CHAPTER 2: Sources of Fiscal Risk
- Lusine Lusinyan, Aliona Cebotari, Ricardo Velloso, Jeffrey Davis, Amine Mati, Murray Petrie, and Paolo Mauro
- Published Date:
- January 2009
Fiscal risks—deviations of fiscal outcomes from what was expected at the time of the budget or other forecast—arise from macroeconomic shocks and the realization of contingent liabilities. Sources of risk include various shocks to macroeconomic variables (economic growth, commodity prices, interest rates, or exchange rates) as well as calls on several types of contingent liabilities (obligations triggered by an uncertain event: including both explicit liabilities—those defined by law or contract, e.g., debt guarantees—and implicit liabilities—moral or expected obligations for the government, based on public expectations or pressures, e.g., bailouts of banks or public sector entities).4
Fiscal risks covered by this paper’s definition will vary in a number of respects, calling for different responses in terms of disclosure and management. For example:
Temporary vs. permanent. Higher-than-expected fiscal deficits resulting from temporary growth slowdowns against a background of low debt may simply require allowing the automatic stabilizers to work. Permanent shocks affecting fiscal sustainability in a lasting manner would have more important implications.
Correlation. Whereas shocks that are likely to offset each other may call for little response, the possibility of positively correlated or mutually reinforcing shocks (e.g., exchange rate, debt, and banking crises) warrants greater policy action.
Forecasting. Deviations of fiscal outcomes from expectations may reflect weak forecasting capacity or “strategic” forecasts (whereby a government might use overly conservative commodity price assumptions to dampen expenditure demands from the legislature or to build a buffer against possible price declines, or optimistic revenue forecasts to facilitate the approval of ambitious spending plans). This highlights the importance of accurate forecasts.
Quantification. Whether fiscal risks are disclosed in a quantitative or qualitative manner depends on whether the fiscal cost of an event and the probability of its occurrence can be reasonably estimated.5Quantification is usually easier for macroeconomic risks and explicit guarantees (which include contractual terms and amounts) than for implicit guarantees.
Sensitivity. Major fiscal risks are often related to areas where expectations of government policies need to be managed carefully, such as problems in the banking system or overvalued exchange rates. This needs to be recognized in designing disclosure modalities.
To gauge the importance of different sources of fiscal risks, this section draws on both realization of risks and forward-looking risk estimates. It analyzes differences between projections and outcomes with respect to variables such as the debt-to-GDP ratio, fiscal deficits, and a residual term in the stock-flow reconciliation. This documents the macroeconomic relevance of fiscal risks, and highlights the importance of debt increases that are not captured in the deficit (examples include assumption of debts and other off-balance-sheet items). Empirical evidence is then presented on the fiscal consequences of each type of shock, based on forward-looking estimates of the implications of changes in macroeconomic variables or the potential costs of contingent liabilities, and ex post estimates of the fiscal costs of shocks such as banking crises and natural disasters.
Macroeconomic Significance of Fiscal Risks
The macroeconomic significance of fiscal risks is highlighted by comparing expectations with outcomes for fiscal variables. A comparison of World Economic Outlook (WEO) forecasts with outturns of fiscal variables such as the debt-to-GDP or deficit-to-GDP ratios shows that unexpected changes are often large and vary widely, although their average is close to zero.6 In a panel of 27 advanced economies for 1995—2007 and 131 emerging and developing countries for 2002—07 (the largest panel for which forecasts of fiscal variables are available), the 10th percentile unexpected worsening (that is, the 10th worst realization of risks in 100 observations) amounts to 7.2 percentage points for the debt-to-GDP ratio and 1.7 percentage points for the fiscal balance to GDP ratio (Figures 1 and 2).7 Unexpected changes in fiscal variables are somewhat larger in emerging/developing countries, but are substantial for advanced countries as well. To confirm that unexpected changes can, in hindsight, be matched to economic shocks, Figures 1 and 2 also illustrate the high frequency of large unexpected improvements in fiscal variables for oil-exporting countries in years when oil prices rose.
Figure 1.Unexpected Changes in the Debt-to-GDP Ratio
Notes: Differences between the debt-to-GDP ratio for year t as forecast in the WEO Fall submission of year t-1 and the realization for year t observed in the WEO Fall vintage of year t+1. The analysis is based on 27 advanced economies for 1995—2007 and 131 emerging/developing countries for 2002—07.
