This Selected Issues paper for Ireland highlights that fiscal consolidation resulted in a tremendous reduction in public debt from nearly 100 percent of GDP in 1991 to about 30 percent in 2004. This has reflected a combination of policy decisions and economic circumstances. Excluding 2001, when the economy has been affected by the global economic slowdown, Ireland has in general consistently enjoyed favorable surprises in its public finances. Indeed, during this period, the actual fiscal outturns have exceeded budget forecasts on average by 0.3 percent of GDP a year.

Abstract

This Selected Issues paper for Ireland highlights that fiscal consolidation resulted in a tremendous reduction in public debt from nearly 100 percent of GDP in 1991 to about 30 percent in 2004. This has reflected a combination of policy decisions and economic circumstances. Excluding 2001, when the economy has been affected by the global economic slowdown, Ireland has in general consistently enjoyed favorable surprises in its public finances. Indeed, during this period, the actual fiscal outturns have exceeded budget forecasts on average by 0.3 percent of GDP a year.

I. Favorable Fiscal Outturns: Is it Just the Luck of the Irish?1

A. Introduction

1. Since the early 1990s, Ireland’s fiscal position has improved significantly. Fiscal consolidation resulted in a tremendous reduction in public debt from nearly 100 percent of GDP in 1991 to about 30 percent in 2004. This has reflected a combination of policy decisions and economic circumstances. Excluding 2001 when the economy was affected by the global economic slowdown, Ireland has in general consistently enjoyed favorable surprises in its public finances. Indeed, during this period, the actual fiscal outturns have exceeded budget forecasts on average by 0.3 percent of GDP a year. Has this been a reflection of prudence in budget forecasts or is it just luck?

uA01fig01

Forecast Error: Fiscal Balance

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

2. The objective of this paper is threefold. First, it compares Ireland’s budget forecasting record with that of other industrialized countries. This comparison makes it possible to assess to what extent the favorable fiscal outturn was Ireland-specific. Second, the paper gauges the main factors affecting the Irish budget forecasts taking into account the institutional environment governing fiscal policy. Third, it discusses some implications for public finances going forward in the context of possibly lower potential growth and significant pressure to raise public spending in Ireland.

3. The analysis presented here suggests that a sustained sequence of stronger-than expected growth and buoyant asset price developments were key contributing factors to the favorable budget surprises in Ireland. Budget macroeconomic projections did not differ much from outside forecasts, including the ESRI, a non-governmental think-tank in Ireland, and World Economic Outlook (WEO). Prudence in preparing budget estimates also played some role. Budget estimates have been based on cautious assumptions regarding tax elasticities. Thus, a combination of stronger growth and a prudent approach employed by the budget in forecasting revenues produced consistent revenue overperformance during this period. However, revenue overperformance was in large part offset by increases in expenditure. As a result, despite the larger upside surprises on growth experienced by Ireland compared to other industrialized countries, the forecast error for the fiscal balance was not significantly large in Ireland’s case.

B. Methodology

4. The paper uses the methodology employed by Muhleisen et al. (2005) to assess Ireland’s budget forecasts. The benchmark group used for comparison purposes consists of 11 countries: the United States, Germany, United Kingdom, France, Italy, Canada, Australia, New Zealand, the Netherlands, Sweden, and Switzerland. Following Muhleisen et al. (2005), the paper compares budget forecasts against the actual outturns reported in the budget two years later to allow for revisions.2 For example, when budgets for Ireland are published in December, the actual outturn for the previous year is still preliminary and subject to considerable changes. The advantage of using the reported data in the subsequent budget over actuals is that it allows a more unbiased assessment of the budget forecast errors based on available information at the time of the preparation of the budget.

5. There are serious data limitations in conducting cross-country comparisons of fiscal forecasts. A relatively short and uneven sample period across the benchmark group makes it difficult to draw statistically significant conclusions. In addition, the forecast errors do not take into account any policy decisions that were not envisaged at the time of the budget, but were implemented during the course of a particular year.

