Assessing Fiscal Risks in Bangladesh

This paper identifies, quantifies, and assesses fiscal risks in Bangladesh. By performing sensitivity analysis and using stochastic simulations, it measures risks arising from shocks to gross domestic product growth, the exchange rate, commodity prices, and interest rates. It also analyzes specific fiscal and institutional risks, including those related to the pension system, issuance of guarantees, state-owned commercial banks, and external borrowing and debt management strategies. The paper finds that fiscal aggregates are particularly sensitive to shocks to commodity prices and the exchange rate. Other factors that could affect fiscal aggregates are the unfunded pension system and limited institutional capacity.


I. INTRODUCTION
Fiscal risks are factors, often outside a government's control, that can cause a country's fiscal aggregates to differ from forecasts. As noted in Cebotari et al. (2009), these differences can be large, and may result from a variety of shocks such as deviations of macroeconomic variables from expectations (e.g. shocks to economic growth, interest rates, exchange rate, and the terms of trade), natural disasters, calls on government guarantees, and also from institutional weaknesses. The 2008-09 global crisis and its aftermath illustrated that the materialization of fiscal risks can lead to significant fiscal liabilities.
This paper assesses fiscal risks in Bangladesh. First, it presents a framework to identify and classify the different sources of fiscal risks in Bangladesh. Second, it assesses the sensitivity of the fiscal balance and the public debt to particular macroeconomic shocks and conducts stochastic analyses of the impact of such shocks on the public debt-to-GDP ratio. Third, it evaluates the impact of specific sources of fiscal risks such as those originating from contingent liabilities and the pension system. Finally, it assesses risks that emerge from the government's institutional capacity limitations, such as budget forecasting errors, external debt management, and data discrepancies. Based on this analysis, the paper also proposes measures to mitigate some of the most severe risks Bangladesh faces.
Results suggest that, in Bangladesh, a variety of factors may cause the fiscal outturns to diverge from forecasts. The fiscal balance is particularly sensitive to shocks to macroeconomic variables such as commodity prices and exchange rates. Additionally, specific factors, such as calls on government guarantees or the recapitalization of stateowned banks could negatively impact fiscal aggregates. Results also highlight the impact of risks derived from the unfunded pension system and the limited institutional capacity.
The paper draws on two strands of the literature: fiscal risks and debt sustainability. Regarding the former, results are consistent with Cebotari et al. (2009), who building on experience from different countries, conclude that macroeconomic shocks and calls on contingent liabilities often have major implications for fiscal sustainability. In addition, Hemming et al. (2006) assess the impact of guarantees and other instruments on debt. They argue that greater use should be made of scenario analysis to stress test debt projections under alternative assumptions about calls on guarantees.
The paper also builds on an extensive literature on debt sustainability and its determinants. 2 When debt rises (in particular the external portion) beyond certain thresholds, a country's fiscal balances become more vulnerable to shocks, as explained by Obstfeld and Rogoff (1996), leading in extreme cases to debt crises. Celasun et al. (2006) study debt sustainability in emerging economies, and find that an explicit quantification of risks could help in designing consolidation strategies. Furthermore, debt sustainability is of particular relevance for low-income countries, given that they generally present high vulnerability to exogenous shocks, political instability, and weak institutions, and their debt structure is usually denominated in foreign currency. 3 The rest of the paper is organized as follows. Section II presents the main framework for analyzing fiscal risks. Section III assesses the impact of macroeconomic risks, by quantifying budget sensitivity to different shocks, and conducting stochastic analyses (fan charts) for the path of public debt. Section IV deals with different contingent and policy-specific risks facing Bangladesh. Section V concludes.

