Unlocking the flow of finance for climate adaptation: estimates of ‘Fiscal Space’ in climate-vulnerable developing countries

ABSTRACT We study the availability of fiscal space in climate-vulnerable developing countries. These countries require urgent climate adaptation and transition investments. However, their governments describe being bypassed for international financial support due to ‘limited fiscal space.’ We suspect that many governments are not close to a point of long-term insolvency but are unable to maneuver fiscally because of what has been called a ‘financial death trap.’ We apply a measure of fiscal space based on an endogenous debt limit reflective of a country’s record of fiscal adjustment consistent with long-term solvency. We find that for many countries, the distance between the endogenous debt limit and forecast public debt ratios – i.e. fiscal space – is fairly ample. Our findings imply that climate-vulnerable countries should be afforded a second look by international financial institutions using a long-term lens, of which this measure of fiscal space is an example. By illuminating the difference between long-term insolvency and short-term liquidity crises, the endogenous debt limit measure could be part of a multi-pronged strategy to unlock greater flows of adaptation finance. It could lower the cost of capital or be useful in the efficient allocation of adaptation financing among countries given current shortfalls. Actions to obviate the financial death trap are also warranted. Climate ambitions will be derailed if otherwise solvent and able governments are unable to access finance for urgent climate adaptation investments. Key Policy Insights Fiscal space, the distance from projected debt ratios to country-specific debt limits beyond which long-term solvency fails, is estimated to be fairly ample in many climate-vulnerable developing countries. A clear understanding of fiscal space in climate-vulnerable developing countries could help unlock greater flows of adaptation finance. Debt thresholds in IMF debt sustainability frameworks should not be confused with limits to fiscal space per se. Doing so could cause otherwise solvent and able governments to be caught in a financial death trap. Delayed or foregone climate adaptation and transition investments, and delay of their mitigation co-benefits, due to a financial death trap warrant action to overhaul the global financial system.


Introduction
Climate-vulnerable developing countries are the most threatened by the physical impacts of climate change. They have little choice but to immediately undertake strategic adaptation, resilience and transition investments at a 'pace and scale that climate science requires' to survive and thrive (Bhattacharya et al., 2020). However, members of the Vulnerable Twenty (V20) Group of Finance Ministers of the Climate Vulnerable Forum -'a dedicated cooperation initiative of economies systematically vulnerable to climate change,' which represent 1.2 billion people and US$ 2.3 trillion in GDP in countries that account for only 5% of global emissionsdescribe how efforts to secure funding are impeded by extreme resource constraints and limited fiscal space, on the one hand, and the dearth of suitable climate finance, on the other (V20, 2021a and 2021b). They report high and rising levels of public debt, driven by recurring climate disasters and disproportionately high sovereign borrowing costs, which reflect premia for physical climate risks (V20 2021b;Beirne et al., 2021;UNEP et al., 2018;Cevik and Jalles, 2020). High debt servicing requirements crowd out urgent crises and recovery and resilience spending, worsening adaptation and sustainable development prospects further. Yet, only 24% of the climate finance mobilized by developed countries between 2016 and 2020 has been for adaptation, and just 25% of public climate finance has been in grant form (OECD, 2022). 1 Global funds for adaptation are, moreover, just 8.4 to 17.9% of the estimated US$140 to 300 billion needed annually by developing countries (UNEP, 2022). Thus, the V20 has called for debt relief and restructuring, alongside an implementation plan for doubling international finance for adaptation and other related actions (V20, 2022). Public grant-based finance is particularly important since adaptation investments are less likely to feature adequate revenue streams and 'the most vulnerable communities are more likely to already be overly indebted. ' 2 The account that climate-vulnerable countries are bypassed for international financial support due to 'limited fiscal space' is our concern. 'Fiscal space' is the scope for further increases in public debt without undermining sustainability. Public debt sustainability is theoretically equated with government solvency, i.e. the ability of the public sector to honour all future financial obligations (Ostry et al., 2010;Tanner, 2013). Solvency is a medium-to-long-term concept and should be distinguished from liquidity, a short-term constraint: an entity is (il)liquid if, regardless of whether it satisfies the solvency condition, its liquid assets and available financing are (in)sufficient to meet or roll over its maturing liabilities (IMF, 2014). However, the International Monetary Fund's debt sustainability frameworks (henceforth, DSF) combine the concepts of solvency and liquidity 'without drawing a sharp distinction between the two' (IMF, 2002). They feature debt thresholds that are commonly interpreted as limits to fiscal space per se, although the IMF explicitly states they should not be (IMF, 2011). DSFs are articulated for both market access countries (MACs), which typically have significant access to international capital markets, and low-income countries (LICs), which rely primarily on concessional finance for external financing needs.
