III. Assessing Indonesia’s Public Debt Level1
1. This chapter examines the issue of what would be an appropriate upper level of public debt for Indonesia. This is an important issue because: (i) the level of public debt can affect economic growth through its impact on the economy’s vulnerability to shocks and the government’s claims on financial resources for debt service; (ii) Indonesia’s public debt and debt service ratios remain high and lingering weaknesses in the fiscal management framework, including an unstable and petroleum-dependent revenue base, add to the vulnerabilities stemming from fluctuations in the exchange rate and interest rates; and (iii) Indonesia’s re-access to domestic and international credit markets has increased the prospects for market borrowing in the future.
2. The appropriate upper level of public debt is estimated by quantifying the public debt threshold at which significant economic events take place. These are the thresholds at which: (i) the potential for a crisis becomes significant—i.e., the “early warning signals” approach; (ii) the debt level would significantly reduce growth2—i.e., the “debt overhang and growth” approach; and (iii) the future debt level would become unsustainable for given shocks to macroeconomic variables—i.e., the “stochastic debt sustainability” approach. By focusing on public debt, this study differs from most other debt threshold studies, which have looked at external debt and examined episodes of large external debt reductions. There have been few empirical studies of public debt sustainability, mainly because of the difficulty in obtaining reliable and comparable cross-country time series data on public debt.3
3. This study’s empirical results suggest an upper level for public debt in the 35–42 percent of GDP range for Indonesia. Though this result needs to be interpreted with caution, it is comparable to results of other studies on external debt thresholds.4 Thus, while showing that “safe” debt thresholds vary from country to country and depend on history, Reinhart, Rogoff and Savastano (2003) found that the risk of a credit event increases significantly when the external debt to GNP ratio exceeds the 30 to 35 percent of GNP range in a debt intolerant country (i.e., a country with a relatively low institutional investor rating). Patillo, Poirson, Ricci (2002, 2003) find that external debt levels above 35–40 percent of GDP are detrimental to growth. IMF (2003) finds that external debt thresholds of 40 percent for the average emerging market and below 55 percent of GDP for low-income countries are appropriate.
B. The Characteristics of Indonesia’s Public Debt
4. Indonesia’s public debt ratio jumped during the crisis, followed by a partial reversal in subsequent years. The public debt ratio increased from about 30 percent of GDP in 1995 to 54 percent of GDP in 2004. The debt ratio peaked at 92 percent of GDP in 2000, with the increase mostly reflecting bank restructuring costs in the wake of the financial crisis and further depreciation of the currency in 2000 (of some 30 percent). Since then, economic recovery, fiscal consolidation, and lower interest rates have helped reduce the debt ratio.
5. The government has improved the debt structure, but it still entails substantial exchange rate and interest rate risk. While the foreign currency denominated debt has been declining, to represent about 49 percent of all government debt in 2004, and is held mainly by official creditors, such a high proportion still implies that exchange rate fluctuations have a sizable impact on the debt level and debt service.5 Moreover, 54 percent of outstanding government bonds carry variable interest rates, which leaves the government highly vulnerable to interest rate shocks. The government has made substantial progress in improving the structure of its tradable debt, by lengthening the duration of debt through debt buybacks and exchanges, smoothing the debt service profile and substantially reducing rollover risks; the average maturity of the domestic bond portfolio is currently about 7.2 years.
6. Debt service costs remain high. While total debt service costs declined to about 6 percent of GDP in 2004, public debt service still represented as much as 53 percent of non oil revenues and 42 percent of primary spending, with the interest bill alone accounting for about one fifth of total spending. The heavy debt burden has hampered the government’s ability to boost spending in priority sectors such as health, education, and infrastructure.
7. Indonesia’s public debt ratio has converged toward the regional average but may still be on the high side. The September 2003 World Economic Outlook study showed that public debt had grown across emerging markets, to about 72 percent of GDP in 2002. While this is substantially higher than Indonesia’s present debt ratio, Indonesia is broadly at par with the regional average. The September 2003 WEO notes, however, that major recent default cases and financial crises have occurred even with a public debt ratios below 45 percent of GDP.6
Public Sector debt
C. Debt Threshold and Vulnerability: an Application of the Early Warning System
8. One yardstick used to measure debt intolerance is the public debt level associated with a currency crisis through the application of the early warning system (EWS). EWS studies typically monitor the evolution of a number of economic indicators that tend to behave differently in a systematic manner prior to a crisis. Such an analysis involves the use of a non-parametric method and high-frequency data. Results have tended to show that external debt levels (either as percent of total exports or GDP) are good predictors of currency crises.78 A high debt level could exacerbate capital account problems and lead to capital flight if it raised concerns about debt sustainability.