II. A FRAMEWORK FOR ANALYZING FISCAL RISKS
As mentioned above, fiscal risks are the government's exposure to short-to medium-term variability in the levels of revenues, spending, and the values of public assets and liabilities relative to forecast. It is helpful to organize fiscal risks into a framework that differentiates between: (a) general economic risks, such as those arising from shocks to macroeconomic variables (e.g. commodity prices, GDP growth, exchange rates); (b) specific fiscal risks, mainly from contingent liabilities, whether explicit or implicit; and (c) structural or institutional risks, such as weak institutional capacity and spending rigidity. 4 These risks are then assessed vis-à-vis their impact on the budget and the debt stock ( Figure 1).
General economic risks operate through a variety of channels, such as shocks to GDP growth, inflation, the exchange rate, interest rates and commodity prices. These shocks affect expenditures (e.g. through the subsidy bill), revenues, and consequently the stock and dynamics of public debt.
Realizations of contingent liabilities (that is, obligations triggered by an uncertain event), can also create substantial fiscal risks. A contingent liability can be explicit or implicit. In the first case, the conditions are clearly stipulated in policies or legal obligations, while in the second case the obligation arises from the expectation that the government will provide support should a particular event occur. 5 Fiscal risk analysis has traditionally focused on explicit contingent liabilities arising from contractual or legal obligations of the government. However, it is clear that non-contractual commitments are also critical for fiscal sustainability (Cottarelli, 2013), particularly those emanating from the financial sector. A particular feature of implicit contingent liabilities is that their hidden and/or uncertain nature may tempt governments to avoid dealing with them in a timely fashion. However, this may exacerbate the problem when they are eventually realized, as the size of the liabilities may have grown in the meantime.
Structural or institutional weaknesses can also create policy risks and constrain the effectiveness of fiscal risk management. Coordination problems between different levels of government can impede the government's ability to implement the desired fiscal policy or hamper the ability to respond to shocks. Limited capacity to identify and manage fiscal risks can exacerbate a country's exposure to existing fiscal risks. When policymakers lack good information, fiscal management becomes more difficult, increasing the likelihood of policy errors. As noted by Budina and Petrie (2013), this situation can be compounded if the institutions and actors responsible for the specific risk management functions are not clearly identified, if those responsible lack the necessary authority, or if budgeting systems undermine effective management.

Figure 1. Types of Fiscal Risks
The framework outlined above will guide the identification in this paper of fiscal risks in Bangladesh.

Types of Fiscal Risks
General Economic Risks Related to general forecasting parameters (e.g. macroeconomic shocks)

Structural or Institutional Risks
Constrain the effectiveness of fiscal risk management (e.g. weak capacity; spending rigidity; revenue earmarking)

Specific Fiscal Risks
Usually unrelated to general economic forecasting parameters. Give rise or contribute to uncertainty about specific fiscal outcomes.

Other Risks
Arise from specific, identifiable sources, but are not fully predictable (e.g. changes in values of assets and liabilities)

Contingent Liabilities
Arise solely from the occurrence of a specific event or condition

Implicit
Based on expectations or pressures to provide support (e.g. failing publicprivate partnerships, bailouts of state-owned enterprises)

Explicit
Based on clear and firm legal obligation or a declared policy (e.g. government guarantees)

III. MACROECONOMIC RISKS AND BUDGET SENSITIVITY IN BANGLADESH
Macroeconomic shocks (e.g. shocks to GDP growth, commodity prices, and interest rates) can be a source of significant risks to both the budget at any point in time and the evolution of public debt in Bangladesh. This section assesses the sensitivity of the fiscal balance and public debt to particular macroeconomic shocks and conducts stochastic analyses of the impact of such shocks on the public debt-to-GDP ratio. 6

A. Sensitivity Analysis
Bangladesh's fiscal aggregates are sensitive to variations in a number of macroeconomic variables, including commodity prices, exchange rates, interest rates, and GDP growth. Shocks to these variables thus impact on fiscal performance, and some of these variables have been particularly volatile in the past few years.
This section examines the impact on fiscal outcomes of changes in the forecast values of key variables. The analysis focuses on one standard deviation permanent shocks to commodity prices (oil and urea), the exchange rate, the domestic interest rate, and GDP growth introduced one at a time ( Figure 2). 7 The shocks are assumed to take place from the start of FY14. The near-and medium-term effects of the shocks are illustrated though their impact on the overall fiscal balance and total public debt (deviations from baseline) in Table 1.

Figure 2. Shocks to Macroeconomic Variables
Source: IMF Staff calculations 6 The analysis uses FY96-base GDP. It is to be noted that the Bangladesh authorities have started publishing a rebased GDP series, with FY06 as base year. Nominal GDP in FY13 was about 16 percent higher under the rebased series compared to FY96-base GDP. 7 Permanent shocks are defined as permanent deviations with respect to the baseline.

