Compiled by: Donald Maphiri, Tania Ajam, Conrad Barberton

Presented by: Donald Maphiri And Conrad Barberton

We thank the Chairperson of the Select Committee on Finance for inviting AFReC to make a submission on the MTEF Figures 2001/02.

Our submission is divided into two sections: The first section deals with various aspects of the MTEF, and the second section deals with the FFC’s costed norms approach and provincial taxation. The main points of the submission are summarised at the end.

1. The MTEF for 2001/2002

It is a fact that the introduction of MTEF has been a major innovation in the budget process. As a result this process has become more transparent in recent years and has served to support expenditure planning in government more effectively. The three-year spending envelope allows departments to have greater certainty about the resources available to them and therefore are better able to plan the deployment of their resources to meet their objectives. The first three-year expenditure framework was for the 1998/99-2000/01 period. The MTEF has thus completed a full three-year cycle. It should be mentioned that the introduction of the MTEF does not necessarily guarantee improved service delivery and/or flow of information. Continual improvements on the present achievements will need to be made to keep up with changing circumstances and needs. The experience that has been gathered over the past three financial years has highlighted a number of issues that need to be taken into account in the MTEF process. These include:

• The compilation of the baseline figures and forward estimates;
• Implementing ABX-type budgeting to facilitate trade-offs as recommended by the Presidential Review Commission on Public Sector Reform;
• Differences between three-year incremental budgeting and MTEF based on cost drivers and policy adjustments; and
• The need for the MTEF scorekeeper.

These issues are dealt with below.

1. Developing baseline figures and forward estimates

It is generally accepted that baseline estimates refer to the cost of conducting existing programmes. The introduction of policy changes (namely, the introduction of new projects and programmes) are evaluated relative to the baseline. Forward estimates are provisional decisions by government about its intention regarding future spending on both existing and new programmes, as well as the phasing out of old programmes. The forward estimates are rolled over every year with the new estimate added for the last year of the relevant MTEF cycle. These estimates are, however, not binding on the government when it comes to compiling the current year’s budget. In addition the forward estimates are revised periodically to take account of policy changes, changes in economic conditions, and other new information on expenditure levels.

Internationally, the forward estimates tend to be projections based on current policy parameters. When the first forward projection for any MTEF cycle becomes the baseline for the current budget year the focus of the budgeting process is on making marginal changes to take account of new programmes or the termination of old ones. In South Africa, however, the MTEF guidelines say that proposed policy changes or shifts in priority should be accommodated within the responsible department’s medium term’s expenditure estimates. Where policy changes imply adjustment to forward estimates, they have to be made within the context of the annual medium term expenditure planning process. This implies that MTEF projections over the cycle cannot be treated as baselines for the relevant years because they reflect both existing programmes and policies as well as programmes and policies that still need to be approved.

In addition, Treasury has adopted a zero change approach to the aggregate value of the MTEF, i.e. departments have to keep within their initial allocations. So when they introduce new programmes they have to find the funding from within their existing allocation either by stopping old programmes or realising efficiency gains. In other words the South African forward estimates contain not only the cost of existing projects/policies, but also any policy changes as well. This means that policy changes are not as transparent as they should be, as they are embedded in the forward estimates. The result is that policy changes are not made in a transparent way.

In view of the above AFReC proposes that MTEF forward figures should distinguish between the costs of existing programmes and those of new programmes, and decisions to phase out programmes. This, we believe, will increase transparency with regards to decisions around trade-offs between new and old programmes, as well as improvements in the delivery of services. A useful format for presenting this information is the ABX format of presenting budget information which is discussed next.

2. Implementing ABX format for the budget to facilitate trade-offs in the MTEF

Prioritisation and reprioritisation presuppose shifting resources away from low-priority activities to high priority activities and often from existing programmes to new ones. The determination of priorities becomes precise and realistic during the preparation of the budget. It is thus important that there should be mechanisms within the budget process that encourage the re-evaluation of policies and priorities and that facilitate the generation of policy alternatives. It is precisely for this reason that AFReC supports the use of the ABX format of presenting budget information to facilitate decision making around tradeoffs. This approach was initially advocated by the Presidential Review Commission on Public Sector Reform. It should be noted that ABX is not a system of budgeting in the same sense that performance budgeting is a system of budgeting. Rather it is a way of presenting budget information to facilitate the making of tradeoffs.

