Asymmetric sovereign risk: Implications for climate change preparation

Gómez González, José Eduardo; Uribe, Jorge; Valencia, Oscar M. Working Paper Asymmetric sovereign risk: Implications for climate change preparation IDB Working Paper Series, No. IDB-WP-1588 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Gómez González, José Eduardo; Uribe, Jorge; Valencia, Oscar M. (2024) : Asymmetric sovereign risk: Implications for climate change preparation, IDB Working Paper Series, No. IDB-WP-1588, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0012853 This Version is available at: https://hdl.handle.net/10419/299414 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ Asymmetric Sovereign Risk Implications for Climate Change Preparation Jose E. Gomez-Gonzalez Jorge M. Uribe Oscar M. Valencia WORKING PAPER No IDB-WP-1588 Inter-American Development Bank Institutions for Development Sector Fiscal Management Division March 2024 Asymmetric Sovereign Risk Implications for Climate Change Preparation Jose E. Gomez-Gonzalez, Lehman College, City University of New York Jorge M. Uribe, Universitat Oberta de Catalunya Oscar M. Valencia, Inter-American Development Bank Inter-American Development Bank Institutions for Development Sector Fiscal Management Division March 2024 Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Gomez-Gonzalez, Jose E. Asymmetric sovereign risk: implications for climate change preparation / Jose E. Gomez- Gonzalez, Jorge M. Uribe, Oscar M. Valencia. p. cm. — (IDB Working Paper Series ; 1588) Includes bibliographical references. 1. Climatic changes-Economic aspects-Latin America. 2. Climatic changes-Economic aspects- Caribbean Area. 3. Debts, Public-Latin America. 4. Debts, Public-Caribbean Area. I. Uribe, Jorge M. II. Valencia Arana, Oscar. III. Inter-American Development Bank. Fiscal Management Division. IV. Title. V. Series. IDB-WP-1588 http://www.iadb.org Copyright © 2024 Inter-American Development Bank ("IDB"). This work is subject to a Creative Commons license CC BY 3.0 IGO (https://creativecommons.org/licenses/by/3.0/igo/legalcode). The terms and conditions indicated in the URL link must be met and the respective recognition must be granted to the IDB. Further to section 8 of the above license, any mediation relating to disputes arising under such license shall be conducted in accordance with the WIPO Mediation Rules. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the United Nations Commission on International Trade Law (UNCITRAL) rules. The use of the IDB's name for any purpose other than for attribution and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this license. Note that the URL link includes terms and conditions that are an integral part of this license. The opinions expressed in this work are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. Oscar M. Valencia, [email protected] Abstract* Climate change adaptation efforts are heavily dependent on a country’s fiscal capacity and the associated costs of undertaking adaptation policies. The current accumulation of high debt levels in emerging and low-income developing countries, which are disproportionately affected by climate change, raises significant concerns. This study shows that sovereign risk, and hence funding costs for governments, exhibits significantly asymmetric reactions to its determinants across the conditional distribution of credit spreads. This aspect, previously overlooked in the literature, has relevant policy implications. Countries with elevated risk levels are disproportionately vulnerable to climate change compared to their lower-risk counterparts, especially in the short term. Notably, investing in climate change preparedness proves effective in mitigating vulnerability to climate change, in terms of sovereign risk, particularly for countries with low spreads and long-term debt (advanced economies), where readiness and vulnerability tend to counterbalance each other. However, for countries with high spreads and short-term debt, additional measures are essential as climate change readiness alone is insufficient to offset vulnerability effects in this case. Results also demonstrate that the actual occurrence of natural disasters is less influential than vulnerability to climate change in determining spreads. JEL Codes: F34, G15, H63, Q51, Q54 Keywords: credit risk, disaster risk, nonlinear dynamics, panelquantile regressions, preparedness, sovereign risk, vulnerability *ĺAuthor information: Jose E. Gomez-Gonzalez, Lehman College, City University of New York, Visiting Professor,ĺ Universidad de la Sabana, [email protected] Jorge M. Uribe, Universitat Oberta de Catalunya, Barcelona (Spain), [email protected] Oscar M. Valencia, Fiscal Management Division, Inter-American Development Bank, [email protected] 2" 1. Introduction Climate change poses a considerable threat to countries’ macroeconomic and financial stability, as well as to development efforts of emerging and low-income economies. The ability of these economies to adapt is closely tied to their fiscal capacity and the cost of adaptation. Countries with high fiscal capacity are better positioned to implement effective mitigation and adaptation strategies. Conversely, limited fiscal capacity, prevalent in emerging markets and low-income developing economies, hinders their adaptation efforts, amplifying their vulnerability to climate change—a vulnerability that surpasses that of advanced economies (Bolton et al., 2022). This study examines how sovereign risk spreads and, consequently, the cost of national funding, respond to vulnerability and preparedness to climate change, while recognizing the different dynamics expected from emerging, low-income, and advanced economies. Sovereign debt determinants, as outlined in the existing literature, encompass macroeconomic, institutional, external sector, and fiscal factors, along with natural disasters and climate change-related fundamentals. We present a novel and comprehensive empirical framework that makes it possible to evaluate the effects of these variables across the entire conditional spread distribution, thereby facilitating a more precise analysis of debt dynamics, which are inherently nonlinear. The nonlinearity primarily stems from the fact that the impacts of vulnerability, preparedness, and other determinants are not uniform across the spread distribution. This is evident, as we anticipate that the adverse effects of climate change will disproportionately affect countries with initially higher spreads and reliance on mainly short-term debt. As anticipated, our findings indicate that climate change vulnerability becomes notably significant for shorter maturities, especially those equal to or less than two years. This impact is particularly pronounced for countries with a highrisk profile that experience elevated borrowing costs in the global debt market. Our results highlight the importance of international efforts aimed at addressing the repercussions of climate change, where such initiatives should recognize the distinct impact of climate change on interest payments for debt, especially for 3" emerging and low-income countries, as the world undergoes a global ecological transition. This study expands the existing body of research that empirically models sovereign risk and sovereign yields and, in particular, the recent literature that investigates the impact of