Commodity prices
Bangladesh's fiscal position is sensitive to certain commodity prices; in particular oil and urea, which tend to move together ( Figure 3) and whose volatility has recently increased. 8 Shocks to these prices operate through both the revenue and expenditure sides. On the revenue side, an increase in commodity prices would result in a rise in import-related tax revenues, which in total account for over 30 percent of tax collections. 9 On the expenditure side, the same shock would translate into an increase in the subsidy bill; in particular, payments related to fertilizers (urea) and fuel subsidies, such as those to Bangladesh Chemical Industry Corporation (BCIC) and Bangladesh Petroleum Corporation (BPC). 10 Consumption of fuel and urea is subsidized in Bangladesh. 11 In FY13 total subsidies were around 3.1 percent of GDP, of which energy related subsidies reached roughly 1.7 percent of GDP, and fertilizer subsidies around 1 percent of GDP.
The impact on revenues, due to the rise in import-related tax collections, is not enough to offset the much larger effect on expenditures, and therefore the overall effect is negative. The analysis suggest that a one standard deviation increase in the prices of oil and urea (roughly a 30 percent price increase) would reduce the overall fiscal balance (that is, increase the fiscal deficit) by around 0.6 percent of GDP on average each year with respect to the baseline. It would also lead to a cumulative increase in the stock of debt of around 3.1 percent of GDP over five years with respect to the baseline. 8 Urea is used as a basic input in the production of rice fertilizers. 9 For simplicity, this analysis assumes zero elasticity of commodity import volumes with respect to prices. 10 The analysis here assumes that the authorities do not adjust retail energy or fertilizer prices, so the fiscal balance absorbs the entirety of the shock. This is clearly a worst-case scenario. 11 For the purposes of this analysis, it is assumed that shocks to oil prices are transmitted one to one to international fuel prices. ©International Monetary Fund. Not for Redistribution

Exchange rate
While the exchange rate has been very stable in the past few years in Bangladesh, a shock to the exchange rate would affect the fiscal balance and debt stock through a variety of channels. 12 A depreciation in the Taka/US$ exchange rate has an impact on domestic prices, and through them on nominal revenues and expenditures. Beyond that, depreciation has a direct impact on both revenues and expenditures. In the case of revenues, the impact is associated with import-related taxes. On the spending side, the main items affected are: (1) the fertilizer subsidy bill; (2) payments to BPC for oil imports (constant volumes assumed); (3) the externally financed portion of the Annual Development Program (capital spending); and (4) interest payments on external debt. 13 Additionally, there's a valuation effect on external debt: the nominal Taka equivalent value of public debt denominated in foreign currency would move at a one-to-one rate with exchange rate changes.
Results show that a permanent 10 percent depreciation in the Taka/US$ exchange rate would reduce the overall fiscal balance (that is, increase the deficit) by around 0.8 percent of GDP on average yearly with respect to the baseline and increase the stock of debt by around 6.6 percentage points of GDP over five years.

Domestic interest rate 14
Interest expenses are a small share of total fiscal expenditures in Bangladesh. Therefore, shocks to interest rates have a limited impact: a one standard deviation rise in domestic interest rates (130 basis points) would reduce the overall fiscal balance by around 0.3 percent of GDP with respect to the baseline and push up the stock of debt by around 1.3 percentage points of GDP over five years.

GDP growth
In terms of its direct impact, economic growth mainly affects the revenue side of fiscal aggregates in Bangladesh, including VAT (import and domestic), import tax, supplementary duties, and income tax. As is standard in studies for other developing and emerging market countries, and following IMF (2009), this paper assumes the elasticity of revenue with respect to growth to be equal to one and the elasticity of expenditure with respect to growth to be equal to zero. 15 Results show that a one standard deviation decline in GDP growth (around 0.7 percentage points) would reduce the overall fiscal balance by around 0.1 percent of GDP with respect to the baseline and push up the stock of debt by around 0.5 percent of GDP over five years.
The relatively small effect is the reflection of two factors: (a) a small tax base: tax revenue collection in Bangladesh is among the lowest in the world, at around 9 percent of GDP; and (b) the low volatility of growth in the past few years, which implies that shocks to growth measured in terms of one standard deviation are small. Of course, the tail event of a larger, more sustained shock to growth would produce a larger deterioration in fiscal aggregates.