9. The EWS methodology requires a precise definition of a crisis and a method for generating predictions. The empirical analysis used in this chapter concentrates on a sample of 14 emerging markets in Latin America and Asia, with monthly data interpolated from yearly and quarterly data for 1990–2002.9 The following method was employed to define the threshold:
Crisis period: A crisis is defined as “a situation in which an attack on the currency leads to sharp currency depreciations, large losses in reserves or a combination of both.” Months in which a weighted index of these variables is above the mean by more than three standard deviations are defined as crises.10
Threshold identification. The public debt to GDP ratio, Xj, is said to signal a crisis in period tif the public debt to GDP ratio crosses the critical threshold point X’.The threshold is defined in relation to percentiles of the distribution of the indicator.
Critical threshold: The critical threshold itself is found by performing a grid search from the 10th to the 40th percentiles of the public debt variable and selecting the value that minimizes the noise-to-signal ratio (i.e., the number of false signals relative to good signals). Any noise to signal ratio greater than one leads to bad predictors as the indicator will be giving more false signals than accurate ones. Any such indicator should be excluded from the analysis.
10. The application of the EWS methodology to the sample of emerging markets shows that:
The public debt to GDP ratio is a good crisis predictor. This is indicated by its minimum noise to signal ratio of about 0.9.
For Indonesia, the analysis points to an upper debt ratio of 42 percent of GDP. This result is derived from the minimum noise-to-signal ratio corresponding to the 34th percentile of the debt variable, which in the case of Indonesia corresponds to a debt-to-GDP ratio of 42 percent of GDP.
The use of this threshold to signal a currency crisis should be nuanced, as it is associated with an unconditional probability of a crisis of only 11 percent.11 Other caveats associated with the method are that the composition of debt may make a difference (e.g., Indonesia’s large share of official creditors may point toward a higher threshold), and that many false signals are still possible.
D. Debt Overhang and Growth
11. The second approach to gauging the appropriate public debt level assesses the link with economic growth. The theoretical rationale for such a link is that at relatively low levels of debt, borrowing may have a positive impact on growth through increased investment, leading to higher transitional growth (Eaton, 1993). On the other hand, higher debt levels would be expected to lead to a higher tax burden in the future, which would lead to lower investor expectations of after-tax returns, thereby lowering investment and growth (Krugman (1988); Agenor and Montiel (1996)). By the same token, a higher debt level would tend to limit resources available for spending on productive activities because of the debt service burden, and to crowd-out the private sector from credit markets.
12. The empirical literature linking growth and debt has concentrated on external debt.Elbadawi (1997) found a growth maximizing external debt to GDP ratio of 97 percent, which seems quite high. More recent papers by Patillo et al. (2002) found that external debt levels above the 35–40 percent of GDP range might be detrimental to growth. None of the previous studies have estimated the empirical relationship between public debt and growth.
13. The empirical analysis presented below uses annual panel data for 51 countries covering 1972 to 2002. The data was collected from WEO and the World Bank Development Indicators, and complemented by the Patillo, Poirson and Ricci (2003) dataset. The analysis used three-year averages of annual growth data in order to allow for a smoothing of short run fluctuations and business cycles effects. The results conform to those shown in studies on external debt. The results could be further tested for robustness by separating the sample between the countries having access to capital markets and those that do not, and by using a more comprehensive public debt data series spanning the last three decades.12
14. The empirical analysis suggests that debt and long-term growth were negatively correlated. Such a relationship is present linearly but also in a multivariate setting after controlling for other factors that affect growth, such as terms of trade, population growth rates and fiscal balance, and is consistent across different specifications (Table).13 The results imply that an increase in the debt to GDP ratio by 2.7 percent would reduce growth on average by about 1 percent.14 These results hold for Indonesia, as suggested by the insignificant coefficient on the interactive debt dummy.