climate change and natural disasters in sovereign risk. Literature in the former set typically emphasizes the significance of fiscal discipline and long-term growth in mitigating sovereign risk and reducing spreads, especially over the long run. According to this literature, in the long term, fundamental factors such as the debt-to-GDP ratio significantly shape market sovereign bond spreads, whereas in the short term, financial volatility becomes a dominant determinant (Bellas et al., 2010; Poghosyan, 2014). Other traditional factors influencing sovereign yields include local and foreign monetary policy conditions, local inflation rates, deficit-to-GDP ratios, terms-of-trade and their volatility, fiscal variables and political factors, alongside the quality of domestic institutions, among others (see, for instance, Afonso and Jalles, 2019; Arora and Cerisola, 2001; Beqiraj, Patella, and Tanzioni, 2021; Brooks, Cunha, and Mosley, 2022; Caggiano and Greco, 2012; Chatterjee and Eyigungor, 2019; Dailami, Masson, and Padou, 2008; De Santis 2020; Eichler, 2014; Hilscher and Nosbusch, 2010; Krishnamurthy, Nagel, and Vissing- Jorgensen, 2018; Liu and Spencer 2013; Mati, Baldacci, and Gupta, 2008; Matsumura and Machado, 2010). Our contribution to this literature is straightforward. We are the first to consider a nonlinear relationship between the explanatory factors outlined above and the sovereign spreads, governed by the level of the spread, that is, according to the level of sovereign risk itself. Although our postulate is innovative, it firmly aligns with the established tradition in the field of distinguishing emerging (and low-income) economies from developed economies when analyzing sovereign risk. Notably, when sovereign risk is examined in advanced economies, the spread is termed as a “convenience yield” (Du and Schreger, 2016; Du, Im, and Schreger, 2018), as the dynamics of spreads are anticipated to diverge when they are high compared to when they are low. Addressing this distinction directly, we employ panel quantile models, demonstrating that certain determinants of spreads hold more relevance for different segments of the spread distribution, while others are virtually unimportant at specific quantiles. At the same time, our model refrains 4" from establishing arbitrary distinctions between countries, particularly in terms of categories like “advanced,” “emerging,” or “low income,” which lack solid economic grounds. In short, we postulate that the different dynamics observed in the data are associated with the level of risk, rather than with some ambiguous country characteristics. Given the predominant role of external influences on sovereign risk, a subset of research has probed the impact of financial and trade openness on sovereign spreads (Maltritz, 2012; Maltritz and Molchanov, 2014) and the importance of considering the high commonality in international debt markets when modeling sovereign spreads (Gilchrist et al., 2022; Gomez-Gonzalez, Uribe, and Valencia, 2023a; Liu and Spencer, 2013; Longstaff et al., 2011). To this literature we own the inclusion of a common international factor in our models. We empirically assess the impact of external factors on country-specific risk and demonstrate that this factor, which we estimate ourselves, remains consistently significant, irrespective of the segment of the spread distribution analyzed or the maturity of the spread. To the best of our knowledge, we are the pioneers in undertaking such an analysis. Our research also is related to a branch of the literature that explores how different maturities of sovereign yields and spreads respond to economic shocks. Theoretically, long-term interest rates reflect expectations about a government’s future solvency and financing needs, while short-term rates indicate concerns about liquidity and short-term performance outlooks (Eichler and Maltritz, 2013; Freixas and Rochet, 2008). The composition of long-term and short-term debt is crucial, especially for emerging market economies, with long-term debt acting as a safeguard against interest-rate spread fluctuations and short-term debt encouraging prompt repayment (see Arellano and Ramanarayanan, 2012; Sánchez, Sapriza, and Yurdagul, 2018). Notably, Eichler and Maltritz (2013) delve into the factors influencing government bond yield spreads. Their findings indicate that low economic growth and greater economic openness amplify default risk across all maturity levels, while heightened indebtedness exclusively heightens short-term risk. We conduct our analysis for different maturities as well and find that the effects of most of the variables are greater in short-term maturities, especially for the highest quantiles of the spreads. 5" The second set of studies to which we contribute, which analyzes the impacts of climate change preparation and vulnerability and natural disasters on sovereign risk, is still in its infancy. Notable contributions have recently been made by Bolton et al. (2022) and Klusak et al. (2023) from a policy-oriented perspective and Mallucci (2022) from a theoretical standpoint that explicitly incorporates natural disasters and climate change risk into a traditional framework of sovereign debt price determination in the vein of Hatchondo and Martinez (2009) and Chatterjee et al. (2023). Bolton et al. (2022) offer a comprehensive overview of the literature linking sovereign debt and climate change risk, examining various dimensions of the interplay between climate and debt. Their analysis involves an exploration of the financial costs associated with climate adaptation and potential fiscal constraints that may impede the implementation of such adaptation measures. Additionally, they investigate the role of green bonds in financing climate adaptation and assess whether a premium, known as a “greenium,” exists in the sovereign debt market for environmentally friendly initiatives. Notably, their findings reveal the absence of a greenium. From the policy perspective, several other organizations, including the Inter-American Development Bank, the International Monetary Fund, and the United Nations, have contributed substantially to this body of work (e.g., Aligishiev, Massetti, and Bellon, 2022; Buchner et al., 2021; Buhr et al., 2018; Delgado, Eguino, and Lopes, 2021; Powell and Valencia, 2023; Voltz et al., 2020). In a nutshell, these reports shed light on the challenges and opportunities faced by both developed and emerging economies as they grapple with the consequences of climate change through fiscal and policy measures. They employ diverse research methodologies, including interviews, surveys of finance ministers and other key stakeholders, and data from a wide array of sources, including national statistics on emissions, energy sources, and fiscal revenue derived from fossil fuel sales. Subsequently, this information is harnessed to project potential scenarios of GDP and fiscal losses due to climate change risks, both from physical impacts and transition-related changes. Together, these reports offer an ample understanding of fiscal policies and global initiatives addressing climate change. However, it is important to note that these recommendations can at times be overly broad and may not fully recognize ! 