B. Stochastic Analysis of Debt Dynamics
In some cases, macroeconomic shocks do not hit an economy in isolation but occur simultaneously. In crisis episodes (tail events), a negative shock to real GDP may occur in parallel with a shock to the exchange rate, interest rates, and inflation. The cumulative impact of such shocks on public debt may be significant.
Using fan chart analysis, this section illustrates the frequency distribution of projected public debt-to-GDP ratio paths generated by shocks to key macroeconomic variables. Fan charts are a tool to depict the possible evolution of the public debt ratio over the medium term and to visually assess fiscal risks from macroeconomic shocks. Sample statistics based on historical data (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012) for the real GDP growth rate, the effective real interest rate on government debt, the primary balance, and the real exchange rate are used to generate the sample means and the variance-covariance matrix that defines a joint normal distribution of these macroeconomic variables. Draws for each one of the variables from the joint normal distribution are used to generate the shocks--calculated as the value drawn minus the sample mean--that are applied to the baseline projections for each of the macroeconomic variables. These "shocked" series of macroeconomic variables are then introduced into the debt dynamics equation to calculate a distribution of projected debt paths (see Annex 1 for details on the derivation).
The results suggest that Bangladesh remains at a low risk of debt distress. After a combined shock to key macroeconomic variables, there is a 50 percent probability that the debt-to-GDP

Scenario C (130 basis points increase in domestic interest rate, or 1SD)
ratio remain between 36 and 40 percent (Figure 4, left-hand side). If the draws wererestricted to only negative shocks (e.g., only draws of negative primary balance; see Figure 4, righthand side), then the probability of higher debt levels would increase. Even under these assumptions, debt levels would remain below reasonable thresholds, such as the 42 percent of GDP average low-income country debt in 2012 (for low-income Asian countries, the average for 2012 was 48 percent).

IV. SPECIFIC FISCAL RISKS FACING BANGLADESH
Fiscal risks in Bangladesh do not only arise from disturbances to general economic variables: they also arise from specific sources such as the realization of contingent liabilities. This section assesses the impact on fiscal aggregates of the hypothetical realization of all government loan guarantees and contingent liabilities from the state-owned banks. It also examines the potential long-term impact from the unfunded pension system.

A. Government Loan Guarantees
The government of Bangladesh customarily provides guarantees against loans contracted by the different state-owned financial and non-financial enterprises. Most loans finance the implementation of diverse public policies and programs. If the contracting organization fails to pay the loan in time, the guarantees are invoked and the liabilities for payment are passed on to the government. Consequently, these guarantees could eventually turn into outright government debt.
The stock of government guarantees issued before FY04 was mainly related to agricultural programs. From that year until FY12, the issuance of guarantees was very small and still related to agricultural credit. In FY12 there was a steep increase in guarantees, mainly those provided to state-owned commercial banks (SOCBs) for lending to non-financial public enterprises, in particular BPC. As a result, the stock of government guaranteed debt (both external and domestic) rose from around 3.5 percent of GDP in FY04 to 5.7 percent of GDP (Taka 592 billion) by end-June 2013 (see Table 2), of which guarantees provided to SOCBs represented around 30 percent of the total.  Risks emanating from government guarantees are sizable. Should they materialize in full, they could noticeably increase Bangladesh's public debt.

B. State-Owned Banks
The weak balance sheets of state-owned banks (SOBs) represent a tangible fiscal risk (contingent liability) for the Government of Bangladesh. There are eight SOBs in Bangladesh, comprising 4 commercial banks and 4 specialized banks (development banks). As a whole, they represent around 32 percent of banking system's assets, or roughly 24 percent of GDP ( Figure 5). These banks account for the majority of the outstanding nonperforming loans (NPLs) in the banking sector. It is to be noted that lending to state-owned enterprises, even to loss-making ones, does not give rise to NPLs, as nearly all of these loans are guaranteed by the government.  The SOBs have come under renewed stress since 2012, reflecting different factors such as a slowdown in economic activity, increasing competition, and weak internal governance (recent financial frauds highlighted significant weaknesses in oversight, internal controls, and risk management in SOBs). As of end-2013, the capital shortfall at these banks, compared to the regulatory minimum, stood at about 2.5 percent of GDP. 16

C. The Pension System
There are two potential sources of fiscal risks arising from Bangladesh's current pension arrangements. First, there are those associated with the civil servants' retirement scheme and the General Provident Fund. A more hypothetical long-term risk arises from potential pressures from the absence of an organized pension system for workers in the private sector, whether formal or informal.