Figure 2.Unexpected Changes in the Fiscal Balance to GDP Ratio
Notes: Differences between the fiscal balance to GDP ratio for year t as forecast in the WEO Fall submission of year t-1 and the realization for year t observed in the WEO Fall vintage of year t+1. The analysis is based on 27 advanced economies for 1995—2007 and 131 emerging/ developing countries for 2002—07.
The largest unexpected increases in the debt-to-GDP ratio are often related to exchange rate depreciations and calls on contingent liabilities. A decomposition of unexpected increases in the debt-to-GDP ratio into: (i) unexpected rises in deficits, (ii) a contribution from unexpected economic growth slowdowns, and (iii) a residual term including factors such as exchange rate depreciation and calls on contingent liabilities points to the importance of the residual term (Box 1).8 Many large and unforeseen increases in the debt-to-GDP ratio reflect the inclusion of debts (e.g., from bailouts of banks or state-owned enterprises) that had not been previously recorded in general government debt.9 Worsenings in the deficit or economic growth are significant but feature less prominently.
Individual Sources of Risk
Unexpected changes in key macroeconomic variables imply substantial fiscal risks. Forward-looking estimates of risks from macroeconomic variables—in the form of standardized bound tests used in IMF debt sustainability templates—show that a one-half standard deviation permanent shock to real growth would increase the debt-to-GDP ratio five years later by 6.8 percent of GDP on average in a sample of 19 advanced and emerging market countries. A one-half standard deviation shock to the primary deficit would raise the debt-to-GDP ratio by 5.2 percentage points. And a one-half standard deviation shock to interest rates would lead to somewhat smaller increases on average, though it would have even more significant effects in countries that rely primarily on floating interest rate debt. In developing countries (based on a limited sample), a decline in economic growth would have an especially notable effect on debt dynamics (Table 1).
Box 1.Sources of Fiscal Risks: Decomposition of Unexpected Changes in the Debt-to-GDP Ratio
Unexpected increases in the debt-to-GDP ratio are decomposed into unexpected rises in deficits, a contribution from unexpected economic growth slowdowns, and a residual term including factors such as exchange rate depreciation and calls on contingent liabilities:
where Δ indicates a change over the previous year and all variables refer to differences between WEO forecasts for year t made in year t–1 and outturns for year t based on data observed in year t+1; λ≡(γ/1+γ), where γ is the nominal rate of economic growth; and ε is the residual term. An analysis based on the magnitude of the 10th percentile worsenings for each component points to the importance of the residual term in accounting for unexpected increases in the debt-to-GDP ratio.
|Emerging Market and Developing Economies|
|All countries||Advanced||All||Oil-exporting||Non oil-exporting|
|Number of observations||415||261||154||27||127|
The 10th percentile worsening is largest for the residual term. Adverse surprises in the deficit or economic growth are somewhat smaller, partly because the exercise is based on changes within one year; the relevance of drops in economic growth and worsening deficits increases at somewhat longer horizons.
A variance decomposition confirms that the residual term accounts for the bulk of unexpected changes in the debt-to-GDP ratio, with surprises in the deficit or in the contribution from growth playing a smaller role.
|Interest Rate||Real GDP Growth||Primary Balance||Combined Shock||Exchange Rate|
|Advanced and Emerging Market Economies|
The impact of exchange rate depreciations is immediate, and can be especially strong when a large share of the debit is in foreign currency. A 30 percent depreciation of the real exchange rate would increase the debt-to-GDP ratio by 8 percent in the year of the shock and (reflecting gains in competitiveness) 6.5 percent after five years in the sample of advanced and emerging economies; and by similar amounts in developing countries. Indeed, turning to information on ex post realization of risks, exchange rate depreciation accounted for a major share of the increase in the debt-to-GDP ratio in the context of several emerging market crises during the 1990s (de Bolle, Rother, and Hakobyan, 2006).10
Changes in commodity prices also have important fiscal implications, especially for commodity producers. For example, a US$20 decline in oil prices would lead the overall fiscal balance to worsen by 10 percentage points of GDP in a sample of oil-producing countries (Ossowski and others, 2008). As seen above, the magnitude of the impact is also apparent in the large negative forecast errors for the debt-to-GDP ratio of oil producers during years characterized by oil price increases. Commodity price changes affect the fiscal sustainability of commodity importers primarily through economic growth, though their direct fiscal impact may be considerable for countries with energy subsidies.
