Posted by Ian Lienert
When budget revenue projections are conservative, governments receive pleasant surprises – they benefit from more tax and nontax revenues than planned, which allows debt to be reduced faster than planned or, alternatively, expenditure to be higher. In contrast, if budget revenue projections are too optimistic, it is difficult to cut back expenditure, at least in the short-term. This can result in public debt increasing to a level higher than desirable, assuming market deficit-financing is available. In countries where debt market financing is unavailable, there is likely to be an undesirable build-up of payment arrears.
An IMF Occasional Paper published in 2007 examined the budget forecasting experience in eleven OECD countries. It found that in 8 of the 11 countries, revenues were underestimated during the late 1990s and early 2000s. Canada was the country that most strongly underestimated tax and nontax revenues. The countries for which revenue projections erred on the optimistic side during the observation period were France, Germany and the United States. It may be concluded that OECD countries leant towards revenue projection conservatism during this period.
In chapter IV of Occasional Paper No. 258, an IMF staff team first examined the institutional arrangements for budget forecasting processes in Australia, Canada, France, Germany, Italy, Netherlands, New Zealand, Sweden, Switzerland, United Kingdom and United States. The study then analyzed forecast errors -- forecasts minus outcomes -- for various macroeconomic and fiscal variables. The statistical analysis was generally conducted for a nine year period, from 1995 to 2003.
Concerning institutional aspects, the study noted that, whereas budget projections are prepared by the Ministry of Finance (or the equivalent) in all countries, the degree to which the forecasting process is formalized varies. Some countries prepare stylized forecasts with some cross-checks with sectoral and revenue experts (e.g., Sweden, Switzerland). Others use detailed model-driven processes maintained by technical experts (e.g., Australia, France and the United States). Some countries reassess their fiscal forecasts after consulting with the private sector (e.g., Australia, Canada). In the United States, the Congressional Budget Office (CBO) plays a similar role – it provides a view independent from that of the “ministry of finance” 1/. In the case of the CBO, there is a mandate to prepare 10-year projections of major fiscal variables based on current legislated policies. A few other countries have established congressional or parliamentary budget offices (see, for example, the blog of July 21, 2008, on the recent establishment of such an office in Canada; http://blog-pfm.imf.org/pfmblog/2008/07/canada-creates.html ).
With regard to revenue forecasting, Canada stands out as the country with the most consistent and largest underestimation of budget revenues. Canada’s GDP forecasts were deliberately conservative during 1995-2003: actual economic growth was on average ½ percentage point higher than that assumed in budget projections. Inflation (the GDP deflator) was also underestimated, by 0.2 percentage points. All tax and nontax revenue components were underestimated (the forecasting records of four different taxes were analyzed).
Statistical tests were conducted to detect whether, during the small sample period, there were systematic tendencies to underestimate nominal GDP, total revenues and nontax revenues during the 1990s and early 2000s. This analysis revealed that Canada, New Zealand, Sweden, and the United Kingdom all exhibited a consistent bias in either the macro forecasts or for aggregate fiscal revenues. In all countries, errors in output projections tend to explain a substantial share of revenue errors. Canada stood out as the country where a number of small, unidirectional forecast errors led to the overall bias towards conservatism in revenue projections.
For full results, see the “Budget Forecasting” chapter in IMF Occasional Study 258 Northern Star: Canada’s Path to Economic Prosperity, published in 2007 and available for purchase – see http://blog-pfm.imf.org/pfmblog/2008/07/canada-creates.html. The chapter is based on a more comprehensive study, IMF Working Paper/05/66, How Do Canadian Budget Forecasts Compare with Those of Other Industrial Countries? prepared by Martin Mühleisen, Stephan Danninger, David Hauner, Kornélia Krajnyák, and Bennett Sutton in March 2005 (see http://www.imf.org/external/pubs/cat/longres.cfm?sk=18080.0). The preparation and publication of these studies were strongly supported by Canadian officials.
Blog comments on the study
The research is solidly-based and includes a battery of statistical tests for errors in macroeconomic and fiscal variables; it also discusses the interaction between forecast variaiable. Although the study is now beginning to become a little dated, it indicates that the majority of selected OECD countries leaned towards pessimism in their budgetary revenue projections. However, it is difficult to generalize this conclusion for the 11 countries examined. The reasons are:
• Limited time series and lack of strong results. The sample period is less than 10 years. Also, statistically significant results that indicate a forecasting bias towards conservatism were obtained for a few countries only.
• Buoyant economic growth in sample period. The late 1990s and early 2000s was characterized by a period a buoyant economic growth. This resulted in both growth and budget revenues being higher than assumed. If the time series were to be updated to include the worldwide weakening of economic growth in 2008, for example, the results would mostly change somewhat.
• Data caveats. The study acknowledges the various data caveats needed for postmortems of forecasting errors. Both “budget projections” and “actual” outcomes can be revised, depending on each country’s budget and statistical reporting processes. These impinge on the empirical analysis. For example, supplementary budgets that are substantially different from initial budgets adopted by parliament can materially affect the entire analysis.
• Country specific factors play an important role in these results—but these can be changed over time. OECD countries periodically review their projection methods, including ways of assuring forecast quality (the study examines relevant issues).
• For example, the “prudence factor” and budget contingency reserves affected the results for Canada. During 1994 to 1998, prudence was explicitly incorporated into the Canadian forecasts by deliberately adopting budget forecast assumptions that were more pessimistic than private sector forecasters. During these years, the Department of Finance based budget projections on lower economic growth and higher interest rate assumptions than the average of private sector forecasters (on which Canada relies more heavily than other countries). This practice was helpful for contributing to the reduction in public debt, which had built up to worrisomely high levels in the early 1990s.
• Is revenue forecasting pessimism desirable? The results for Canada suggest that GDP and revenue forecast pessimism is desirable when a strong fiscal adjustment and reduction in debt is necessary. However, Canada stopped deliberately making conservative model-based forecasts after the worst of the fiscal crisis of the early 1990s was over. Clearly there are limits to which a strong ministry of finance can continue deliberate revenue projection pessimism, especially given that, in annual budget negotiations, spending ministries are likely to be more aggressive in obtaining the available “fiscal space” to maximize their budget revenues. Also, in countries where legislatures have considerable scope for changing the executive’s draft annual budget – strong amendment powers – parliament/Congress may increase budget expenditures if it suspects that the executive has deliberately been pessimistic in its revenue projections. When these tradeoffs occur, there may be a case for the ministry of finance to prepare realistic, rather than pessimistic, GDP and/or revenue projections. Clearly the case for realistic forecasts is stronger in countries whose fiscal and public debt positions are sufficiently comfortable.
1/ The “ministry of finance” function in the United States is performed by an Office of Management Budget (for spending projections) and a Department of Treasury (for revenue and debt projections). These two executive departments are complemented by a Council of Economic Advisors – the trio agrees on macroeconomic projections and forecast assumptions.
2/This is the reason why Japan was excluded from the sample. In the late 1990s, Japanese fiscal policy was implemented largely through supplementary budgets, whose projections differed significantly from those in initial budgets (because of intentional policy changes). In view of Japan’s fiscal policy responses during the Asian crisis (in particular), it would be very difficult to compare Japan’s forecast errors with those of the other countries, where supplementary budgets were admittedly adopted, but had a lesser budgetary impact.