Civil Servant Retirement Scheme
As in other South Asian countries, the government of Bangladesh provides its own employees with a noncontributory, defined benefit pension, including survivor benefits. Civil servants are eligible to receive a pension at the age of 59. 17 Pensions depend on the length of the public service. The civil servants' salary structure is divided into 20 grades or categories, with the basic salary ranging between Tk 5,000 and Tk 40,000 and an average of Tk 20,000. After 25 years of service (or at the age of 59) a civil servant is entitled to a pension of 80 percent of his/her prorated last basic salary (with proration based on years of service if less than 25), half of it as pension payment every month and the other half in a lump sum.
Pension spending on the Civil Servant Retirement Scheme (CSRS) is captured in fiscal aggregates under current expenditures. In FY13 the government assigned Tk 60 billion to the pension bill (0.57 percent of GDP).
Bangladesh currently employs roughly 1.2 million civil servants, of which around 35,000-40,000 retire every year. However, demographic trends will drive up the number of retirees per year, with an impact on pension expenditures. Current United Nations projections estimate that the elderly (individuals aged 60 years and above) will more than triple as share of Bangladesh's total population by 2050, from the current 6 percent ( Figure 6). And, as the figure shows, the increase in the ratio of elderly population to working-age population (known as the old-age dependency ratio) is even more dramatic.

Figure 6. Actual and Projected Population Over 60
Source: United Nations, World Population Prospects.
To estimate the potential fiscal impact (via spending on the CSRS) from expected changes in demographics, it is helpful to decompose pension spending to GDP into three factors: The first term is the old-age dependency ratio. The second term is the benefit ratio, defined as the ratio spending per pensioner to GDP per worker, which provides a measure of the generosity of pension benefits. Absent any changes in the benefits formula, this ratio is assumed to remain constant at its value as of end-2013 (about 1.32). And the final term is the eligibility ratio, defined as the ratio of the number of individuals receiving a pension to the population age 60 and older, which provides a measure of pension system coverage. This is assumed to be constant at 0.04 (civil service pensioners were 4 percent of the elderly population at end-2013), under the assumption that the covered population (in this case civil servants) and eligibility condition for a pension (such as the retirement age or minimum years of service) will not change over time.
Based on these parameters, pension spending is projected to increase from 0.5 percent of GDP today to 0.9 percent of GDP in 2030, and to around 1.8 percent of GDP in 2050 in line with the expected acceleration of aging after 2030 (Table 3). old-age dependency ratio benefit ratio eligibility ratio

General Provident Fund
In addition to the CSRS, there is the General Provident Fund (GPF) for civil servants, which is a mandatory, defined contribution system, in which civil servants contribute a minimum of 10 percent of their salaries (there is no upper limit). The notional accounts accrue interest of around 12 percent of the GPF stock at year end. When civil servants retire, they can withdraw the whole amount plus interest. Before retiring, at any point in time, civil servants can borrow up to 80 percent of their cumulative contributions from the fund. As of end-FY13, the GPF stock of contributions amounted to Tk 204 bn, plus Tk 24 bn in interest (roughly 2.2 percent of GDP). Unfortunately, despite its name, the GPF is unfunded: the cash flows it generates are not being saved, but used to finance the deficit. 18 Indeed, GPF is currently generating sizable surpluses (i.e. contributions to the fund minus withdrawals) of around Tk 30billion. However, as civil servants age and start to retire in larger numbers, the net cashflows may become negative, posing a clear financial risk. 19