Whether a government is close to a point where fiscal solvency would be in question, or whether it is solvent but finds itself lacking in liquidity and room for fiscal maneuver, is a critical distinction. Sachs (2021) calls the latter a 'financial death trap,' where 'developing-country governments are pushed into defaultnot out of bad faith or because of long-term insolvency, but for lack of cash on hand.' It occurs when a government experiences a 'sudden stop' to new lending because it has debt falling due that requires refinancing; credit rating agencies or international lenders have decided, 'often for some arbitrary reason,' that the government is no longer creditworthy (ibid). If the flow of finance to otherwise solvent governments, which pursue climate-resilient, low-carbon sustainable development, is impeded in this manner, the Paris Agreement itself is undermined.
The interest rate-growth differentials (IRGDs) of countries, or the average interest paid on government debt less the growth rate of the national economy, are an important consideration. Negative IRGDs have prevailed in emerging markets and developing countries for decades , and are expected to persist in many V20 countries for at least the next five years. 3 Negative IRGDs could indicate that, ex-ante, 'fiscal space' may be wider than acknowledged. This is because when output growth is greater than interest rates on government debt, the debt-to-GDP ratio can be stabilized or even decline without governments having to generate primary surpluses (i.e. revenues that are greater than non-interest expenditures) (Blanchard et al., 1990). Supplement A provides a brief explanation.
Information on the extent of fiscal space in climate-vulnerable developing countries could be a vital component of a multi-pronged strategy to unlock greater flows of adaptation finance. The information could help mitigate the physical climate risk premia on sovereign debt, lowering the cost of capital across the economy, which could help mobilize private funds without necessarily increasing sovereign debt burden. It could help crowd-in private investments for mitigation, thereby freeing up more public climate finance grants for adaptation. 4 It could also be useful in the efficient allocation of scarce adaptation grants and loans among countries, including for adaptation investment planning. For example, countries with less fiscal space and less capacity for investment planning could be considered for more grant-based adaptation planning. 5 Strong technical capacity for adaptation planning is crucial in fact: well-chosen adaptation investments, with large multiplier effects and sustainable development dividends, can have a positive impact on growth, potentially helping an economy outgrow its debt or preventing growth from being derailed by climate change. 6 On the other hand, the lack of clarity on priority investments (e.g. costs, timeframes, integration in national spending plans), and high up-front costs and risks (e.g. regulatory, technical), deter initiatives and impede private finance for climate-aligned investments (Hourcade et al., 2021;Prasad et al., 2022). Other components of a multi-pronged strategy include, among others, debt-for-climate swaps, debt suspension and cancellation, multi-sovereign guarantees and national policy integration and investment planning (e.g. for climate, SDGs, and pandemic recovery) (Hourcade et al., 2021).
Our objective is to estimate whether and to what extent fiscal space exists in climate-vulnerable developing countries. We use the approach of Ostry et al. (2010) and Ghosh et al. (2013a) (henceforth Ostry/Ghosh) which entails the estimation of a country-specific 'debt limit' implied by its historical record of fiscal adjustment. This debt limit is understood as the point beyond which debt can grow without boundslong-term solvency failsabstracting from liquidity/rollover risks. 'Fiscal space' is the difference between current levels of public debt and estimated debt limits.
Our choice of approach is deliberate. First, as earlier mentioned, IMF DSF thresholds already incorporate liquidity concerns in a significant way. Second, any difference between DSF thresholds and estimated debt limits may be a gauge of the opportunity cost of liquidity constraints due to a financial death trap. Third, the model's use of a country's historical record of fiscal behavior to draw 'implications about the sustainability of public debt positions at the present time' is appealing (Ostry, et al.). This information could prompt international funders and credit rating agencies to look beyond short-term liquidity issues and consider longerterm climate-aligned growth potential in making financing decisions. It may also help support a change of Basel III liquidity guidelines that, among others, 'discourage investing in long-term assets even with longterm negative interest rates of AAA countries' (Hourcade et al., 2021). At the very least, the information could facilitate a deeper reckoning within the international financial system about how to prevent financial death traps. Sachs (2021) has a number of suggestions in this regard.