12" We assume x = #x!,x",…,x#& to be a n×p dimensional data matrix. Following Stekhoven and Bühlmann (2012), we directly predict the missing values using RF estimated on the observed variables present in the dataset. For any arbitrary variable x$, including missing points at entries i$ %& ∈{1,…,n} the dataset can be split into four parts: (1) the non-missing values of x$, denoted y$ '($ ; (2) the missing observations, y$ %&; (3) variables different from s, with observations i$ '($ ={1,…,n}\i$ %& denoted as x$ '($, and (4) the variables other than x$ with observations i$ %&, denoted by x$ %&. The RF model first makes an initial conjecture for the missing values in x, in our case the mode value. Then, it sorts the variables x$, s = 1,…,p according to the number of missing observations. For each variable x$ the missing values are filled in by estimating a RF model with response variable y$ '($ and predictors the rest of the variables in a given year, x$ '($. Then, the algorithm proceeds by predicting the missing values y$ %& by applying the estimated RF to the x$ %&. This procedure is repeated until a pre-specified stopping criterion is met. This stopping criterion is met when the difference between the newly imputed data matrix and the previous one increases for the first time with respect to both continuous and discrete variables. For the N continuous variables, that is: ∆%=∑*+!"# $%&'("),+*+) $%&'(")-, -∈/ ∑*+!"# $%&'(")-, -∈/ , (1) while for F discrete variables it takes the form: ∆.=∑ ∑ /0!"# $%&'(")10*+) $%&'(") ! $23 -∈4 #%& , (2) where #NA is the number of missing entrances in the categorical variables. 2.2. Factor-Augmented Panel Quantile Model Once we have completed the data for the spreads, we used the completed vectors as the response variable in a panel-quantile framework for four different maturities, 1 year, 2 years, 5 years and 10 years, separately, which become 𝑦12 34256127 in the following presentation, where we will omit the superscript to avoid unnecessary notation. Abrevaya et al. (2008) and Koenker (2004), among others, have extended panel quantile models to longitudinal contexts. In their approach, the dynamics of ! 13" the τ-quantile of the dependent variable are characterized by the following equation: 𝑄8(𝑦12|𝑏1,𝛽,𝑥12)= 𝑏1+𝑥12 9𝛽8, (3) where, for a given quantile 𝜏 ∈ (0,1), 𝛽8 summarizes the relationship between the explanatory variables 𝑥 and the 𝜏-th response quantile, for a country whose spread baseline level is equal to 𝑏1. 𝑥 consists of some key indicators previously identified by the literature on sovereign risk, including the inflation rate, real growth, the terms of trade of the country, the economic complexity indicator, the debt-to-GDP ratio, natural resource rents as a percentage of GDP, and the Rule of Law indicator. Crucially, 𝑥 also contains vulnerability and readiness indicators, which both add up to the ND-GAIN indicator, which assess a country’s exposure to and socio-economic capacity to face climate change and will be explained in detail in the next data section. The model in equation 3 is akin to traditional panel data models of the yield spreads and can be equivalently written as: 𝑦12 = 𝑏1+𝑥12 9𝛽8+𝜀12, (4) where,?𝑄8(𝜀12?|𝑏1,𝛽,𝑥12)= 0. There are two distinct approaches to estimate such (conditional) quantile regression in longitudinal data, with a distinction between distribution-free methods and likelihood-based methods. In the distribution-free approach, fixed individual-specific intercepts are considered, treated as location shift parameters common to all conditional quantiles. This implies that the conditional distribution for each individual has the same shape but different locations, as long as the 𝑏1’s are different. Koenker (2004) introduced fixed effect quantile regression for longitudinal data in this vein. In contrast, within the likelihood framework, individual-specific parameters 𝑏1’s are assumed to be independent and identically distributed random variables. This framework effectively allows for explaining differences in the response variables across individuals (countries) and quantiles (different spread levels, associated with varying degrees of sovereign risk). It also allows us to introduce into our model, in the last section of our results, dummy variables that measure different dimensions of the occurrence of natural disasters in a given country, in a given year, within our sample period, in a parsimonious and natural way. ! 14" Let 𝑏1=(𝑏1!,…,𝑏18) represent a 𝜏-dimensional vector of individual random parameters, which density is given by 𝑓:(∙;𝐷8). 𝐷8 is covariance matrix dependent on 𝜏. In such a case, a linear quantile mixed model is defined as follows: 𝑄8(𝑦12|𝑏1,𝛽,𝑥12,𝑧12)= 𝑥12 9𝛽8+𝑧12 9𝑏1, (5) where 𝑧12 denotes an additional set of variables. In the simplest case, followed in our baseline model, 𝑧12 can be set to a vector of ones, which specifies time fixed effects per country, defined over the mixture densities 𝑓:(∙;𝐷8). Or 𝑧12 may account for the natural disaster variables as well, like in our final model specification (section 4.3). The random structure of 𝑏1?in equation 5 enables the consideration of between-individual heterogeneity without necessitating orthogonality between the observed and omitted covariates (Geraci and Bottai, 2014; Marino and Farcomeni, 2015). Alternatively, the equation above can be written as follows: 𝑦12 = 𝑥12 9𝛽8+𝑧12 9𝑏1+𝜀12, (6) where,?𝑄8(𝜀12?|𝑏1,𝛽,𝑥12,𝑧12)= 0. Recall that 𝑏1 is a vector of country- and quantilespecific random coefficients which account for unobserved heterogeneity that is not captured by the elements in 𝑥12 and describe the dependence between repeated measurements from the same country/unit over the time. Moreover, for a given quantile level 𝜏, 𝑦12 is assumed to have an Asymmetric Laplace Density (ALD) (e.g., Yu and Moyeed, 2001) given by: 𝑓7|:#𝑦12|𝑏1,8;𝜏& = E8(!,8) ?5FexpH−𝜌8E767,@67,5 ?5FK. (7) where, 𝜌8(∙) denotes the quantile asymmetric loss function (Koenker and Bassett, 1978), while 𝜎8, and 𝜇12,8, stand for the scale location parameters of the distribution, respectively. All in all, the ALD facilitates maximum likelihood estimation. Furthermore, the location parameter 𝜇12,8 is modeled as follows: 𝜇12,8= 𝑥12 9𝛽8+𝑧12 9𝑏1. (8) The modeling strategy is completed by the mixing distribution 𝑓:(∙;𝐷8) introduced before. At this point, instead of specifying a distribution parametrically, Alfó, Salvati, and Ranalli (2017) and Geraci and Bottai (2014) proposed estimating it directly from the data via a Non-Parametric Maximum Likelihood approach (NPML). This leads to the estimation of a quantile-specific discrete mixing distribution defined over the set of locations N𝜁!,...,𝜁A,8Q, with mixture probabilities ! 15" 𝜋A,8 ?= ?𝑃𝑟#𝑏1?=?𝜁A,8&, ?𝑖? = ?1,...,𝑛, 𝑔?? = ?1,...,𝐺8?, and 𝐺8≤ ?𝑛. Following this proposal, the location parameter of the ALD in equation (8) becomes?𝜇12,8= 𝑥12 9𝛽8+𝑧12 9𝜁A,8, and the likelihood for estimation is defined accordingly as follows: 𝐿(∙|𝜏)=∏ ∑ ]∏𝑓7|:#𝑦12|𝑏1,8 = 𝜁A,8;𝜏& B6 2C! ^ D5 AC! E 1C! 