6. Forecast errors are defined as the difference between the reported actuals and budget projections. A positive (negative) value implies the outcome has exceeded (underperformed) budget expectations. The extent of forecast accuracy is assessed on the basis of a simple average of forecast errors (the mean error—ME) and root mean squared error (RMSE), which indicates the magnitude and variance of the errors independent of the direction of forecast errors. A standard set of statistical tests is applied to assess the presence of forecast bias and efficiency; the latter to assess whether forecasts were based on all information available at the time of budget preparation.

C. International Comparison of Forecast Accuracy

Fiscal projections

7. During 1995–2003, Ireland’s fiscal balance was on average 0.4 percent of GNP better than projected.3 While there was substantial overperformance, Ireland was not the only country that enjoyed such stronger fiscal outturns. Indeed, a number of countries in the benchmark group experienced similar levels of overperformance on their budget balance. Compared to the benchmark group, Ireland’s mean error was slightly above the group average; and its forecast accuracy, measured by the RMSE, was slightly weaker than the average.

uA01fig02

Fiscal Balance: Mean Error

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

uA01fig03

Fiscal Balance: RMSEs

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

8. Overperformance on fiscal balance was mainly driven by higher-than-expected revenue. The large majority of the benchmark group benefited from revenue overperformance. The revenue forecast error was particularly large for Ireland, followed by Canada and New Zealand. Excluding the economic downturn in 2001-02, revenues were consistently higher than budgeted in Ireland, by about 0.8 percentage point of GNP. The forecast accuracy of Ireland’s revenue projections was relatively weak, reflecting larger standard deviation compared to other countries.

uA01fig04

Revenue Forecast Error

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

uA01fig05

Expenditure Forecast Error

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

9. In contrast, expenditure in Ireland has consistently exceeded budget estimates despite smaller-than-expected interest payments. Interest on government debt in Ireland was on average over 0.1 percent of GNP lower than projected, by far the largest deviation among the benchmark countries. This notwithstanding, expenditure tended to overshoot the budget projections by an average of 0.3 percentage point of GNP. In general, forecast accuracy was higher in expenditure than revenue.

10. A striking feature of Ireland’s fiscal forecast errors was a high correlation between revenue and expenditure errors. A decomposition of the forecast errors for the fiscal balance indicates that, in most of the benchmark countries, higher-than-expected revenue was accompanied by lower-than-expected expenditure. This can be explained in part by lower-than-budgeted welfare payments such as unemployment benefits on the back of stronger economic activity. Two exceptions were Ireland and Australia where revenue and expenditure surprises tended to be positively correlated.

uA01fig06

Decomposition of Fiscal Forecast Errors

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

Macroeconomic assumptions

11. The accuracy of macroeconomic projections underpinning budgetary forecasts plays a key role in explaining fiscal forecast errors. Based on budget forecasts and actuals in the United States over the 1980s and 1990s, Auerbach (1994) shows that macroeconomic forecast errors can account for a large part of the deterioration in fiscal position during that period. More recently, evaluating the performance of budget and growth forecasts in convergence and stability programs across euro-zone countries, Strauch (2004) concludes that the cyclical position and the form of fiscal governance have been key determinants of forecast biases. Jonung and Larch (2004) show that the presence of overly optimistic forecast bias in macroeconomic projections in several euro-area countries have affected fiscal forecasts.

12. A comparison of forecast errors among the benchmark group shows that there were significant errors in macroeconomic projections but with notable differences across countries. Within the benchmark group, forecast errors for growth in Ireland were consistently one-sided, with large upward surprises during the sample period. The only exception was lower-than-expected growth in 2001; however, most countries in the benchmark group failed to predict the effects of the 2001 global economic downturn. Canada and the United States experienced a similar pattern of more favorable outturns for growth, but only in the second half of the 1990s. Most of the other countries experienced forecast errors in both directions.