In the next section, we describe our methodology and model. The third section is a discussion of our estimation results. The fourth section concludes.

Methods and data
Ostry/Ghosh posit that governments exhibit fiscal fatigue, whereby a government's ability to increase primary balancesthe difference between a government's revenue and its non-interest expenditurecannot keep pace with rising debt. Put another way, it becomes increasingly difficult for a government to raise taxes or cut non-interest expenditures to cover higher interest payments. Thus, after exhibiting little (or even decreased) sensitivity to rising debt (while debt levels are low), primary balances first increase and then slow down as debt risesand possibly decrease at very high debt levels.
Ostry/Ghosh show that fiscal fatigue is a sufficient condition for an endogenous debt limit. Above this limit, and without an 'extraordinary' fiscal adjustmentone that would break with its historical reaction functionpublic debt would increase without bound because the primary balance would never be enough to offset a growing debt service. 7 Their model is illustrated in Supplement B and involves finding points of intersection between an estimated fiscal reaction function and, in a deterministic example, a 'risk-free' interest payment schedule. The lower intersection yields d*, the 'conditionally stable' long-run average debt ratio to which the economy normally converges; the upper (or right-most) intersection yields d**, our debt limit of interest. In the presence of an endogenous response of the interest rate to rising risk, the intersection of the fiscal reaction function and the interest payment schedule yields d ls , which is the debt limit in the stochastic case. d ls is necessarily lower than d** and the available fiscal space implied by it, smaller.
Ostry/Ghosh confirm fiscal fatigue in 23 advanced economies ( Figure 1) and obtain country-specific debt limits in the deterministic case ranging from 152.3% to 263.2% of GDP. The debt limits in the stochastic case are lower by an average of nine percent.
Importantly, Ostry/Ghosh emphasize that 'since behavior can change, a finding of no or little fiscal space is not a prediction that public debt will explode or that the government will default.' Rather the finding indicates that 'fiscal policy may need to react more strongly to debt than past behavior would suggest'. Estimated debt limits are also not to be construed as 'desirable' or 'optimal' levels of debt. In fact, since liquidity risks are not trivial in practice, 'governments will want to ensure that they do not exhaust their fiscal space and that debt remains well below its calculated limit.' 8 Moody's Analytics has adopted the Ostry/Ghosh approach to fiscal space in its assessment of fiscal risks in advanced economies (Ostry et al., 2015). The IMF does not, however, and instead, has designed a new framework for providing 'bottom line' assessments of fiscal space for MACs. This new framework continues to emphasize the availability of market funding rather than long-term solvency. 9 It is not considered to be appropriate for low-income countries (LICs) however, except for those countries which can obtain a significant amount of nonconcessional financing. Instead, the insights from LIC debt sustainability assessments are viewed as adequate (IMF, 2018a).

Empirical implementation
We estimate fiscal space for a sample of climate-vulnerable countries in three steps following Ostry/Ghosh. First, we estimate a reduced form fiscal reaction function, designed to capture a possible nonlinear response of the primary balance to lagged gross debt. Specifically, we estimate the cubic function: where pb i, t is the ratio of the primary balance (budgetary central government) to GDP in country i at time t; d i, t-1 is lagged general government gross debt; X i,t is a vector of macroeconomic and other determinants of the primary balance (drawing from Ostry/Ghosh; Abiad and Ostry, 2005;and Mendoza and Ostry, 2008); 10 α i are country fixed effects, and ε i,t are error terms which are corrected for within-country first-order autocorrelation.
Next, we compute country-specific debt limits, d**, using the estimated coefficients from equation (1) and country-specific IRGDs (see Supplement B, equation 2). We use three IRGDs for each country, obtaining debt  Ghosh et al., 2013a, p. F13 limits in three deterministic scenarios. The first IRGD is the historical average (from 2009 to 2019) of the effective interest rates on government debt relative to GDP growth rates. 11 The second is the projected average (from 2021 to 2025) of effective interest rates relative to GDP growth rates, sourced from IMF reports. The third is the second IRGD shocked (thus, 'projected, shocked'), whereby projected effective interest rates are increased by one standard deviation and projected GDP growth rates decreased by one standard deviation.