𝜋A,8. (9) Left to include are the time-varying common factors that measure global macroeconomic forces that are expected to influence all debt maturities and spreads for all countries at the same time, but in a distinctive fashion. Doing so equation 6 above can be written as follows: 𝑦12 = 𝑎1𝑓2+𝑥12 9𝛽8+𝑧12 9𝑏1+𝜀12, (10) where, 𝑓2 has dimensionality 𝑘 = 1 in our baseline model, and it is estimated via PCA in a preliminary step. Note that the inclusion of this time-varying factor is a valid and parsimonious alternative to explicitly including in the model general macro-forces such as the VIX, oil prices, uncertainty, world interest rates, TED spreads, and other proxies for global financial cycles, inflation cycles, or commodity cycles, as far as this single factor adequately captures the variation in the common dynamics affecting sovereign credit spreads globally. While this is trivially true when 𝑘 = 𝑛FG5E261HI, it holds only approximately when 𝑘 = 1. The quality of this approximation is determined by the percentage of variance explained by the first principal component in the factor model of credit spreads. In our case, the first factor accounts for 42.8 percent of the variability in the 272 series of spreads (4 for each of the 68 countries), demonstrating a remarkably high explanatory power and validating our factor-based approach. We also assess the sensitivity of our model to the inclusion of a second factor and other modeling choices. Our results remain unaltered in this case. We conclude this section by underlining the essential role of panel quantile models in estimating the direct effects of economic and climate-related determinants on spreads, similar to traditional panel models. In a longitudinal setting, panel-like structures enable the modeling of unobserved heterogeneity between countries. Ignoring this heterogeneity would lead to biased estimates of quantile effects. The inclusion of the factor structure is driven by the need to tailor the methodology from a statistical medicine context to an international macroeconomic setting. This adjustment is essential given the well-documented ! 16" presence of risk commonality across countries not only in terms of sovereign risk, but also across any given economic fundamental. 3. Data This study uses two datasets. The first dataset, which we call the main dataset, is the one used in our main regression results in Section 5.2. This section contains the panel-quantile models explained in Section 3.2. The second dataset consists of 66 additional variables in three dimensions, macroeconomic, debt-related and institutional/political variables which help us to train the RF, as explained in Section 3.1. The results of this imputation exercise are described in detail in Section 5.1 of the results. 3.1. Main Regression Variables Table 1 shows the variable description, variable short-name, source of information, mean, median standard deviation, maximum and minimum values of our main variables, while Table 2 consists of the country names, ISO-3 codes, and whether a country is considered to be advanced or otherwise. Only the spreads contain imputed observations in Table 1. The yields for different maturities were downloaded from Bloomberg. As the table shows, there is significant variability in the spreads, as highlighted by the substantial disparities between the maximum and minimum spread values across various maturities. The broad spread variation reflects the diversity among the countries included in the sample. ! 17" Table 1. Summary of Main Variables Statistics Indicator Abreviation Source Mean Median Std.Dev Max. Min. Value sovereign spread with respect to the US 1 year maturity ValSpread_1Y Bloombergown elaboration 4.78 3.22 6.95 90.51 -5.88 Value sovereign spread with respect to the US 2 years maturity ValSpread_2Y Bloombergown elaboration 5.47 2.79 10.62 108.47 -5.86 Value sovereign spread with respect to the US 5 years maturity ValSpread_5Y Bloombergown elaboration 4.72 2.87 7.4 95.7 -5.08 Value sovereign spread with respect to the US 10 years maturity ValSpread_10Y Bloombergown elaboration 4.4 2.51 7.31 92.62 -4.22 The number of people affected > 100.000 a year ndisaster1 EMDAT 0.22 0 0.42 1 0 The number of deaths > 1,000 a year ndisaster2 EMDAT 0.03 0 0.17 1 0 Economic damage is > 2% of GDP ndisaster3 EMDAT 0.01 0 0.1 1 0 At least one of the three conditions above is met ndisaster EMDAT 0.23 0 0.42 1 0 ndisaster 1 is met and there are weather disasters ndisaster1_weather EMDAT 0.22 0 0.41 1 0 ndisaster 1 is met and there are geophysical disasters ndisaster1_geophysical EMDAT 0.08 0 0.28 1 0 ndisaster 2 is met and there are weather disasters ndisaster2_weather EMDAT 0.03 0 0.17 1 0 ndisaster 2 is met and there are geophysical disasters ndisaster2_geophysical EMDAT 0.02 0 0.13 1 0 ndisaster 3 is met and there are weather disasters ndisaster3_weather EMDAT 0.01 0 0.09 1 0 ndisaster 3 is met and there are geophysical disasters ndisaster3_geophysical EMDAT 0.01 0 0.09 1 0 At least one of the three conditions above is met for weather ndisaster#_weather ndisaster_weather EMDAT 0.23 0 0.42 1 0 At least one of the three conditions above is met for weather ndisaster#_geophysical ndisaster_geophysical EMDAT 0.08 0 0.28 1 0 Natural resources rents as % of GDP rents WEO-IMF 2.66 0.4 5.48 43.08 0 Rule of law rle World Bank 0.54 0.52 0.93 2.13 -1.43 Terms of trade change in % tot WEO-IMF 102.35 100 15 183.84 30.73 Real GDP growth growth WEO-IMF 3.65 3.59 3.49 28.08 -15.1 Inflation rate, inf_avg WEO-IMF 4.32 2.9 5.01 55.04 -4.87 Gross debt as % of GDP, general government debt WEO-IMF 56.02 46.35 36.81 260.96 3.82 Readiness Indicator readiness ND-Gain Web Page 0.5 0.48 0.14 0.81 0.2 Vulnerability Indicator vulnerability ND-Gain Web Page 0.39 0.38 0.08 0.6 0.25 Economic Complexity Indicator eci Harvard Growth Lab Web Page 0.58 0.58 0.9 2.82 -2.34 Note: The table shows the main variables used in this study, the variables’ description, sources of information, and summary statistics in the five right-hand columns. Source: Authors’ elaboration. ! 18" Table 2. Countries Included in the Sample # Country name ISO3 Advanced Emerging/ low-income # Country name ISO3 Advanced Emerging/ low-income 1 Australia AUS 1 0 35 Korea KOR 1 0 2 Austria AUT 1 0 36 Lebanon LBN 0 1 3 Belgium BEL 1 0 37 Sri Lanka LKA 0 1 4 Bangladesh BGD 0 1 38 Lithuania LTU 1 0 5 Bulgaria BGR 0 1 39 Latvia LVA 1 0 6 Brazil BRA 0 1 40 Morocco MAR 0 1 7 Botswana BWA 0 1 41 Mexico MEX 0 1 8 Canada CAN 1 0 42 Mauritius MUS 0 1 9 Switzerland CHE 1 0 43 Malaysia MYS 0 1 10 Chile CHL 0 1 44 Namibia NAM 0 1 11 China CHN 0 1 45 Nigeria NGA 0 1 12 Colombia COL 0 1 46 Netherlands NLD 1 0 13 Costa Rica CRI 0 1 47 Norway NOR 1 0 14 Cyprus CYP 1 0 48 New Zealand NZL 1 0 15 Czech Republic CZE 1 0 49 Pakistan PAK 0 1 16 Germany DEU 1 0 50 Panama PAN 0 1 17 Denmark DNK 1 0 51 Peru PER 0 1 18 Egypt EGY 0 1 52 Philippines PHL 0 1 19 Spain ESP 1 0 53 Poland POL 0 1 20 Finland FIN 1 0 54 Portugal PRT 1 0 21 France FRA 1 0 55 Qatar QAT 0 1 22 United Kingdom GBR 1 0 56 Romania ROU 0 1 23 Greece GRC 1 0 57 Russia RUS 0 1 24 Croatia HRV 0 1 58 Singapore SGP 1 0 25 Hungary HUN 0 1 59 Slovak Republic SVK 1 0 26 Indonesia IDN 0 1 60 Slovenia SVN 1 0 27 India IND 0 1 61 Sweden SWE 1 0 28 Ireland IRL 1 0 62 Thailand THA 0 1 29 Iceland ISL 1 0 63 Turkey TUR 0 1 30 Israel ISR 1 0 64 Uganda UGA 0 1 31 Italy ITA 1 0 65 Ukraine UKR 0 1 32 Japan JPN 1 0 66 Vietnam VNM 0 1 33 Kazakhstan KAZ 0 1 67 South Africa ZAF 0 1 34 Kenya KEN 0 1 68 Zambia ZMB 0 1 Note: The table shows the countries included in our sample, with their respective ISO3 codes and a dummy variable of whether they are advanced or otherwise in terms of development. They appear in alphabetical order according to the ISO3 codes. Source: Authors’ elaboration. ! 