13. During 1995–2003, output growth (GNP) in Ireland was on average nearly 1 percentage point higher than budget projections. At the other end of the spectrum, Germany recorded lower-than-expected growth on average of 1 percentage point. Another salient feature of the Irish forecast errors was the high RMSE, reflecting not only a large mean error but also a higher standard deviation than the other countries in the benchmark group.4

uA01fig07

Real GDP Growth: Mean Error

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

uA01fig08

Real GDP Growth: Forecast Accuracy

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

14. An important part of revenue overperformance can be explained by higher-than expected growth. A simple pooled regression of the forecast errors for revenue projections on the errors for the output projections suggests that generally one percentage point higher output growth than budgeted would lead to higher revenues by about 1½–2 percentage points. However, the size of the estimated elasticity may be biased upward because systematic biases other than those in the macroeconomic variables may be embedded in the revenue projections.

15. Consistent with the favorable outturns in growth, labor market conditions in Ireland turned out stronger than envisaged, and inflation (as measured by either GDP or GNP deflator) was higher. Actual unemployment rates in Ireland were on average 0.4 percentage points lower than projected, and inflation 0.6 percentage points higher. Forecast errors for both variables show similar characteristics to the growth forecast error, with large mean errors and standard deviations. The result may reflect in part high volatility in these key Irish macroeconomic variables due to the large role played by external demand as a source of growth. Commodity-exporting countries, such as Australia and New Zealand, also have relatively higher standard deviations in their forecast errors; however, these countries have relatively small mean errors on average in comparison to Ireland.

D. Characteristics of the Forecast Errors in Ireland

16. Ireland’s large and consistently one-sided forecast error for output growth was primarily driven by upward surprises in external demand. Indeed, the forecast errors for domestic demand, both for private consumption and fixed capital formation, were two-sided and their mean errors were relatively small, especially for private consumption (Table 1 and Figure 1). On the other hand, export growth (goods and services) recorded mostly large, one-sided forecast errors. During 1991–2003, export growth was on average 3½ percentage points higher than budget projections. The Irish economy was also subject to strong fluctuations, as indicated by the large RMSEs in the forecast errors for investment, imports, and exports. Standard statistical tests confirm forecasting biases in Ireland’s macroeconomic projections (Table 2). As expected, the forecasts for output growth and the unemployment rate suggest the presence of a statistically significant bias. Among the components of output, forecast errors for exports exhibit bias in both median and mean tests.

Table 1.

Ireland: Descriptive Statistics of Macroeconomic Forecast Errors, 1991–2003 1/

article image
Sources: Fund staff calculations. See Appendix for a description of descriptive statistics.

For each variable, columns lis mean error (ME), mean absolute error (MAE), standard deviation (SD), and root mean squared error (RMSE).

Percent deviation in level.

Average for 1997–2003.

Figure 1.
Figure 1.

Ireland: Forecast Errors for Macroeconomic Projections

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

Sources: Fund staff calculations.
Table 2.

Ireland: Macroeconomic Projections: Results of Bias and Efficienty Tests 1/

article image
Sources: Fund staff calculations. See Appendix for a description of the methods.

Marks indicate tests that reject a zero median or mean at the 10 percent level.

17. Given the large uncertainties in predicting the strength of external demand, the Irish record of under predicting export growth may indicate that the budgets have relied on prudent macroeconomic assumptions, but other macroeconomic forecasts for Ireland show similar results. A comparison of budget macroeconomic forecasts against those by outside forecasters (ESRI and the WEO), available around the same time that budgets were released, show that errors in these forecasts were equally large with similar biases and large RMSEs (Table 3).5 They also under predicted output growth (GDP) by about 2¼ percentage points; in fact, their deviation was slightly larger than that of the budget forecasts. Similarly, while their forecast errors for private consumption growth were small, they consistently under predicted export growth. Overall, budget forecasts are in line with the other forecasts. A more formal test to examine the presence of statistical dominance of the budget forecasts over the others confirms that there is no clear evidence of such dominance. Regressions of the actual values of macroeconomic variables on both budget and outside forecasts, following the technique employed by Fair and Shiller (1990), show that neither set of forecasts statistically encompass the other.