The ABX budgeting consists of three components:

1. The A-budget. This component specifies the baseline or the starting point including the costs of existing projects and programmes. This presupposes that the actual costs of conducting the current service delivery are known as well as the rate at which they will increases in future years based on the evolution of cost drivers. The most common question in constructing the baseline has been whether the baseline should be based on current levels of operational efficiency or on best practice level of efficiency. Standard practice, however, has been the calculation of baseline based on the current level of efficiency except for cases where improvements have been made that will certainly yield improved efficiency, in which case the baseline should reflect these developments.
2. The B-budget. This is the second component and should reflect all proposals for new services, expanded programmes and enhanced levels of service delivery. This covers all proposed policy that will require increased expenditure, including improved norms and standards, new policies, new or expanded programmes, and legislation with financial implications. The aim here is to subject all of the above proposals to budget decision making before final approval thus avoiding a situation where government decides on policy without determining how to pay for it. The analysis of these proposals would determine whether they are compatible with the government's medium term priorities and with the resources available for new initiatives. Since the costing of all new programmes cannot be accommodated all at once in the budget preparation process it is important to build an 'evaluation time' into the budget process. This would allow Cabinet to evaluate the allocation of resources on an ongoing basis, and to plan the orderly phasing-in of new programmes.
3. The X-budget. The third component of ABX budgeting is the compilation of the potential or planned reductions in expenditures whether they be from administrative savings or programme reductions or the discontinuation of programmes. The savings for these can be utilised for funding new initiatives or to make up for shortfalls between the resource requirements of existing programmes (as determined in the A-budget) and available resources.

Note that with the information presented in these three ‘budgets’ it is possible to make rational tradeoffs between existing and new programmes and to plan both the phasing out and the phasing in of programmes over time. Once a new programme from the B-budget is approved by Cabinet (and where relevant parliament) it would be shifted to the A-budget and phased in as part of the existing set of policies the department has to budget for and implement. Similarly when programmes are phased out they would be gradually removed from the A-budget.

1. Difference between three-year incremental budgeting and MTEF based on cost drivers and policy adjustments

There is a fundamental difference between three-year incremental budgeting and an MTEF based on cost drivers and policy adjustments. The former is not different from conventional incremental budgeting where current budgets are increased by some margin for the following year. The introduction of the MTEF was an attempt to move away from this kind of budgeting to a budgeting system based on the actual cost of service delivery. The latter system takes cost drivers and policy adjustments into account when making forecasts over the MTEF period. Figures based on these factors should be more realistic and support effective service delivery.

Our view therefore is that unless departments identify comprehensive and relevant cost drivers during budgeting and quantify their exact level of costs, there is a risk of budgeting in South Africa to boiling down to incremental budgeting with a three-year horizon. Therefore the problems associated with incremental budgeting will not be resolved. Ideally, the MTEF should be preceded by a move to performance budgeting so as to lay a credible foundation for fiscal resource allocation. The following prerequisites are therefore indispensable:

• A credible baseline budget for each department and/or agency based on current programmes;
• A co-ordinated policy package to ensure that actual spending patterns are within an acceptable margin of the budgeted estimates
• An effective implementation structure that ensures the MTEF approach establishes and gains increasing credibility
• Linking the MTEF to service delivery to avoid it boiling down to a financial numbers game.

As noted above key to an effective MTEF is the implementation of a performance budget. This would put in place a firm foundation for the projection of costs and outputs. In addition both the ABX format of presenting budgets idea noted above and the ‘scorekeeper’ idea noted below would contribute to the effective use of the MTEF in public decision making.

1. Introducing the MTEF scorekeeper

In order to ensure the integrity of the MTEF a ‘MTEF scorekeeper’ is needed to maintain the MTEF baseline and to record the impact of new (approved) policies on the MTEF. Currently, departments maintain their own MTEF numbers and the national and provincial treasuries review and compile these into consolidated MTEF tables. It is our view that the treasuries should play a far more active role in maintaining the integrity of their respective MTEFs by acting as ‘scorekeepers’. Very briefly we envisage that the role of the scorekeeper would entail:

• Maintaining a single MTEF database for the province or national government as the case may be;
• Adjusting the baseline numbers for demographic and certain economic variables, such as inflation;
• Maintaining a data-base of current policy proposals and the extent of their expected budgetary impact, as well as any cost-saving measures (i.e. maintain an ABX budget for government);
• Ensuring that departments cost and integrate any new (approved) policies into the MTEF baseline.
• Ensuring that departments have plans to ensure that approved policies are financed by cost savings or by cutting existing policies or can be paid for by increasing revenues.

The idea is that the scorekeeper ‘referee’ the MTEF process to ensure that departments play by the rules and ‘keeps’ score of the current state of the MTEF. The idea is not for the scorekeeper to do the MTEF thinking for the departments. Rather the departments are the ones that instruct the scorekeeper when to make changes, and what changes to make. The scorekeeper would check that any adjustments to the MTEF proposed by a department comply with the rules for making such adjustments and then record the results of the adjustments. The idea is to develop a MTEF with integrity.