19" The minimum spread is negative across all maturities. This phenomenon is primarily due to the inclusion of countries such as Germany, Japan, and the United Kingdom, which have sometimes exhibited lower sovereign spreads than those of the United States throughout the sample period. Most of the remaining countries have consistently maintained positive spreads, with some emerging nations exhibiting high spreads. The average spreads are notably higher in the short term (especially the 2-year spreads) compared to the medium (5-year) and long-term (10-year) maturities. As can be seen in Table 2, our sample consists of 68 countries, roughly 44 percent of which are advanced and the remaining 56 percent are emerging or lowincome developing nations. Our country sample is larger than the previous samples. It represents an increase of 70 percent compared to the 40 countries in Beirne et al. (2021), and 1.4 times the dataset available from Du and Schreger (2016) and Du, Im, and Schreger (2018). It also includes earlier years, as it starts in 2000. Our analysis excludes the years 2020 and 2021, for most data are readily available due to the extraordinary disruptions caused by the COVID-19 pandemic, which significantly influenced international debt market dynamics in a way orthogonal to our interests (see, for example, Candelon and Moura, 2023). 3.1.1. Macroeconomic, Fiscal, Institutional Covariates, and Natural Disasters Consistent with prior research, we address potential confounding factors by incorporating several macroeconomic, fiscal, and institutional covariates into our empirical model. Table 1 also provides descriptive statistics for these variables, presenting information such as their source, mean, median, standard deviation, maximum, and minimum values. We additionally incorporate binary variables that take the value of 1 when a country experiences a natural disaster in a specific year, according to a variety of criteria. This set of variables offers valuable insights, as countries that have endured natural disasters could be more susceptible to climate risk vulnerabilities. Moreover, the repeated exposure to such disasters may incentivize a country to enhance its preparedness for future occurrences. ! 20" The variables measuring the occurrence of weather disasters were retrieved from the EMDAT (or EM-DAT), the international disasters database. The rents resulting from natural results exploitation, real growth, inflation, terms of trade, and debt-to-GDP ratio were obtained in different public datasets by the International Monetary Fund, among them the World Economic Outlook, 2019. The Rule of Law estimate was extracted from the World Development Indicators of the World Bank. The Economic Complexity Indicator (ECI) comes from the Harvard Growth Lab, which reports the ECI index developed by Hidalgo and Hausmann (2009). 3.1.2. Measuring Climate Vulnerability and Readiness For Adaptation Regarding the indexes to measure vulnerability and exposure to climate change, we used those provided by the Notre Dame Global Adaptation Initiative (ND-GAIN). We employ the ND-Gain index, along with its constituent elements, to assess both climate vulnerability and the capacity for adapting to climate change. According to the ND-GAIN website, the ND-GAIN Country Index is designed to consolidate a country’s susceptibility to the consequences of climate change, as well as its readiness to bolster resilience in the presence of climate-related challenges. The primary index can be dissected into two fundamental dimensions, namely vulnerability and readiness, as illustrated in Figure 1. Vulnerability pertains to a country’s predisposition to being adversely affected by climate-related hazards, while readiness signifies the nation’s level of preparedness to undertake adaptive measures, incorporating responses from both the public and the private sectors. Vulnerability indicators can be further subdivided into six life-supporting sectors: health, food, ecosystems, habitat, water, and infrastructure. Each of these is evaluated across three key dimensions: exposure, sensitivity, and adaptive capacity. Readiness, on the other hand, can be broken down into three distinct categories: economic, social, and governance. This division can be extended to yield highly actionable indicators tailored for policymakers, with the exception of exposure indicators, which lack actionable aspects. Figure 1 presents a visual representation of the ND-GAIN country index and its constituent elements. Summary statistics for ND-GAIN indicators and their ! 21" components are provided in Table 1. The readiness and vulnerability components of the ND-GAIN index are expressed on a scale from 0 to 1. A higher value on the readiness index signifies better preparedness for climate events, whereas a higher value on the vulnerability index indicates a greater likelihood of climate event occurrence. These two indexes also exhibit substantial variation. The readiness index spans from 0.20 (Nigeria in 2014) to 0.81 (Singapore in 2014) within our sample, while the vulnerability index ranges from 0.25 (Switzerland in 2015) to 0.60 (Uganda 2004). Broadly, there is a positive correlation between both the ND-GAIN and the readiness index and a country’s level of development, while the correlation between the vulnerability index and development is negative. Figure 1. Graphical Description of the ND-GAIN Country Index and Its Components Note: This figure was adapted from the webpage of NDGAIN. It shows the components of the NDGAIN indicator, vulnerability and readiness, and the subcomponents of each: six sectors in the former case, and three dimensions in the latter. It also shows the number of original series that are used to construct each of the sectors and dimensions in brackets. Source: Authors’ elaboration. 3.2. Auxiliary Variables for Imputation Table A1 of the Appendix displays the variables used in the RF that we train to impute the missing observations of spreads in the first part of our results. We include 66 variables, in addition to the variables unrelated to climate change from Table 1, that is, the ND-GAIN indexes and the EMDAT natural disasters dummy variables. These variables can be broadly categorized as related to debt or fiscal management, macroeconomic indicators, external sector indicators, or variables measuring the countries’ institutional frameworks. The dataset also includes all maturities in our sample and the spreads available from Du et al. (2016, 2018). ND-GAIN Vulnerability Health" (6) Food"(6) Ecosys." (6) Habitat" (6) Water" (6) Infra." (6) Readiness Economic" (1) Govern." (4) Social"""""""""" (4) ! 