Table 3.

Ireland: Forecast Errors:1992–2003 1/

(Actual minus forecast, in percentage points)

article image
Sources: Department of Finance, ESRI and the WEO.

For each variable, columns lis mean error (ME), mean absolute error (MAE), standard deviation (SD), and root mean squared error (RMSE).

18. Consistent with the presence of forecast biases in macroeconomic projections, a series of statistical tests finds broad evidence of biases in Ireland’s fiscal forecasts between 1991 and 2003 (Tables 45 and Figure 2). While the tests do not find biases in total revenue or expenditure, almost all of the major subcomponents of tax revenue and current expenditure have forecast biases, as indicated by means and medians of the forecast errors that were significantly different from zero.

Table 4.

Ireland: Descriptive Statistics of One-Year Budget Forecast Errors: 1991–2003 1/

article image
Sources: Fund staff calculations.

For each variable, columns lis mean error (ME), mean absolute error (MAE), standard deviation (SD), and root mean squared error (RMSE).

Table 5.

Ireland: Fiscal Forecasts: Results of Bias and Efficienty Tests 1/

article image
Sources: Fund staff calculations. See Appendix for a description of the methods.

Marks indicate tests that reject a zero median or mean at the 10 percent level. Tests are conducted on forecast errors as a percentage of actuals.

Figure 2.
Figure 2.

Ireland: Revenue Forecast Errors

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

Sources: Fund staff calculations.

19. On the revenue side, a significant part of the forecast error came from deviations in the projections of taxes on income and wealth.6 In contrast, in line with small errors for private consumption growth, the mean forecast errors for both VAT and excise taxes were remarkably small. The RMSEs were particularly large for corporate taxes, capital taxes, and stamp duties. High RMSE for corporate tax revenue may reflect the impact of large export volatility, but also difficulties in projecting the tax base given sustained inflows of FDI to Ireland during the 1990s. Rapidly rising property markets and strong equity markets also contributed to significant upside surprises in stamp duty collections and capital taxes. Indeed, while they account for only about 6 percent of total tax revenues, their overperformance has averaged roughly a third of the forecast errors for tax revenue. As expected, statistical tests confirm forecasting biases in most of the subcomponents of revenue, with the notable exceptions of VAT and excise taxes.

uA01fig11

Tax Revenue Forecast Errors 1995–2003

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

20. Revenue forecast errors can be largely explained by errors in the outlook for growth. Adding the forecast errors for output growth in the mean tests, the null hypothesis of unbiased forecasts can no longer be rejected for the tax revenue projections. Two exceptions are stamp duties and capital taxes, which, as described above, may have been affected by factors beyond the economic cycle in Ireland. Between GNP and GDP growth, GNP appears to be a better proxy for Ireland’s tax base, especially with regard to the base for indirect taxes. Forecast errors for inflation and unemployment do not explain revenue forecast errors well.7

21. The evidence suggests that revenue forecast errors were primarily driven by stronger-than-expected economic activity, but the budget’s cautious approach in projecting revenue also played a role. A calculation of implicit tax elasticity over the period 1991–2003 suggests that budget forecasts relied on a prudent assumption. While the long-term tax elasticity in Ireland is about one, budget estimates appear based on a slightly smaller value of 0.9.

uA01fig12

Non-Interest Current Expenditure

Citation: IMF Staff Country Reports 2005, 370; 10.5089/9781451818826.002.A001

22. On the expenditure side, tests show clear evidence of forecasting bias in current expenditure and interest payments. What makes Ireland rather unique among the benchmark group of countries is a positive statistical relationship between forecast errors in revenue and expenditure. The procyclicality of Irish errors in expenditure forecasts can be illustrated by adding errors in the output projections to the right-hand side of the regression testing for bias in the mean forecast error. When growth errors are included, the null hypothesis of unbiased forecasts for expenditure can no longer be rejected.