For low-income countries, volatile aid flows and the need to cushion the poor from external shocks present special challenges. In some highly aid-dependent countries, aid is more volatile than fiscal revenues, and shortfalls in aid and domestic revenues tend to coincide. More generally, uncertainty about aid disbursements is large and the information content of commitments made by donors is limited (Bulíř and Hamann, 2003). Moreover, sharp increases in staple food prices may unexpectedly require incurring sizable fiscal costs.
However, some of the largest fiscal costs have arisen from contingent liabilities. Examples include:
Banking crises. A review of the fiscal costs of systemic banking crises identified 24 episodes in which cumulative costs exceeded 5 percentage points of GDP, based on a sample of 117 banking crises that occurred in 93 countries during 1977—98. It estimated costs at 30—55 percent of GDP in Argentina, Chile, and Uruguay in the early 1980s, 25—50 percent of GDP in Indonesia, Korea, and Thailand in 1997—98, and about 20 percent of GDP in Japan in the 1990s (Honohan and Klingebiel, 2000).11 Such costs arise primarily from depositor and debtor bailouts, open-ended liquidity support, and repeated recapitalization programs—and are often larger when incurred after years of implicitly subsidized lending by state-owned financial institutions.
Natural disasters. Direct economic losses from natural disasters have often exceeded 10 percentage points of GDP in developing countries and amounted to a few percentage points of GDP in some advanced countries (Freeman, Keen, and Muthukumaral, 2003); such losses are unevenly distributed across countries, as disasters usually revisit the same geographic zones. The fiscal implications are clearly substantial, though estimates are available only for a limited sample; a study on Latin American and Caribbean countries found several episodes when the fiscal deficit rose substantially in the aftermath of natural disasters (Caballeros Otero and Zapata Martí, 1995).
State-owned enterprises. Public enterprises have often been a significant source of contingent government liabilities, especially as a result of political interference, mismanagement, or irresponsible borrowing. Losses or excessive debt have resulted in costly government bailouts, especially in the aftermath of crises.12
Subnational government bailouts. Subnational government defaults or bankruptcies have often led central governments to provide rescue packages, occasionally with large costs: examples include Brazil (7 percent of GDP in 1993 and 12 percent of GDP in 1997; Bevilaqua, 2002), Argentina (1 percent of GDP, cumulative, in the mid-1990s; Nicolini and others, 2002), and Mexico (1 percent of GDP in the aftermath of the Tequila crisis; Hernández-Trillo and others, 2002).13
Legal claims. Governments have paid compensation in legal cases related to disparate claims; the amounts, often difficult to predict prior to a ruling, can be sizable. Examples include war claims and frozen foreign currency deposits (Bosnia and Herzegovina, 12 percent of GDP); litigation on domestic arrears (Chad, 9 percent of GDP); claims related to privatization (Brazil); liquidation of SOEs (Brazil and Indonesia); personnel management (Brazil and France); compensation for real estate and other property losses (Lithuania and Poland); tax refunds (Indonesia); bank restructuring guarantees (Czech Republic); and environmental cleanup (e.g., related to defense or nuclear power; Canada and United States).
Guarantees. Although systematic information on actual calls on guarantees is limited, it is clear that potential risks from guarantees are substantial. Information on exposure is available for explicit guarantees legally binding the government to take on an obligation should a specified event occur (e.g., price guarantees, loan guarantees, or profit guarantees): these amounted to 12 percent of GDP on average in a sample of then pre-EU accession countries as of end-2002 (European Commission, 2004) and to 5 percent of GDP in the countries for which questionnaire responses were available.
Public-private partnerships. PPPs have gained importance as a source of fiscal risks in many advanced and emerging market economies—particularly for large investment projects in transportation infrastructure and the power sector (Hemming and others, 2006).14 They often entail fiscal obligations not captured in the fiscal accounts: for example, state guarantees for concessionaire borrowing, minimum revenues, or exchange rate losses. Indeed, there is growing anecdotal evidence of costly PPP failures due to unrealistic demand projections or other shortcomings in project planning and management.15
Looking ahead, the relative importance of various types of contingent liabilities may increase. For example, survey respondents identified guarantees, especially those linked to PPPs, as among the most important sources of fiscal risks in the future. These developments will need to be borne in mind when turning to appropriate policies in fiscal risk disclosure and management.
