the cyclically adjusted budget balance
5.1. Approaches to estimating structural balances
In an ideal world with sufficient information on all bud- getary items and discretionary policy measures, it would
be possible to adjust each budget item directly to reflect their ‘true’ structural position. In practice, information of such quality is not available. Consequently, indirect meth- ods are used where the cyclical budgetary component is inferred from the co-variation of government revenues and expenditures with output fluctuations.
In estimating cyclically adjusted balances, the Commis- sion uses the same method for all countries. This method is described in detail in Chapter 1 of Part VI. The cycli- cal budget component is inferred from the estimated cyclical position of the economy and estimates of the budget sensitivity parameters. Whereas there is a broad agreement on the magnitude of the estimated budgetary sensitivities, there is less agreement on the best approach for estimating potential output and output gaps. The issues of how to estimate the output gap and budget sen- sitivities are now addressed in turn.
Measuring the output gap
The output gap plays an important role in the calculation of cyclically adjusted budget balances. To compute the output gap, an estimate of the potential or trend output is needed. Because these are not directly observable, a number of assumptions must be made to disentangle trend and cycle in actual GDP developments.
The Commission services have so far used the so-called Hodrick-Prescott filter (hereafter, HP) to estimate trend output and related output gaps. This method is simple to use and the results can be easily reproduced. It also min- imises the need for judgmental interventions and so allows a consistent treatment of Member States. Moreover, the method is parsimonious on data requirements and the cal- culations can be made on the basis of information pro- vided in the stability and convergence programmes. The HP filter has also the advantage that the estimated output gaps, and hence the cyclical components of the budget balance, cancel out over the cycle. A clear disadvantage of the HP filter, however, is its lack of economic foun- 63 (1) Opinion of the Monetary Committee on the content and format of
stability and convergence programmes, document MC/II/482 final of 16 September 1998, endorsed by the Ecofin Council of 12 October 1998. This document was published as an annex to the conclusions of the Council.
dations which makes its results and underlying assump- tions difficult to interpret economically.
Another commonly used method to compute potential output and output gaps involves estimating a production function. Under the production function approach, potential output estimates, beings based on theoretical grounds, can be interpreted from an economic stand- point. This allows the assessment of the underlying factors driving the results. On the other hand, the results depend strongly on assumptions on the functional form of the production technology, e.g. returns to scale, the trend growth of technical progress as well as on estimates of the structural unemployment rate. All these assump- tions are subject to heated economic debate.
The Commission services use the production function from the QUEST model to check the robustness of the results obtained by the HP filter. Output gaps generated by the QUEST production function approach do not dif- fer strongly from the output gap calculations based on the HP filter basically because the HP filter is used to estimate the structural unemployment. The correlation between the output gaps produced with the two methods are in the range between 0.8 and 0.9 for all EU countries. Such high correlation can also be observed when compar- ing HP filter gaps to the production function based gaps made by the OECD and the IMF, even if their specifica- tion is different from the QUEST production function. While high correlations indicate that the estimated out- put gap changes produced by various methods are quite similar, in level terms the output gap estimates may still show considerable variation.
These differences are particularly evident in calculating potential output during the disinflation period of the 1980s and the 1990s. A production function approach usually establishes a direct link between growth and inflation by incorporating a Phillips curve capturing the short-term trade-off between output and inflation. The output loss needed to bring down inflation is considered to have little effect on the production potential of the economy. Hence, large negative output gaps emerge. A statistical filter, instead ‘interprets’ part of the subdued growth performance as a fall in potential output. Hence, the estimated output gaps are smaller.
This implies that unlike the HP filter trend method, a method based on production function does not necessar- ily produce symmetrical output gaps. In the context of budgetary surveillance, the question is how much of these accumulated negative output gaps will in the end
be retrieved. If the accumulated output loss is not going to be recovered, the corresponding revenue will actually never materialise. However, with a more stable macro- economic environment, the importance of this issue should be reduced in the future. The stability-oriented macroeconomic framework of EMU would itself con- tribute to smooth out these differences.
Budget sensitivity parameters
The budgetary sensitivity parameters used by the Commission services are based on tax and expenditure elasticities which have been recently calculated by the OECD (OECD, 1999a). The different tax elasticities (indirect taxes, personal income taxes, corporate taxes and social security contributions) are weighted using their relative shares in overall tax income over the 1985–99 period. The government expenditure sensitivity parameter refers to unemployment related expenditures. This approach implies that the budget sensitivity para- meters express the average cyclical response of the bud- get to the cycle and not the response in any one year (1).
Generally this poses no problem when considering bud- getary trends in the longer term, but it can complicate the interpretation of results for an individual year for sever- al reasons. Firstly, tax reforms at different points in time are not reflected in changes of the budget sensitivity parameters. Specifically, tax reforms after 1996 will only be reflected when the tax elasticities will be updated the next time. Secondly, as shown in the next section, in individual years different tax bases may react atypically to the cycle. A third issue is the lag in the collection of taxes, especially when there are abrupt cyclical swings; uncertainty concerns particularly corporate taxes, which are not only quite volatile but also collected with sub- stantial lags.
An additional issue is the coverage of the budgetary items included in the measurement of the cyclical com- ponent. Other expenditure items besides unemployment benefits — for instance, social and health care expendi- ture — may fluctuate with the cycle. However, it has proven empirically difficult to find a consistent pattern. A related issue is how to deal with the different bud- getary rules on expenditures and revenues that have been introduced in several Member States in the last few years. For example, the Dutch budget system includes
64
(1) In particular, on the revenue side the OECD tax elasticities are
based on the 1996 tax codes.
specific budgetary rules which partially offset the bud- getary impact of the automatic stabilisers, making it dif- ficult to distinguish between automatic and discretionary changes.
More generally, it may be argued that governments tend to react systematically to the cycle and that this ‘govern- ment reaction function’ should be taken on board. A number of studies, as well as evidence provided in other parts of this report, have found that ‘quasi-automatic’ behaviour linked to systematic reactions by budgetary authorities is not symmetric over the cycle and, in gen- eral, tends to be pro-cyclical. Hence, ‘actual’ stabilisers may be smaller than ‘purely automatic’ stabilisers. However, the SGP is largely about breaking this tradi- tional behaviour in order to let the automatic stabilisers play freely over the cycle. Thus, to the extent that the Pact’s monitoring system works effectively, this kind of policy reaction should not characterise the future bud- getary behaviour.
Comparing cyclically adjusted budget balances
The Commission, the OECD and the IMF all use the same overall framework to compute cyclically adjusted figures, that is, applying a set of budgetary sensitivity parameters to an estimated output gap. As signalled above, the differences that appear are mainly due to dif-
ferent methodologies in the estimation of the output gaps where the OECD and the IMF use a production function while the Commission services use a trend estimation method.
Table 18 below shows figures on cyclically adjusted bud- get balances computed by the Commission, the OECD and the IMF for all EU Member States for the 1997–99 period in their respective, most recent forecasts (1). For
example, for 1997 the IMF indicates that the structural deficit in the euro area was 1.3 % of GDP while the Commission has a figure of 2.0 % of GDP, a substantial difference. On average, differences in levels are gradually reduced in 1998 and 1999, even if they remain sizeable for some countries, especially for the large economies.