., Pages, C., 2000. The Cost of Job Security Regulations. NBER Working Papers, vol. 7773. Lazear, E., 1990. Job security provisions and employment. The Quarterly Journal of Economics 105, 699–726. OECD, 1994. OECD Jobs Study. OECD.