The term “contingent liability” throughout this paper refers to its general use as “spending that may be triggered by a future event.” This differs from the accrual accounting definition of “contingent liability” (not recognized on the balance sheet as a liability) as linked to events that are less than likely to occur.
Under Knight’s (1921) definition, situations where an event’s expected cost (the product of the event’s cost times the probability of its occurrence) cannot be quantified would be labeled as “uncertainty,” whereas situations where probabilities and costs can be estimated would represent “risk.”
Unexpected changes in debt-to-GDP ratios are computed as the difference between forecasts for year t based on the October vintage of the year t-I WEO, and outturns for the year t recorded in the WEO’s October vintage of year t +1. Unexpected changes for other variables, such as the fiscal deficit as a ratio of GDP, are computed in a similar manner. Instances in which the debt-to-GDP ratio changed unexpectedly because of debt restructurings or changes in the debt concept reported to the WEO are omitted from the sample. sytematic studies of the accuracy of WEO forecasts found little, if any, bias in WEO forecasts of macroeconomic variables (Timmermann, 2007) or fiscal variables (IMF, 2003).
The 10th percentile is chosen to reduce the influence of extreme values, or outliers.
Such a residual term, often referred to as the “hidden” deficit, is a key determinant of debt dynamics (Kharas and Mishra, 2001; Panizza, Jaimovich, and Campos, 2006; Polackova Brixi and Schick, 2002). The largest residual terms found within the sample analyzed by IMF staff often relate to exchange rate depreciations (recent examples include Egypt, 2003; Iceland, 2001; and Israel, 2002) and recognition of public sector obligations (e.g., Canada, 1999—2000; Cape Verde, 2005—6; Egypt, 2003; Greece, 2002 and 2004; Japan, 1998 and 2006; and Mauritius, 2003).
Many revisions apply retroactively to the debt-to-GDP series for several years prior to the year in which the “surprise” is observed. While unexpected increases in debt often reflected improved recording of existing obligations, they sometimes revealed that obligations had accumulated in various parts of the public sector and had to be recognized on the government’s books.
Exchange rate depreciation accounted for about half of the increase in Brazil (1998) and Indonesia (1998); essentially all of the increase in Argentina (2001), the Philippines (1998), Turkey (2001), Ukraine (1998), and Uruguay (2002); and more than all of the increase (the debt-to-GDP ratio was reduced by other factors) in Ecuador (1999), Mexico (1995), and Russia (1998). The debt-to-GDP ratio jumped by more than 30 percentage points of GDP on average during these crises.
In a number of cases, Honohan and Klingebiel’s (2000) method does not fully reflect recoveries and may thus be considered an upper bound on the net present value of the fiscal and quasi-fiscal costs. At the same time, banking crisis interventions were often financed with central bank debt that remained on the central bank’s balance sheet for many years (Stella and Lönnberg, 2008).
Examples relate to the power sector (Indonesia, where during the 1998 crisis the central government paid for the electricity company’s fuel costs, amounting to 4 percent of GDP; and the Philippines); airlines (subsidies/bailouts averaging US$2 billion each for several airlines in Europe); railways/metro (Colombia, Hungary, Japan, Malaysia, and Thailand; 1–5 percent of GDP); and water authorities (Jordan, 3 percent of GDP).
In Italy, central government bailouts of subnational government health units ranged between 0.2 and 0.6 percent of GDP yearly over the past five years.
It is important to note that many PPP contracts involve even larger fiscal risks for the long term than they do for the medium term.
During the 1990s, calls on demand guarantees related to PPPs in power, telecoms, and toll roads in Colombia resulted in cumulative payments of 2 percent of GDP by 2004. Substantial obligations on PPP contracts in power plants and roads also became due in Indonesia, Malaysia, and Thailand during the Asian crisis. More recently, governments have provided new state guarantees, equity contributions, operating subsidies, or full bailouts and renationalization in the transportation infrastructure sector, in countries including Australia, Hungary, Mexico, and the United Kingdom (OECD/ITF, 2008), with gross costs for individual projects often amounting to ½ percent of GDP.