28" similarly find that vulnerability to climate change positively influences sovereign borrowing costs. Notably, our study diverges from Beirne, Renzhi, and Volz (2021) in highlighting the importance of climate change readiness. Contrary to their emphasis on vulnerability, our results suggest that the positive effects of vulnerability can be offset by proportionate increases in climate change preparedness. This is evident in the effects of increments in the readiness indicator, which generally mirror the magnitude (but with opposite signs) of vulnerability effects. The only exception is the 90th quantile of the 2-year spread, where vulnerability’s impact is most pronounced within our sample. Table 3. Panel A. Determinants of Yield Spreads at Short Maturities Spread 1 year Quantile=0.9 Quantile=0.5 Quantile=0.1 Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Common factor -0.07 0.01 -4.89 0.00 -0.08 0.01 -8.03 0.00 -0.05 0.01 -5.90 0.00 Readiness -0.15 0.05 -2.86 0.00 -0.12 0.06 -2.11 0.03 -0.07 0.05 -1.44 0.15 Vulnerability 0.14 0.05 2.67 0.01 0.14 0.05 2.79 0.01 0.07 0.05 1.40 0.16 Economic complexity -0.17 0.05 -3.59 0.00 -0.19 0.04 -4.93 0.00 -0.24 0.04 -5.55 0.00 Real GDP growth -0.09 0.02 -5.25 0.00 -0.09 0.02 -4.25 0.00 -0.09 0.02 -3.96 0.00 Inflation 0.27 0.04 6.71 0.00 0.17 0.04 4.75 0.00 0.18 0.04 5.13 0.00 Terms of trade -0.06 0.02 -3.27 0.00 -0.10 0.02 -4.10 0.00 -0.04 0.02 -2.14 0.03 Rents (% GDP) 0.00 0.03 -0.03 0.96 0.09 0.03 3.18 0.00 0.04 0.03 1.02 0.30 Rule of law -0.07 0.06 -1.21 0.22 -0.03 0.07 -0.40 0.67 -0.01 0.07 -0.10 0.90 Debt (% GDP) 0.03 0.04 0.73 0.46 -0.01 0.02 -0.61 0.53 0.05 0.03 1.86 0.06 Spread 2 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Common factor -0.05 0.01 -3.92 0.00 -0.06 0.01 -9.75 0.00 -0.03 0.01 -5.84 0.00 Readiness -0.12 0.06 -2.14 0.03 -0.10 0.04 -2.32 0.02 -0.07 0.04 -1.60 0.11 Vulnerability 0.27 0.07 4.14 0.00 0.13 0.04 3.13 0.00 0.07 0.03 2.34 0.02 Economic complexity -0.16 0.06 -2.93 0.00 -0.18 0.04 -4.43 0.00 -0.07 0.02 -3.83 0.00 Real GDP growth -0.07 0.02 -3.61 0.00 -0.05 0.01 -3.68 0.00 -0.06 0.02 -3.40 0.00 Inflation 0.21 0.03 7.69 0.00 0.12 0.02 5.23 0.00 0.08 0.01 5.69 0.00 Terms of trade -0.06 0.03 -2.27 0.02 -0.05 0.02 -2.17 0.03 -0.04 0.01 -2.85 0.00 Rents (% GDP) -0.01 0.04 -0.19 0.83 0.04 0.02 2.35 0.02 0.04 0.02 2.19 0.03 Rule of law -0.01 0.10 -0.07 0.93 -0.06 0.06 -0.98 0.32 -0.10 0.04 -2.56 0.01 Debt (% GDP) 0.11 0.05 2.43 0.01 0.08 0.02 3.39 0.00 0.02 0.02 0.98 0.32 Note: The table shows the effect of the determinants of sovereign yield spreads with respect to the US at short maturities (1 year, 2 years) and three quantiles of the spreads distribution (0.1, 0.5 and 0.9). All the variables have been scaled and have zero mean and unit variance, which makes comparison of the effects easier. Significant effects in bold and shadow. Source: Authors’ elaboration. ! 29" There is an intriguing pattern in the effects of vulnerability on sovereign spreads: in Panel A, for short maturities, the impact of vulnerability is more significant at higher quantiles. That is, the substantial effects observed at the highest quantile (e.g., 0.14 and 0.27 for 1- and 2-year yield spreads) diminish to zero for 1-year maturities and 7 percent for 2-year maturities at the lowest quantile. In contrast, in Panel B, the effect is more pronounced for lower quantiles (e.g., 0.12 and 0.13 for five and ten years. respectively) than for the higher quantiles, where the effect is non-significant. This suggests that the market places a high price on climate risk for short-term maturities, particularly for countries with riskier profiles that typically incur higher funding costs. At longer terms, the effects are smaller but disproportionately penalize low-risk countries. That is, while climate change vulnerability is more of a rollover risk, for high-risk countries (typically emerging and low-income countries), it is more of a structural long-term solvency risk for advanced economies, which usually face lower borrowing costs. All in all, our analysis reveals that climate change preparation can mitigate the exposure to climate risk. This is especially evident when focusing on longer maturities and lower quantiles of the spreads, but the attenuation provided by readiness is considerable smaller for high-quantile spreads at shorter maturities. ! 30" Table 3. Panel B. Determinants of Yield Spreads at Longer Maturities Spread 5 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Common factor -0.05 0.01 -4.64 0.00 -0.05 0.01 -4.65 0.00 -0.03 0.01 -4.65 0.00 Readiness -0.09 0.04 -2.02 0.04 -0.21 0.05 -4.30 0.00 -0.12 0.05 -2.63 0.01 Vulnerability 0.05 0.05 1.01 0.30 0.06 0.06 1.07 0.28 0.11 0.03 3.98 0.00 Economic complexity -0.24 0.04 -6.51 0.00 -0.19 0.05 -3.64 0.00 -0.14 0.02 -6.48 0.00 Real GDP growth -0.10 0.02 -5.16 0.00 -0.07 0.02 -2.96 0.00 -0.09 0.02 -4.92 0.00 Inflation 0.20 0.04 4.95 0.00 0.15 0.03 4.76 0.00 0.13 0.02 5.75 0.00 Terms of trade -0.05 0.03 -1.56 0.12 -0.09 0.03 -3.33 0.00 -0.02 0.02 -1.25 0.21 Rents (% GDP) 0.02 0.04 0.55 0.57 0.09 0.04 2.12 0.03 0.08 0.02 4.66 0.00 Rule of law -0.11 0.07 -1.73 0.08 -0.10 0.07 -1.52 0.13 -0.13 0.05 -2.50 0.01 Debt (% GDP) 0.03 0.03 1.17 0.24 0.08 0.04 2.01 0.04 0.06 0.02 2.68 0.01 Spread 10 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Common factor -0.03 0.01 -2.22 0.03 -0.04 0.01 -4.57 0.00 -0.03 0.01 -4.98 0.00 Readiness -0.16 0.05 -3.00 0.00 -0.15 0.05 -3.22 0.00 -0.13 0.04 -3.34 0.00 Vulnerability 0.15 0.10 1.56 0.12 0.08 0.04 2.13 0.03 0.12 0.04 3.34 0.00 Economic complexity -0.23 0.05 -4.83 0.00 -0.23 0.03 -7.11 0.00 -0.13 0.02 -5.94 0.00 Real GDP growth -0.08 0.02 -3.93 0.00 -0.06 0.02 -3.35 0.00 -0.09 0.02 -5.45 0.00 Inflation 0.20 0.05 4.40 0.00 0.13 0.02 5.41 0.00 0.09 0.02 3.95 0.00 Terms of trade -0.05 0.03 -1.60 0.11 -0.06 0.02 -2.39 0.02 -0.04 0.01 -2.93 0.00 Rents (% GDP) 0.04 0.05 0.87 0.38 0.05 0.03 2.07 0.04 0.01 0.02 0.38 0.69 Rule of law 0.03 0.09 0.34 0.72 -0.10 0.07 -1.42 0.15 -0.11 0.04 -2.76 0.01 Debt (% GDP) 0.03 0.04 0.73 0.46 0.07 0.02 3.56 0.00 0.06 0.02 2.57 0.01 Note: The table shows the effect of the determinants of sovereign yield spreads with respect to the US at long maturities (5 years, 10 years) and three quantiles of the spreads distribution (0.1, 0.5 and 0.9). All the variables have been scaled and have zero mean and unit variance, which makes comparison of the effects easier. Significant effects in bold and shadow. Source: Authors’ elaboration. Our findings make a significant contribution by emphasizing the varying effects observed across the quantiles of the spreads distribution. This distinction is crucial for emphasizing the actual risks posed by climate change, especially to emerging and low-income countries, which as shown are different from those of advanced economies. Additionally, our results offer a valuable framework for contextualizing and understanding the mitigating effects of climate change preparation. They are also related to the literature that advocates for differentiating the determinants of short- and long-term maturities. ! 31" Economic Complexity and Growth The economic complexity indicator can be understood as an index of export clusters within the international trade network. Economies characterized by higher complexity demonstrate increased productivity, innovation, and intricate production networks. These attributes facilitate specialization in the production of high-value-added goods, valued in global markets for their lower price volatility and better preparedness for future higher growth trajectories. Conversely, less complex economies often specialize in commodities and low-value-added goods, exposing them to greater market fluctuations, particularly in fiscal revenues. A recent study by Gomez-Gonzalez, Uribe, and Valencia (2023b) establishes that economic complexity is a reliable predictor of future fiscal crises. As a result, it is expected that economies with higher complexity would also experience lower borrowing costs, indicative of a lower risk profile. This expectation is further substantiated by the results presented in Table 3. Our findings extend this understanding by revealing that the impact of economic complexity tends to be more pronounced for