15. The methodology was then refined in order to identify the debt level at which the overall impact on growth becomes negative. That is, the analysis seeks to determine the level from which any increase in debt would lead to lower growth. To this end, the following regression was estimated:
|Log (Lagged income)||-1.56||***||-0.75||*||-8.53||***|
|Terms of Trade||0.045||***||0.053||***||0.03||*|
|Log (pop. Growth)||-4.32||***||-4.49||***||-6.6|
|Indonesia debt interactive dummy||0.21||0.59||**|
|Number of observations||394||380||394|
Where y is growth in income per capita, X represents factors controlling for growth, and Dj represents debt dummy variables divided into quintiles (see Table for different thresholds, which were established by ranking all debt observations and establishing thresholds corresponding to each quintile). The regression results should be interpreted by assessing the effect of debt on growth in comparison to the lower quintile. They indicate that debt levels belonging to the third, fourth, and fifth dummy would reduce the growth level, implying that debt levels above 38 percent of GDP would be detrimental to growth.15 The introduction of an Indonesia dummy variable interacted with the debt quintile dummies do not change the result.
|Number of observations||Thresholds|
|OLS and fixed effects|
|Interactive Indonesia Dummy 3/||1.69||1.57||0.95|
Coefficient is with respect to Quintile 1 dummy.2/ Control variable coefficients are consistent with those found on page 6.
Interacted with debt dummy of the second quintile.
Coefficient is with respect to Quintile 1 dummy.2/ Control variable coefficients are consistent with those found on page 6.
Interacted with debt dummy of the second quintile.
16. The inclusion of additional factors in the analysis could potentially have affected the appropriate debt range chosen. A number of studies have found that countries with better institutions experience stronger growth, which boosts revenues and eases the debt servicing burden, while weak political systems often delay fiscal adjustment and accumulate short-term debt.16 Low government revenues, as in the case of Indonesia, would also imply that governments will often have difficulty meeting their desired expenditures, thereby increasing the pressure to borrow. In addition, a different composition of the debt structure including its financing and maturity would have affected the appropriate debt level chosen.
E. Stochastic Debt Sustainability
17. The third measure of debt intolerance looks at the effects of stochastic shocks on the debt level. The basic premise of this method is that a country’s public debt is deemed sustainable as long as the public debt to GDP ratio remains constant at its current level or declines over time even under adverse circumstances.17 Debt dynamics are modeled in the following way:
where dis the debt-to-GDP ratio, pb is the primary balance, f is a weighted average of domestic and foreign interest rates, α is the share of foreign-currency denominated public debt, π is the change in the domestic GDP deflator, and g is the real GDP growth rate. Changes in the exchange rate (local currency per U.S. dollar) are denoted by σ, with σ >0 indicating a depreciation of the local currency.
18. Debt sustainability is assessed based on projections of key variables affecting future debt ratios. The method postulates that the net present value of the primary surplus must equal today’s debt to GDP ratio as a condition for debt sustainability. Underlying variables such as interest rates, economic growth, and the primary surplus, are projected. The appropriate debt threshold would be the medium-term debt target that would be resilient to “reasonable” adverse circumstances (such as shocks to individual variables of one or two standard deviations, or alternative scenarios that combine several shocks).
19. The analysis suggests that Indonesia’s debt position would be sustainable with a public debt ratio of about 35 percent of GDP or less. Again, this is based on the premise that the public debt is sustainable as long as the debt-to-GDP ratio is projected to remain constant at its current level (a little above 50 percent of GDP) or declining, even when the economy is subject to “normal” shocks. The 35 percent of GDP threshold is predicated in the baseline medium-term scenario presented in the staff report and would be sensitive to changes in underlying assumptions. Under the baseline scenario, a debt level of 35 percent of GDP or less would be sustainable under the following (see figure):
If the exchange rate depreciated by 30 percent in 2006, the debt level would remain at about 50 percent of GDP in 2010.
A temporary upward shock to the real interest rate, by two standard deviations above its historical average in 2006 and 2007, would raise the debt ratio to 54.3 percent of GDP by 2010.
A combination of permanent slowdown in growth, to 3.5 percent in the medium term and the lowering of the primary fiscal deficit to the 0.5–1 percent of GDP range, would lead to a 2010 debt ratio of 51 percent of GDP.
20. The empirical analysis above suggests that Indonesia’s current public debt level is on the high side and that the government’s medium-term debt target would bring debt to a “safe” level. Indeed, while it is not possible to arrive at precise estimates of the appropriate upper public debt level, and drawbacks exist for each method utilized above, the analysis suggests that Indonesia’s “safe” debt level would be in the 35–42 percent of GDP range or below. This confirms that the current public debt level is on the high side, while also lending support to the government’s 32 percent of GDP target for 2009.