E. Institutional Environment

23. The institutional framework for fiscal policy and budget forecasting practices in Ireland are relatively strong compared with other countries (Box 1). The Department of Finance is solely responsible for budget forecasts and prepares its own macroeconomic projections. Over the past decade, economic growth turned out to be generally stronger than projected by the Department of Finance, but the comparison with other forecasters suggests that the latter were not any more accurate. Given the strong powers of the executive branch in Ireland’s parliamentary system, there is little risk of the budget being amended by the legislature, which could potentially undermine the consistency of the budget forecasts.

Key Elements of the Institutional Framework

Fiscal forecasting

  • Department of Finance is solely responsible for macroeconomic and budget projections. There is no involvement of non-government agencies.

  • Three-year forecasting horizon, including the budget year.

Budget system: the executive branch has strong budgetary powers in Ireland’s parliamentary system.

  • Budget is submitted to the legislature 4 weeks before the fiscal year begins. There is no pre-budget parliamentary discussion. The Cabinet’s proposed budget cannot be amended.

  • Mid-year taxation changes require either full supplementary budget or an additional Finance Act. Increases in spending require approval of supplementary estimates (augmentation of programs already included in the budget) or additional estimates (new programs).

Fiscal rules

  • Ireland is subject to the Stability and Growth Pact (SGP), which sets a medium-term fiscal objective of close-to-balance or surplus, with a limit of 3 percent of GDP deficit and 60 percent of GDP debt ceiling.

  • There are Ireland-specific rules and guidelines about expenditure. For example, capital spending (by all sectors of government) is currently capped at 5 percent of GNP, and local government borrowing is capped at €140 million. However, there is no formal mechanism to enforce these rules.

Fiscal relations with sub-national levels of governments

  • The central government enjoys relatively strong control over the sub-national governments. There is no cost sharing arrangement between the central and the sub-national level.

  • There is limited fiscal autonomy at the sub-national level: borrowing by the sub-national governments is governed by the Local Government Act, which requires the approval of the central government to any borrowing proposal. A local authority which fails to agree on an appropriate budget is suspended, and replaced by a commissioner appointed by the central government.

24. However, when government revenue was higher than projected, there was a tendency to increase spending. Ireland’s budget framework allows for supplementary or additional spending during the course of the fiscal year. On the whole, the increases in spending were prudent, as they amounted to less than the revenue overperformance. However, there is the risk that spending increases could be procyclical, adding to demand pressures at times when there is limited slack in the economy.

25. Spending increases were in fact procyclical during 1997–2002. A precise estimate of spare capacity in Ireland is difficult to derive given the effects on the level of potential output of many of the factors accounting for the boom in the 1990s, such as the increased employment and productivity catch up. However, the IMF staff’s estimate indicates that the resource constraint was starting to bind in the period after 1997. Assuming a fiscal multiplier of one, the contribution to growth from expenditures above budgeted levels averaged about 0.6 percentage point of GNP a year during this period.

F. Concluding Remarks

26. The analysis presented in this paper suggests that fiscal overperformance in Ireland was largely due to stronger–than-expected output growth. During 1995–2003, GNP growth in Ireland exceeded the budget forecast by about 1 percentage point on average—highest among the countries sampled—contributing to better fiscal outturns by 0.4 percentage point of GNP. In addition, buoyant asset price developments contributed to the favorable budget surprises. Prudence also played some role. Budget estimates have followed a cautious approach based on prudent assumptions for tax elasticities. A combination of stronger-than-expected growth and the cautious approach employed in forecasting revenues produced consistent revenue overperformance during the period. While revenue overperformance was in part offset by increases in expenditures, spending increases did not exceed revenue overperformance, another aspect of fiscal prudence in Ireland.