higher quantiles, with the exception of the very short 1-year maturity. Across various maturities and quantiles, the effects are both statistically and economically significant, ranging between -0.24 and -0.07. These effects of economic complexity are intricately linked with structural factors, reflecting long-term productivity and growth. Accordingly, we include the annual real growth rate of economies as an explanatory variable. To proxy for shortrun performance and generation of fiscal revenues. This variable consistently proves significant across all maturities and quantiles. In contrast to complexity and most other factors in our model, the effect of growth remains consistently sized in all cases, ranging between -5 and -10%. Terms of Trade, Rents Highlighted by Bulow and Rogoff (1989) and Hilscher and Nosbusch (2010), changes in a country’s terms of trade affect its ability to generate dollar revenue from exports, thereby influencing its capacity to meet obligations on externally denominated debt in dollars. The volatility in terms of trade holds significance for the broader economy as well. Terms of trade play an essential role in explaining ! 32" fluctuations in output at business cycle frequencies, as stressed by Mendoza (1995), and have adverse effects on long-term economic growth, as per Mendoza (1997). In a related context, for countries dependent on commodities, in addition to the effects of terms of trade, the unpredictability of export revenues stemming from high volatility in commodity prices is also expected to impact sovereign yields (e.g., Céspedes and Velasco, 2012; Igan et al., 2022; Van der Ploeg and Poelhekke, 2009), demanding separate consideration. However, the expected sign of rents remains ambiguous, as more commodity-dependent countries tend to exhibit more volatile growth trajectories. On the other hand, higher rents should facilitate the repayment of sovereign obligations. Therefore, both negative and positive signs could be justified. According to our results, a modest and negative impact, ranging between - 4% and -10%, is attributable to terms of trade. This effect is consistently significant for short-term debt in Panel A, while for Panel B it holds significance, especially at the center of the distribution. The magnitude of this effect remains relatively stable across quantiles. Conversely, natural resource rents, expressed as a percentage of a country’s GDP, result in a marginal increase in spreads. However, this effect consistently proves to be very slight, with most instances, as detailed in Table 3, Panel A, not reaching statistical significance. Institutions Institutional factors have been identified as key contributors to variations in crosscountry credit risk. Notably, Eichler (2014) presents evidence suggesting that a higher level of political stability and the capacity to enforce austerity measures significantly reduce sovereign yield spreads. Cole and Kehoe (1995) explore one theoretical foundation for this association, arguing that the effectiveness of reputation in supporting debt is intricately linked to a country’s institutional framework. Specifically, in cases where bankers are permitted to default on payments owed to governments, nurturing a positive relationship with bankers confers lasting benefits on the government, enabling substantial borrowing supported by its reputation. In contrast, if bankers are obligated to honor contracts, the government experiences only transient benefits from cultivating a positive ! 33" relationship, and its reputation can sustain minimal (or even zero) borrowing. Following this line of reasoning, this mechanism is anticipated to influence not only the quantity of credit extended to the government but also its pricing. In our findings, the institutional quality of a country, gauged by the Rule of Law estimate in the Worldwide Governance Indicators (Kaufmann, Kraay, and Mastruzzi, 2010), exhibits the anticipated negative sign as per theoretical expectations. However, it only proves statistically significant for long maturities and at lower quantiles. This underscores its role as a long-term structural determinant, particularly in market scenarios characterized by low volatility. Lastly, in Table 3, Panel C, we present the results associated with the idiosyncratic components in our panel quantile model. In this specification, the effects adhere to tradition, as we include only a constant in modeling the effects for all quantiles. For example, for the 1-year spreads at the 0.9 quantile, the results reveal a clear differentiation into three groups. The first group centers around spreads with a mean of zero (including those below the average value), the second encompasses countries close to spreads around 0.36 standard deviations, and a high-risk group clusters around 1.17 standard deviations. As we progress to the right in the table, the estimated location parameters consistently shift to the left, and the groups cluster around negative values at the 0.1 quantile. The estimated parameters exhibit similarity across maturities, with one exception for a 2-year spread, where a notably high component is recorded at the 90th quantile, possibly indicative of outliers. This observation further underscores the motivation behind our approach. It’s essential to highlight that one of the significant advantages of quantile regressions lies in their robustness to outliers, given that they are constructed based on order statistics. In the subsequent section, we introduce additional explanatory variables for these location parameters, emphasizing the role of natural disasters in determining yield spreads. ! 34" Table 3. Panel C. Idiosyncratic Components Results Spread 1 year Quantile=0.9 Quantile=0.5 Quantile=0.1 Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Component 1 0.00 0.02 0.18 0.84 -0.28 0.03 -8.47 0.00 -0.67 0.04 -16.75 0.00 Component 2 0.36 0.05 6.95 0.00 0.05 0.03 1.48 0.14 -0.37 0.02 -15.44 0.00 Component 3 1.17 0.23 5.07 0.00 0.70 0.13 5.25 0.00 -0.06 0.08 -0.78 0.43 Spread 2 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Component 1 0.01 0.03 0.22 0.81 -0.25 0.03 -8.20 0.00 -0.56 0.03 -21.80 0.00 Component 2 0.52 0.05 10.64 0.00 0.01 0.04 0.30 0.75 -0.30 0.01 -22.35 0.00 Component 3 8.28 3.37 2.46 0.01 0.22 0.15 1.49 0.13 -0.03 0.07 -0.54 0.58 Spread 5 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Component 1 -0.06 0.04 -1.57 0.11 -0.33 0.04 -8.45 0.00 -0.68 0.03 -23.51 0.00 Component 2 0.24 0.05 5.09 0.00 -0.02 0.03 -0.61 0.53 -0.31 0.02 -19.15 0.00 Component 3 0.95 0.07 14.04 0.00 0.37 0.08 4.49 0.00 -0.08 0.08 -1.04 0.29 Spread 10 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Component 1 -0.02 0.03 -0.66 0.50 -0.30 0.05 -6.52 0.00 -0.64 0.03 -20.14 0.00 Component 2 0.23 0.05 5.09 0.00 -0.05 0.03 -1.60 0.11 -0.31 0.02 -18.02 0.00 Component 3 1.09 1.03 1.05 0.29 0.19 0.04 4.61 0.00 -0.11 0.04 -2.74 0.01 Note: The table displays the idiosyncratic components of the model, which are modeled in all specifications as a mixture of the three distributions described by the location parameters presented in the table. Source: Authors’ elaboration. 