1. The costed norms approach and provincial revenue raising powers

This section discusses certain aspects of the costed norms approach to the division of nationally collected revenue proposed by the FFC, as well as the issue of provincial revenue raising powers.

1. FFC's proposed costed norms approach

In a perfect world, of perfect information and zero research costs the FFC’s costed norms approach would undoubtedly be a sounder instrument of resource allocation across spheres of government than the current approach used by Treasury. In a second best world where information is not available and the cost of research to obtain such information is very high, the FFC has to make use of all sorts of proxies to measure what it really wants. In such a world, it is by no means certain that the FFC’s proposed costed norms formula is a better intergovernmental resource allocation instrument than the Treasury formula. In the ideal world the FFC’s formula based on a costed norms approach would be irreconcilable with the Treasury’s approach. In the second best world, it is probably possible (and may be desirable) to include some of the FFC’s recommendations into the Treasury’s formula, thus creating a hybrid formula where certain components are based on costed norms and others based on the proportional division of revenue approach used by Treasury.

For instance. AFReC has consistently argued that for social security it would make more sense to allocate funds based on actual cost of grants, rather than the proportional demand for such grants as is currently the practice. It should be possible to estimate the demand for social grants in each province with a high degree of accuracy and therefore to calculate the exact amount of nationally collected revenues each province requires in order to meet this statutory obligation. This approach would obviate the current problem where increases in social grants put pressure on other expenditures especially social welfare services.

Overtime the ‘proportional components’ of such a hybrid formula could be converted to take account of costed norms as explicit norms and standards for other social services are set by national government and explicit baseline unit costs are established with the implementation of performance budgeting.

It is our view that the FFC’s costed norms approach cannot be implemented in education, health and social services in the short term for a number of reasons:

• National norms and standards have not been defined in these areas – for instance what package of medical treatments constitute minimum secondary health care standards etc – this is based on medical protocols etc
• The required information on costs and uptake of services are not available – this problem is acknowledged both by the FFC and the Treasury
• Political will – International experience has shown that it is not only for information reasons that binding minimum norms are not defined. Establishing explicit minimum norms and standards for service delivery is often difficult politically and so governments prefer to leave such standards undefined.

In essence the FFC is trying to overcome the proportional split approach of the Treasury formula where the relative allocation may or may not accord with the actual cost of service delivery. (As noted above the tension between relative share and actual cost is probably most acute in the area of social welfare grants where the norms are well defined and entitlement is set at national level.). But to the extent that the FFC’s cost projections are not accurate, their costed norms approach will not necessarily overcome this problem. What the FFC’s current proposals do is that they set a figure for say – average unweighted cost per learner in a province; then they weight it by factors which are likely to impact on the relative cost off service provision to arrive at an adjusted cost factor – eg (special versus ordinary learners, rural vs urban etc). But how accurate are the initial cost estimates, how accurate are the adjustments? Until we have a budgeting system that can deliver on these questions we simply don’t know! In addition the FFC does not present any sensitivity analysis in its final document so we can gauge nothing about the possible variances and their impacts on costing.

The FFC is trying to estimate a production function (which is essentially the ratio of inputs to outputs). But outputs are not well defined – what is the cost of a "basic education" - is it taken as years spent in a classroom or in terms of educational performance. So what are we costing the average cost of? Secondly the FFC makes the point that parameters should be policy-invariant i.e. out of the control of provincial governments. But this is assuming there is some notional least cost production function on which the FFC can base its average cost information. But varying the input mix is within the control of provinces. And by using historic cost data as the basis for the projection of average costs for the "normative" production function, surely violates the objective of policy invariance. In fact if the (actual) average cost is used in this way, how is it different from Treasury formula which uses past expenditures to determine the weightings given to each are of expenditure?

This is especially true since the FFC are using proxies like learner type rather than actual cost. How accurate is the production function when using proxies as instruments? Is it sufficient that the factors modifying the "average cost" should be policy invariant, or should the "average cost" be policy invariant as well (which would be the case given a normative least cost benchmark production function)?

The weightings on the current formula for dividing revenues across spheres reflect provinces past expenditures on the social services. As far as can be determined, in 1998 and 1999 they were based on some weighted average of province's past expenditures and MTEF projections. This year it is stated explicitly that they "reflect a 3-year average expenditure on these services" (Explanatory Memorandum, page 11). The use of expenditures to calculate the weightings is problematic because it amounts to nothing more than a sophisticated form of incremental budgeting. It is also well known that the baseline for these expenditure patterns was primarily determined by apartheid spending priorities. Moving towards costed norms would be a step in the right direction, but we need to be certain that the average costs we use do not simply reflect past expenditures in a different guise yet again.