Finally, we compute fiscal space as the difference between our estimated d** and IMF-projected debt for 2025. There is always uncertainty around point estimates such as these however, thus we also calculate fiscal space in terms of probabilities (e.g. the probability that additional fiscal space at a given IRGD is greater than, say, 50% of GDP) using coefficient estimates obtained from Monte Carlo simulations. We also compute fiscal space in the stochastic case by assuming that d ls is lower than the estimated projectedshocked d** by 20%, which is a conservative assumption. It is more than double the reduction from d** observed by Ghosh et al. (2013) for their sample of advanced economies. Put another way, we assume that the increase in risk premiums as debt approaches its limit is greater among climate-vulnerable countries than among the advanced economies studied in Ghosh et al. (2013).
Our sample consists of members of V20 and ASEAN. ASEAN is included since six of its members are among the 15 countries most affected by extreme climate from 2000 to 2019 (Eckstein et al., 2021). There are 48 V20 and 10 ASEAN members, with 3 overlaps, or a total of 55. Of the 55, spotty data eliminated 8 outright (Comoros, Gambia, Haiti, Niger, South Sudan, Tuvalu, Yemen, Brunei) and 6 during modelling (Afghanistan, Grenada, Malawi, Papua New Guinea, Palau, Sta. Lucia). Three more were considered outliers (Sudan, Barbados, Lebanon), leaving 38 in our estimations. The 38 consists of 23 low-income countries (LIC) and 15 market access countries (MAC), which we study separately using data from 1990 to 2019. Because data are uneven across the economies, we have two unbalanced panels and a parsimonious set of controls. Variable definitions and data sources are described in Supplement C.

Fiscal reaction function
The results of our fixed effects panel regressions suggest that fiscal fatigue is a robust feature of the LICs and MACs in our sample. This is shown in Figures 2 and 3, and supported by estimates in Supplement D. The figures present mean primary balances (blue line) as well as fitted values from our estimation results (orange line), computed for a specified debt range (0-10%, 11-20%, and so forth). The lines indicate that at low levels of debt, the primary balance does not respond positively to debt, but the response changes as debt increases, becoming positive before eventually weakening. For LICs, the response is predicted to start declining when debt to GDP is a little over 92%. For MACs, the inflection point is around 87%.

Estimated debt limits
Summary statistics of the IRGDs and debt ratios used in estimating debt limits and associated fiscal space are reported in Table 1; individual country statistics are in Supplements E.1 and E.2. LICs have relatively large and negative IRGDs; a positive average occurs in just two instances. 12 The IRGDs for MACs are also negative on average but less negative than those for LICs by about 3.4 to 4.7 percentage points. Historically, two MAC countries have had positive IRGDs; three are projected to have positive IRGDs from 2021 to 2025.
Average debt ratios in LICs have been about 9.9 to 13.1 percentage points less than the debt ratios in MACs. They are expected to be about 16 percentage points less in 2025, consistent with the fact that MACs have better access to debt markets. In 2025, all but three MACs are projected to breach the 50% debt-to-GDP DSF threshold for emerging market economies.
Estimated debt limits (d**) and 'conditionally stable' long-run average debt ratios (d*) are reported in Tables  2 and 3. Estimated debt limits are, on average, higher in LICs than in MACs, which is consistent with the larger, negative IRGDs of the LICs. Estimated debt limits for LICs range from 123.7 to 160.3%, with mean at 141.7% of GDP in the historical case. They are 113.9 to 148.5%, with mean at 137.4% of GDP in the projected-shocked case. For MACs, estimated debt limits range from 118.8 to 162.4%, with mean at 132.1% of GDP in the historical case. They are 79.2 to 151.7%, with mean at 121.4% of GDP in the projected-shocked case.  Interestingly, MACs and LICs are converging to long-run average debt ratios (d*) that are in the vicinity of DSF debt thresholds. This is shown in Table 4. For instance, the estimated d* for MACs average 42.4 and 41.7% using historical and projected-shocked IRGDs respectively, which is below the 2013 DSF debt-to-GDP threshold of 50% for emerging MACs. 13 This could indicate a disciplining effect of the DSF threshold, but it also confirms that DSF thresholds should not be construed as limits to fiscal space per se. The differences between estimated debt limits (d**) and DSF thresholds are, in fact, large. Differences suggest significant opportunity costs if, say, markets react poorly to a government's actual or anticipated breach of the DSF threshold, triggering a sudden stop, despite its fiscal policy remaining consistent with long-run solvency.