Al-Mashat andMati (2004) “Currency crises in the Middle East North Africa (MENA) Region: What are the Early Warning Signals”? Proceedings of the Conference on “Rising Challenge: International Crisis and Economic Management in Egypt.”
Barro (1979) “On the determination of Public Debt” Journal of Political Economy.
Berg and Patillo (2000) “The Challenges of Predicting Economic Crises” Economic Issues No. 22)International Monetary Fund.
Eaton (1993) “Sovereign Debt: A Primer” World Bank Economic Review Vol. 7 no. 2 pp. 137–172.
Elbadawi et al (1997) “Debt Overhang and Economic Growth in Sub-Saharan Africa” External Finance For Low Income CoIMF institute,Washington DC.
Hemming KellSchimmelpfennig (2003). “ Fiscal Vulnerability and Financial Crisis in Emerging Market Economies” Occasional Papers 218International Monetary Fund.
International Monetary Fund (2002) “Assessing sustainability” (www.imf.org)
__________ (2003) World Economic Outlook, Public Debt in Emerging Markets:Is it too high?Washington DC.
KaminskyLizondo and Reinhart (1999) “The Twin Crises: The Causes of Banking and Balance of Payments Problems” American Economic Review.
__________ (1998) “Leading Indicators of currency crises” IMF Staff PapersInternational Monetary Fund,Washington DC.
Kraay and Vikram Nehru (2003) “When is External Debt Sustainable?” Work Bank.
Manasse Roubini (2005) “Rules of Thumb for Sovereign Debt Crises” IMF Working paper.International Monetary Fund.
__________ and Axel Schimmelpfennig (2003) “Predicting Sovereign Debt Crisis” IMF Working Paper WP/03/221.
PatilloPoirson and Ricci (2002) “External Debt and Growth” IMF Working Paper.International Monetary Fund.
__________ (2003) “What Are The Channels Through Which External Debt Affects Growth?Brookings Trade Forumpp 229–58.
Reinhart Rogoff and Miguel Savastano (2003) “Debt Intolerance” Brookings Papers on Economic Activity.
Prepared by Amine Mati (FAD).
For example, by reducing productive public expenditure or by depressing private investment as a result of higher interest rates and/or expectations of higher taxes in the future.
A notable attempt was made in the 2003 World Economic Outlook study, which presents a comprehensive dataset on public sector debt.
Thresholds found often vary across country samples and time periods and are highly dependent on the specification used.
Only 4 percent of total domestic government bonds are held by foreigners. Multilateral creditors represent about 25 percent of all external debt, with the rest held by Paris Club creditors.
For example the public debt to GDP ratio was 28 percent of GDP in Indonesia, 36 percent of GDP in Thailand, and 45 percent of GDP in Argentina before the onset of the crisis (see 2003 WEO).
Although not always accompanied by debt defaults (Roubini et al.).
Several methods of interpolation were used, including a linear trend, log, and exponential. As shown in Kaminsky and Reinhart, interpolation can be justified given the “persistence” of the debt variable. Data used was from WEO, supplemented by higher-frequency data from Harvath Analytica.
The 24 months prior to the currency crisis itself represent the crisis periods.
Calculated as the number of monthly good signals divided by the total number of months in the crisis period.
The extensive public debt data obtained for the WEO (2003) are on a general government basis for industrial countries, and were constructed as a sum of separate series for external and domestic public debt. Note that this data set is different from the one used in section C, as domestic debt data is mostly central government data (with a longer time series).
OLS, Instrumental variables (using two stage least squares to instrument for endogeneity in the schooling, openness, investment and public debt variables) and fixed effects were all used. Time dummies were included for the crisis period (1997–1998). The estimated effects of variables on growth are all of the expected sign, except openness and debt service to exports. Patillo, Poirson, and Ricci (2002) find similar results, using the same approach.
The effect would still be significant but smaller using instrumental variables.
The differential impact on growth is very sensitive to time dummies used for crisis periods, and does not seem to hold under the fixed-effects specification.
IMF (2003), Von Hagen and Harden (1995). See also Chapter II of this volume.
This is the approach used in the standard debt sustainability analysis included in IMF staff reports.