Potential pressures from the absence of pension coverage for the private sector
Bangladesh does not have a formal pension program for the vast majority of the population. First, most of the workforce (an estimated 89 percent of the total, and an even higher proportion for women) is employed in the informal sector, mainly in agriculture. 20 Also, other than a gratuity benefit at retirement, employees of formal private sector firms do not have access to any formal old-age benefits program. Overall, only around four percent of the population over 60 is covered by the pension system in Bangladesh. The rest rely on their own savings to sustain themselves in retirement. 18 For more details, see Alam (2012 The absence of a formal pension scheme for most of the population in Bangladesh might eventually lead to pressures on the government to provide some minimum pension. As a way to illustrate the potential costs involved it would be useful to estimate the costs of setting up a universal scheme. The best way to do this is to draw from international experience. A number of countries -both developed and developing -have put in place universal pension schemes (See Table 4). These pension schemes can be affordable, simple to administer and have been successful in tackling old age poverty (Willmore, 2007). To illustrate how much it would cost to institute a universal pension scheme in Bangladesh, two key parameters need to be taken into consideration:  Age of eligibility (the age at which people get entitled to the pension; the higher the age, the lower the overall cost of the scheme). The illustrative exercise below considers the costs of universal coverage system under different eligibility ages (over-60, over-65, over-70, and over-75). The number of potential beneficiaries, using 2012 population estimates, ranges from 2.1 to 9.9 million. 21  Size of grant, which is the amount provided to beneficiaries. It is common to use the poverty line as a benchmark. In Bangladesh, the poverty line was calculated in 2005 at Tk 861.6 per month. Applying the CPI inflation rate, that poverty line translates into roughly Tk 1,500 per month by end-FY13. As shown in Table 6 a universal pension scheme that provides such an amount would cost between 0.4 and 1.9 percent of GDP, depending on the age threshold. Of course, the fiscal cost of a universal pension will increase over time as the population ages.
Assuming a starting poverty line of Tk 1,500, a constant inflation rate of 6 percent (equal to the average for the last 20 years) and a nominal GDP growth of 12 percent, Figure 5 shows the fiscal cost of the universal scheme by age of eligibility (Figure 7.A). Alternatively, it is possible that the poverty line increases faster than inflation over the long term, as the basic needs basket widens with development. Figure 7.B shows that path, allowing the pension per capita to grow in line with GDP per capita. Since the qualifying population is expected to grow as a share of total population, then the total pension spending would grow as a share of GDP. For the most expensive case (60 and over), the fiscal costs would be almost at 7 percent of GDP in the long run.
As stated in previous paragraphs, the costs of different universal pension schemes vary from 1 to 7 percent of GDP in the medium term. To contain these costs, the literature generally suggests: (a) that these schemes should be means-tested to target only the needy; and (b) the design of such programs should aim for benefits that are sufficient to alleviate poverty but low enough to minimize incentives to remain outside of the formal pension system. 22

V. INSTITUTIONAL CAPACITY
Risks to the budget and the public debt also emerge from the government's institutional capacity. This section focuses on three specific areas that may pose risks to fiscal aggregates in Bangladesh: budgeting practices, external debt management, and data discrepancies.

A. Budgeting Practices and Forecasting
Significant deviations in outturns vis-à-vis budget figures have systematically been observed in recent years in Bangladesh. Consistently, both revenue and expenditure outturns have fallen behind budget target numbers. On average, during the last four years total revenues were below budget target by around 4 percent (0.5 percent of GDP). The highest difference has been in non-tax revenue, with an average deviation of 16 percent (Table 6). Similarly, expenditure outturns fell behind by 8.5 percent vis-à-vis the budget, or around 1.4 percent of GDP. The main driver has been underexecution in capital spending, falling on average around 19 percent below target (Table 6). The main problem associated with this pattern is that while revenue forecasts in a budget document are merely projections, the expenditure allocations are legal spending authorizations. Thus, if revenues fail to materialize, there is a risk that line ministries may still execute in full their spending envelopes, leading to larger-than-expected fiscal deficits and financing needs.

B. External Borrowing and Debt Management
Efficient debt management strategies are important to mitigate the effects of shocks to fiscal aggregates (i.e. macroeconomic shocks and contingent liabilities) and keep borrowing under control. This is particularly true of external debt, which is more likely suffer from shocks to the exchange rate or international interest rates. While Bangladesh's total public debt remains below 40 percent of GDP, there has been a fast increase in non-concessional external borrowing: the annual average external debt disbursement in FY12 and FY13 is around 180 percent higher than the annual average for the period FY05-FY11 (Figure 9).
Bangladesh's government has taken significant steps toward improving the monitoring and contracting of external debt, including through the creation of a Technical Committee on  Nonconcessional Borrowing. Continued efforts to strengthen the assessment, approval, and monitoring of external loan contracts and guarantees are needed.

Figure 9. External Debt Disbursement and Amortization (USD millions)
Note: Borrowing by SOEs supported by government guarantees not included. Sources: Bangladesh Authorities and IMF Staff calculations.