To conclude this brief discussion: the most important consideration when devising a method for dividing nationally collected revenues between the different spheres of government and then dividing the provincial equitable share between the provinces is that the outcome of the process must be generally regarded as "substantially fair". It is simply not possible to devise an objective method for ensuring that the division is 100% fair. Roleplayer’s very definition of what is fair differs depending on their interests and who’s interests they are representing. Moving towards a costed norms approach is likely to foster a greater sense of "fairness", although it runs the risk of becoming extremely complicated. This brings us to the second most important consideration, namely whatever method is used to make the division of revenue it should be both simple and understandable. This inevitably means making certain generalisation about the demand for services in each of the provinces and the cost of service delivery. To seek to incorporate every factor that either increases or decreases the demand for or the cost of services in each of the provinces would result a very complicated, technical formula that very few people would be able to understand. There would always be pressure to make the formula even more comprehensive and therefore even more complicated. The result would be technocrats would take control of he formula, rather than it being an instrument of policy.

1. Provincial revenue raising powers

The structure of South Africa’s fiscal constitution assigns most of the buoyant taxes such as corporate income tax, personal income tax and VAT to national government. This means that – unless the Constitution is fundamentally amended – provincial governments will in the foreseeable future remain heavily dependent on intergovernmental transfers and a revenue sharing process in which nationally collected revenue is allocated across and within the various spheres of government. However, there is also a clear indication in section 228 of the Constitution that provincial legislatures should have at least some latitude in raising their own taxes and flat rate surcharges on national tax bases. In addition, there is the cogent argument of the FFC that provincial tax powers could enhance accountability at the margin.

The Constitution requires that provincial taxation be regulated by national legislation, and already there is movement towards an "allowed list" – a menu of tax options which would be open for provincial governments to pursue. In crafting the enabling framework for provincial taxation, AFReC offers the following considerations:

1. A long term vision: Provinces should have a clear idea of the maximum limits of their taxing powers. Even though many of the options may be difficult to implement at present (such as problems with abatements in implementing a surcharge on the national income tax base), these may in future become viable options as factors such as information technology and administrative efficiency improve.
2. Short term degrees of freedom: Provincial legislatures should have a clear idea of what options are available to them in principle (in terms of legal powers and the long term vision) and in practice (in terms of their current capacity to administer such taxes).
3. A development trajectory based on graduated criteria: Before being granted addition revenue raising options, all provinces should comply with a set of objective criteria (e.g. capacity to administer the proposed tax, utilisation of current revenue sources etc). The same criteria would be applied to all the provinces equally, but only those provinces comply with specific criteria would be allowed to "graduate" to exercising increased powers. Thus all provinces would be treated equally in that the same conditions apply to all of them, but they would differ to the extent that they develop the necessary capacity to take on more taxing powers at differential rates.
4. Systemic coherence: It must be borne in mind that economic theory suggests that the choice of tax powers would also determine what sort of borrowing powers provincial governments have. Provincial borrowing powers in the absence of fairly significant own revenue raising powers could exacerbate moral hazard problems in capital markets. Provincial borrowing powers legislation must be congruent with legislation on tax powers.

1. Summary of main points

In summary this submission makes the following points:

1. The MTEF figures should include information on costs of existing projects and those associated with policy changes or introduction of new programmes, as well as the phasing out of programmes.
2. Departments should compile detailed information on cost drivers and base their MTEF figures on those and policy changes to avoid reverting to incremental budgeting with a three-year horizon.
3. To improve prioritisation and trade-off analysis within the context of the MTEF the ABX method of presenting budget information and new policy options should be adopted government-wide.
4. The national and provincial treasuries should act as MTEF ‘scorekeepers’.
5. The FFC's costed norms approach should only be implemented in areas where the necessary information is available, for instance in social security, i.e. the Treasury formula should be adjusted to take on board the costed norms approach in components where sufficient information is available.
6. The government needs to draw up very specific minimum norms and standards in health, education and social services so that the costed norms approach can be applied in these areas.
7. The any revenue sharing formula should remain simple and transparent so that it dose not devolve into a technical exercise but remains a useful policy instrument.
8. Provinces should be well-informed about the maximum acceptable level of taxing capacity irrespective of the present problems.
9. Provincial legislatures need to be informed about heir powers to administer such taxes.
10. Before provinces are granted increased taxing powers there needs to be an objective graduation criteria governing the graduation of all provinces to certain taxing powers.
11. Provincial borrowing powers need to be consistent with taxing powers to avoid financing provincial borrowing costs from equitable shares.