27. Looking ahead, growth surprises are less likely to be consistently on the upside. Ireland’s growth prospect remains strong, but there are several important downside risks. These include lower potential growth; upward wage pressures resulting from the next national wage agreement and public sector benchmarking exercise which could have an impact on the public sector bill and the overall fiscal position; and deterioration in Ireland’s competitiveness that can undermine the base for corporate taxes. Moreover, asset developments may not continue to contribute significantly to large upside surprises. Given these risks, a continuation of the authorities’ prudent approach to budget forecasts would prove useful.

28. At the same time, the risk of procyclical spending increases in response to revenue performance needs to be addressed. Discretion to increase spending during the course of a year has been used prudently in the past, such that additional expenditures have not exceeded revenue overperformance. Expenditure over budgeted levels on average has been limited to about half of revenue overperformance. This behavior was prudent from a budgetary perspective, but spending increases at times added stimulus to aggregate demand when the economy was already showing signs of limited slack. This tendency appears to have weakened over the recent few years on the back of tighter expenditure management in place. However, large and long-standing pressures to increase public spending in Ireland are likely to contribute to a continuing potential for procyclical stimulus in the event of upside surprises to growth. The risk of such procyclical stimulus could be mitigated by requiring an assessment of cyclical conditions in the economy before additional spending is approved.

References

  • Artis, and M. Marcellino, 2001, “Fiscal Forecasting: The Track Record of the IMF, OECD, and EC,” Econometrics Journal, Vol. 4, pp. S20– 36.

    • Search Google Scholar
    • Export Citation
  • Auerbach, A.J., 1994, “The U.S. Fiscal Problem: Where We Are, How We Got Here and Where We’re Going,” NBER Working Paper No. 4709.

    • Search Google Scholar
    • Export Citation
  • Blöndal, J.R., 2003, “Budget Reform in OECD Member Countries: Common Trends,” OECD Journal of Budgeting, Vol. 2, No.4, pp.–25.

  • Crippen, D., 2003, “Countering Uncertainty in Budget Forecasts,” OECD Journal of Budgeting, Vol.3, No.2, pp.139– 151.

  • Ireland Department of Finance, Public Financial Procedures. Available on the Internet at http://www.irlgov.ie/govacc/contents.htm

  • Jonung, L., and M. Larch, 2004, “Improving Fiscal Policy in the EU: The Case for Independent Forecasts,” European Commission, Directorate-General for Economic and Financial affairs, Economic papers No. 210.

    • Search Google Scholar
    • Export Citation
  • Lienert, I., 2005, “Who Controls the Budget: the Legislature or the Executive?IMF Working Paper 05/115, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Mathieu, C., and H. Sterdyniak, 2005, “In Search of an Appropriate European Fiscal Framework,” ESRI Budget Perspectives 2005.

  • Mühleisen, M., S. Danninger, D. Hauner, K. Krajnyák, and B. Sutton, 2005, “How Do Canadian Budget Forecasts Compare with Those of Other Industrial Countries?IMF Working Paper 05/66, (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Nordhaus, W.D., 1997, “Forecasting Efficiency: Concepts and applications,” The Review of Economics and Statistics, Vol. 69, pp. 667– 74.

    • Search Google Scholar
    • Export Citation
  • Organization for Economic Cooperation and Development and the World Bank (OECD/WB), 2003, OECD/World Bank Budget Practices and Procedures Database.

    • Search Google Scholar
    • Export Citation
  • Strauch, R., M. Hallerberg, and J. von Hagen, 2004, “Budgetary Forecasts in Europe: The Track Record of Stability and Convergence Program,” ECB Working Paper No. 307.

    • Search Google Scholar
    • Export Citation

APPENDIX I Data Sources and Definitions

  • Budgets in Ireland are published in early December, with the new fiscal year starting on January 1. Budgets do not consistently provide detailed information about the outturn for other than the previous year (t-1). Missing information, especially on the macroeconomic variables, was supplemented by using Economic Review and Outlook published by Department of Finance around the same time as the budget each year. Finance accounts were also used to obtain the fiscal outturns.