Data discrepancies
Problems associated with fiscal data quality and timeliness may also pose fiscal risks. One significant example is the discrepancy between revenue collection data provided by the National Board of Revenue (NBR) and that provided by the Office of the Controller General of Account (CGA). Part of this discrepancy reflects a timing issue. Taxes are registered by the NBR when they are effectively paid, but they are only booked by the CGA when the amount is deposited into the Treasury Single Account (TSA). If the definition of revenue is exactly the same and the only difference was one of timing, at year-end the numbers should be reconciled. However, this is not the case, and the gap between the two reported series is increasing. As shown in Figure 10 the cumulative gap between NBR and CGA over the period FY12-FY13 was roughly Tk 90 billion (almost one percent of GDP), with CGA typically well below the NBR data.
These inconsistencies produce uncertainty for fiscal policy-making and undermine transparency and accountability.

VI. CONCLUSIONS AND POLICY IMPLICATIONS
Several factors have the potential to drive actual fiscal aggregates away from projections in Bangladesh. These include but are not restricted to, macroeconomic shocks, contingent liabilities, and institutional weaknesses. This section summarizes the paper's key findings and draws policy implications.
The analysis in this paper suggests that the fiscal balance in Bangladesh is sensitive to macroeconomic shocks, in particular those to commodity prices and exchange rates. A one standard deviation increase in commodity prices, or a 30 percent devaluation, may raise the deficit by between 0.6 to almost 1 percent of GDP on average per year when compared to the baseline.
Specific factors, such as calls on government guarantees or increased recapitalization needs of state-owned banks, could also have a significant negative impact. Should they materialize in full, calls on government guarantees and further recapitalization needs could certainly add pressure to the budget and increase Bangladesh's Public debt.
In addition to the most immediate risks of shocks to macroeconomic variables and calls on contingent liabilities, risks arising from the civil servants' pension scheme and the General Provident Fund could materialize in the medium to the long-run. If no changes were made to the system, the fiscal cost of the unfunded pension scheme could increase from 0.5 percent of GDP to about 2 percent of GDP by 2050. Furthermore, if a universal pension system were to be implemented (only four percent of the old age population is covered by the current system), the fiscal cost would raise up to 6 percent of GDP per year, or even higher.
Finally, risks derived from the government's institutional capacity could also take a toll on Bangladesh's fiscal aggregates. Risks for instance emerge from budget practices, the management of external debt, and data discrepancies. Bangladesh has a tradition of overstating expected revenues and expenditures in the budget, which could lead to excessive spending pressures in the short run. Weaknesses in debt management could lead to riskier debt structures, while data discrepancies produce uncertainty for fiscal policy-making and undermine transparency and accountability.
Policies that could help mitigate the incidence and impact of fiscal risks could include:  Full integration of risks into government policy decision-making, both in fiscal management and in the design of an integrated asset and liability management strategy in coordination with Bangladesh Bank. 23  Building government capacities to analyze and measure fiscal risks. 24 In order to achieve this, a system of treasury cash flow forecasts should be implemented. Even though there are attemps to do so, no formal mechanism is in place yet.
 Improvements regarding debt records and reporting. A key step towards this objective would be the modernization of the National Saving Directorates by linking of the interest rates on the instrument to market or benchmark securities and though making issuance and record keeping electronic.


Measures to reduce currency risks in the government liability structure. For example, a cap in the amount of foreign denominated debt as well as on foreign denominated government guarantees.


Publication of a Debt Statistical Bulletin for the total central government debt (external, domestic, and loan guarantees).
 A full set of policies and procedures for issuance of loan guarantees, as well as prioritization and limitation on the amounts of new guaranteed obligations.
 Implementation of a contributory pension scheme for civil servants to replace the current non-contributory regime, and reforms to the GPF via the creation of notional accounts and of an investment fund to accumulate the system's assets. Consideration could also be given to institutionalizing a non-contributory pension regime for the poor, as existing transfer mechanisms to the elderly poor are very low. Additionally, Bangladesh could aim to develop a voluntary defined contribution retirement scheme for all adults regardless of their employment status. These schemes are important sources of long-term investment funds in the domestic financial markets in developed and developing economies. 23 The current fragmentation among debt management entities adds costs to any planning strategy by the Ministry of Finance and BB. 24 The evidence suggests that the introduction of fiscal rules and the setting up of independent fiscal councils to monitor fiscal developments can help reduce fiscal risks (Debrun et al., 2009). at the end of period t+1.