Estimated fiscal space
Fiscal space associated with our estimated debt limits is computed as the distance from debt ratios projected for 2025 to these limits. These point estimates are reported in Supplements F.1 and F.2. Fiscal space in terms of probabilitiesspecifically, that additional fiscal space at projected-shocked IRGDs is greater than 0, 25, 50, 75 and 100% of GDPare reported in Tables 5 and 6 below. Probabilities are calculated from coefficient estimates obtained from Monte Carlo simulations as discussed in Section 2.
There are high probabilities that LICs have additional fiscal space at projected-shocked IRGDs. All countries barring three have a very high chance (90% probability or more), or a high chance (70% to 90% probability), d 'Shocked' is the difference between country-specific mean of real interest rate plus one country-specific standard deviation, and countryspecific mean of GDP growth minus one country-specific standard deviation.
that additional fiscal space of between 75 and 100% of GDP exists. The last three -Ghana, Maldives and Samoa have high probabilities (70% to 90% probability) that additional fiscal space of between 50 and 75% of GDP exists. The picture is less rosy for MACs, although fiscal space is reasonably ample for many. Two countries (Vietnam and Indonesia) have high probabilities that additional fiscal space is between 75 and 100% of GDP. Five (Mongolia, Philippines, Guatemala, Morocco, Malaysia) have moderate or high probabilities that additional space is between 50 and 75% of GDP. Three (Dominican Republic, Columbia, Thailand) have a 40 to 50% chance that additional space is between 50 and 75% of GDP but also very high probabilities that additional space is between 25 and 50% of GDP. The probability of another 25 to 50% of GDP of fiscal space is high for Tunisia, moderate for Sri Lanka, and low, near-zero or zero for Fiji, Costa Rica, and Singapore.
The higher probabilities of larger amounts of additional fiscal space for LICs relative to MACs are consistent with the relatively higher estimated debt limits and lower projected debt in 2025 among LICs.
When we reduce estimated debt limits by 20%, to approximate the case of stochastic shocks and increasing risk premiums, results suggest that LICs will probably continue to have room for fiscal maneuver. A handful of MACs (Costa Rica, Fiji, Sri Lanka, Tunisia, Singapore) will have little or no room (Supplements G.1 and G.2). 'Shocked' is the difference between country-specific mean of real interest rate plus one country-specific standard deviation, and countryspecific mean of GDP growth minus one country-specific standard deviation.

Conclusions
We study the availability of fiscal space among the membership of the V20 and ASEAN who require urgent climate adaptation and transition investments. We estimate country-specific debt limits, beyond which fiscal solvency fails, using a model that abstracts from liquidity risks. This allows us to illuminate the difference between long-term insolvency and short-term liquidity crises and measure the size of the fiscal space available from a long-term perspective. This is given by the distance between estimated debt limits and IMF-projected debt ratios. At the same time, we are able to gauge the opportunity cost of the so-called financial death trap. This is suggested by the difference between estimated debt limits and prescribed IMF DSF thresholds. We find that, with a handful of exceptions, fiscal space is ample, although less so for MACs. We confirm that DSF thresholds should not be construed as limits to fiscal space per se. The distances between estimated debt limits and DSF thresholds are not small, suggesting significant opportunity costs for otherwise solvent governments if and when breaching DSF thresholds trigger sudden stops of financing for climate investments. If  foregone or delayed adaptation investments include mitigation co-benefits, then opportunity costs are shared by the global community.
Our findings imply that climate-vulnerable developing countries should be afforded a second look by international financial institutions, using a long-term lens. This measure of fiscal space, which reflects how consistent a country's record of fiscal behavior has been with long-term solvency, is one such lens. Since climate change adaptation itself requires a long-term perspective, this measure could be an important component of a multipronged strategy to unlock greater flows of adaptation finance at lower costs.
This measure can also help allocate scarce adaptation grants and loans among countries more efficiently without absolving developed countries from their obligation to mobilize more funds. For instance, countries with less fiscal space and less capacity for investment planning could be considered for more grant-based adaptation planning. Building capacity for adaptation planning, especially for investments with large multiplier effects and sustainable development dividends, is crucial. By having a positive impact on growth, or by preventing growth from being disrupted by climate change, such investments can help an economy outgrow its debt.
Our findings imply that actions to obviate financial death traps are warranted. Opportunity costs of doing nothing are high. National and global climate ambitions could be derailed if otherwise solvent and able climate-vulnerable governments are caught in such a financial death trap.