  • Data used in this analysis are on a cash basis, focusing on the exchequer borrowing requirement.

  • Ireland implemented a change from the European System of Accounts (ESA) 79 to ESA95 starting in 1995. Following a partial transition between 1995–97, the ESA95 was fully implemented in 1998. There are data issues related to the transition to ESA95 as budget macroeconomic forecasts were still based on ESA79, particularly GDP and GNP levels, while budget outturns were reported based on the ESA95.

  • The study focuses on non-interest current spending because the Irish budget does not provide estimates for discretionary and mandatory spending.

  • Data for the benchmark group was obtained from Mühleisen et al. (2005). It was updated for the United Kingdom with the most recent March 2005 budget figures.

Descriptive Statistics

Forecast errors

Forecast errors for variables projected in budgets are defined as the difference between the actual outturns reported in the budget two years later (t+2) and forecasts in the budget year (t). Thus, forecast error for variables x, Et(x), in percentage points, such as real GNP growth, is defined as:

Et (x) = xt+2 - xt

Forecast errors for variables in levels, for example revenue or expenditure, are calculated as:

Et (x) = ln(xt+2) - ln(xt)

Forecast errors in percent of GDP, such as fiscal balance, are calculated as forecast errors in nominal terms in percent of actual GDP (or GNP):

Et(x) = (xt+2xt)/GDPt+2

Mean error and root mean squared error

Mean error (ME) and root mean squared error (RMSE) are defined as:

MEt=1/Tt=1TEtRMSEt=(1/Tt=1TEt2)1/2

where the following relation holds:

RMSEt=(MEt2+σt2)1/2

Bias tests

The study uses three rank-based, nonparametric median tests for forecast bias:

  • Binomial sign test, which tests whether the sample is drawn randomly from a binomial distribution such that the sample proportion above and below the true median should be one-half;

  • Wilcoxon signed ranks test, which is based on the idea that the sum of the ranks for the samples above and below the median should be similar; and

  • Van der Waerden test, which is similar to the Wilcoxon test but is based on smoothed ranks. The signed ranks are smoothed by converting them to quantiles of the normal distribution.

Mean tests are conducted by regressing forecast errors on a constant and testing whether or not it is significantly different from zero.

Efficiency tests

Tests of forecast efficiency check whether forecasts were based on all of the information available at the time they were made. They are based on a test of whether forecast errors are independently, if not normally distributed. More formally, following a regression of the actual outturn on its forecast,

yt+2 = β1 + β2yt + εt

it tests the joint hypothesis of β1 = 0 and β2 = 1.

1

Prepared by Keiko Honjo.

2

Appendix I provides an explanation of the data and statistical tools used in this study.

3

There is a substantial contribution of multinationals to Irish output, and associated profit flows (including the effects of transfer pricing) creates significant differences between measures of GDP and GNP. Hence, the paper uses GNP for Ireland as it reflects a better measure of Ireland’s tax base.

4

The relatively weak forecast accuracy in Ireland may be in part related to Ireland’s higher average growth rate than other countries during this period.

5

The Economic and Social Research Institute (ESRI) is a private think tank in Ireland. For the WEO projections, this study uses projections published in April or May each year reflecting the time lag between when the projections are made and when they are released.

6

Excluding social security contributions, taxes on income and wealth (income tax and corporation tax) account for about 45 percent of total tax revenue; VAT and excise taxes about 40 percent; and capital tax and stamp duty about 6 percent.

7

Forecast errors for either nominal GDP or GNP are also strong candidates for explaining errors in revenue projections. However, a shift from the European System of Accounts (ESA) 79 to ESA95 during this period hampers accurate analysis.

Ireland: Selected Issues
Author: International Monetary Fund