4.3. The Effects of Natural Disasters The impact of natural disasters on sovereign risk has garnered recent attention in the literature. According to Mallucci (2022), natural disasters diminish governments’ capacity to borrow from abroad and depress overall welfare. In Mallucci’s framework, disasters are modeled as exogenous shocks to income and are calibrated to replicate the frequency and intensity of major hurricanes in a sample of seven small Caribbean economies. In the absence of disaster risk, sovereign spreads are lower, as disasters constrain governments’ access to financial markets. To our knowledge, we are the first to introduce natural disasters as a determinant of sovereign risk in a comprehensive sample of countries. To this end, we utilized EMDAT indicators, as explained in Table 1. These indicators present high correlation observed within the same category (from 1 to 3) and low correlation between categories. The categories represent different ways of measuring the disasters’ impact (see Table 1). ! 35" Figure 5 illustrates the clustering of variables. It indicates that while the correlation within categories is high, the correlation between different ways of measuring the effects of the disasters is low. For instance, the correlation between disasters estimated as a percentage of economic loss to GDP (category 3), the number of deaths (category 2), or the number of people affected (category 1) is low. The correlation between disasters in category 1 (ndisaster1) and category 2 (ndisaster2) is 0.22; between category 2 and category 3 (ndisaster3) is 0.17 and it is 0.23 between the second and third categories. Figure 5. Correlation among Natural Disasters Variables Note: The figure shows the correlation among different proxies for natural disasters considered in the literature, which differ in the way that disaster intensity is measured. Source: Authors’ elaboration. From this information we decide to incorporate the three disaster variables simultaneously into the location equation of the idiosyncratic country-specific components of our model. That is, as components of 𝑧12 in equation 10. In this way, our model is able to capture the heterogeneous characteristics of disaster occurrences and considers the impact on the location of the yield distribution across quantiles and maturities. 0.84 0.66 0.26 0.18 0.19 0.25 0.28 0.14 0.08 0.16 0.1 0.84 0.89 0.22 0.23 0.24 0.21 0.23 0.17 0.12 0.19 0.14 0.66 0.89 0.25 0.27 0.28 0.15 0.18 0.15 0.15 0.16 0.17 0.26 0.22 0.25 0.72 0.74 0.4 0.42 0.23 0.23 0.23 0.24 0.18 0.23 0.27 0.72 0.98 0.27 0.28 0.22 0.22 0.33 0.31 0.19 0.24 0.28 0.74 0.98 0.28 0.29 0.23 0.23 0.32 0.32 0.25 0.21 0.15 0.4 0.27 0.28 0.99 0.56 0.54 0.55 0.53 0.28 0.23 0.18 0.42 0.28 0.29 0.99 0.55 0.54 0.55 0.54 0.14 0.17 0.15 0.23 0.22 0.23 0.56 0.55 0.99 0.98 0.97 0.08 0.12 0.15 0.23 0.22 0.23 0.54 0.54 0.99 0.96 0.98 0.16 0.19 0.16 0.23 0.33 0.32 0.55 0.55 0.98 0.96 0.98 0.1 0.14 0.17 0.24 0.31 0.32 0.53 0.54 0.97 0.98 0.98 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 ndisaster3_geophysical ndisaster3 ndisaster3_weather ndisaster2_geophysical ndisaster2 ndisaster2_weather ndisaster1_geophysical ndisaster_geophysical ndisaster1 ndisaster1_weather ndisaster ndisaster_weather ndisaster3_geophysical ndisaster3 ndisaster3_weather ndisaster2_geophysical ndisaster2 ndisaster2_weather ndisaster1_geophysical ndisaster_geophysical ndisaster1 ndisaster1_weather ndisaster ndisaster_weather ! 36" The results of the new model, following this strategy, are presented in Table 5, equivalent to Table 3 but explicitly considering the occurrence of natural disasters in determining cross-country heterogeneity in sovereign debt markets. While the overall results remain very consistent, there is a noticeable improvement in the models, particularly for 5-year maturities, with an increase in the number of statistically significant variables. Green highlights in Table 5 indicate variables that become significant compared to Table 3 in this new specification. Only a couple of variables seem to be less significant (highlighted in red in Table 3), notably the debt-to-GDP ratio, which loses its significance in two specifications and gains significance only on one occasion. The Rule of Law becomes significant on five occasions, while losing its significance in one. Overall, the model adjustment appears to improve with the incorporation of natural disasters, with no changes in the magnitudes or signs of the effects provided in Table 3 and discussed earlier. Table 5. Models Including Natural Disasters Spread 1 year Quantile=0.9 Quantile=0.5 Quantile=0.1 Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Estimate Std.Error z.value P(>|z|) Common factor -0.06 0.02 -3.73 0.00 -0.06 0.01 -5.58 0.00 -0.04 0.01 -5.12 0.00 Readiness -0.16 0.04 -4.51 0.00 -0.21 0.06 -3.45 0.00 -0.09 0.04 -2.15 0.03 Vulnerability 0.11 0.05 2.50 0.01 0.11 0.04 2.66 0.01 0.13 0.03 4.06 0.00 Economic complexity -0.22 0.04 -4.98 0.00 -0.20 0.04 -4.82 0.00 -0.04 0.03 -1.65 0.10 Real GDP growth -0.08 0.02 -3.94 0.00 -0.08 0.02 -3.97 0.00 -0.09 0.02 -3.79 0.00 Inflation 0.28 0.04 7.32 0.00 0.20 0.04 5.12 0.00 0.15 0.03 5.32 0.00 Terms of trade -0.05 0.03 -2.05 0.04 -0.09 0.03 -3.21 0.00 -0.08 0.02 -3.96 0.00 Rents (% GDP) -0.02 0.03 -0.57 0.55 0.10 0.04 2.77 0.01 0.07 0.03 2.41 0.02 Rule of law -0.04 0.05 -0.88 0.37 -0.08 0.08 -1.03 0.30 -0.11 0.04 -2.85 0.00 Debt (% GDP) 0.00 0.04 0.11 0.89 0.06 0.03 2.04 0.04 0.01 0.02 0.58 0.55 Spread 2 Years Quantile=0.9 Quantile=0.5 Quantile=0.1 Common factor -0.06 0.01 -5.44 0.00 -0.06 0.01 -8.67 0.00 -0.04 0.01 -4.84 0.00 Readiness -0.04 0.05 -0.90 0.36 -0.13 0.05 -2.80 0.00 -0.03 0.03 -1.06 0.28 Vulnerability 0.16 0.04 4.28 0.00 0.18 0.05 3.84 0.00 0.15 0.03 4.51 0.00 Economic complexity -0.15 0.05 -2.86 0.00 -0.16 0.04 -4.42 0.00 -0.11 0.02 -5.70 0.00 Real GDP growth -0.07 0.02 -3.93 0.00 -0.06 0.01 -3.99 0.00 -0.06 0.02 -3.58 0.00 Inflation 0.14 0.04 4.00 0.00 0.10 0.02 5.12 0.00 0.10 0.01 8.65 0.00 Terms of trade -0.06 0.02 -3.04 0.00 -0.05 0.02 -2.41 0.02 -0.03 0.01 -2.93 0.00 Rents (% GDP) 0.00 0.05 -0.10 0.90 0.06 0.03 1.93 0.05 0.02 0.02 1.35 0.17 Rule of law -0.11 0.04 -2.43 0.01 -0.01 0.05 -0.23 0.80 -0.06 0.04 -1.72 0.08 Debt (% GDP) 0.09 0.06 1.57 0.11 0.06 0.02 2.73 0.01 0.04 0.01 2.81 0.00 ! 37" Spread 5 years Quantile=0.9 Quantile=0.5 Quantile=0.1 Common factor -0.04 0.01 - 4.41 0.00 -0.05 0.01 - 5.76 0.00 -0.04 0.01 - 5.69 0.00 Readiness -0.13 0.03 - 4.20 0.00 -0.18 0.05 - 3.81 0.00 -0.09 0.04 -2.12 0.03 Vulnerability 0.11 0.03 3.19 0.00 0.10 0.04 2.61 0.01 0.15 0.02 6.13 0.00 Economic complexity -0.21 0.03 - 8.27 0.00 -0.22 0.03 - 7.69 0.00 -0.12 0.02 - 5.78 0.00 Real GDP growth -0.10 0.02 -5.17 0.00 -0.06 0.02 - 2.85 0.00 -0.06 0.02 - 3.59 0.00 Inflation 0.21 0.04 5.92 0.00 0.13 0.03 4.04 0.00 0.13 0.02 5.61 0.00 Terms of trade -0.07 0.02 - 3.90 0.00 -0.09 0.02 - 4.47 0.00 -0.04 0.02 - 2.00 0.04 Rents (% GDP) 0.10 0.03 3.94 0.00 0.07 0.03 2.86 0.00 0.06 0.02 3.29 0.00 Rule of law -0.12 0.04 - 3.04 0.00 -0.17 0.06 - 2.92 0.00 -0.11 0.05 - 2.46 0.01 Debt (% GDP) 0.03 0.02 1.48 0.14 0.06 0.03 2.07 0.04 0.06 0.02 2.65 0.01 ! Spread 10 years Quantile=0.9 Quantile=0.5 Quantile=0.1 Common factor -0.02 0.01 -2.31 0.02 -0.04 0.01 - 3.52 0.00 -0.03 0.01 - 4.93 0.00 Readiness -0.19 0.06 -3.31 0.00 -0.09 0.03 - 3.10 0.00 -0.09 0.03 - 3.06 0.00 Vulnerability 0.13 0.07 1.83 0.07 0.12 0.06 2.20 0.03 0.19 0.03 7.39 0.00 Economic complexity -0.23 0.05 - 4.28 0.00 -0.18 0.03 - 5.92 0.00 -0.13 0.02 - 5.98 0.00 Real GDP growth -0.08 0.02 - 4.59 0.00 -0.07 0.02 - 3.34 0.00 -0.08 0.01 - 5.57 0.00 Inflation 0.28 0.04 6.46 0.00 0.12 0.03 4.31 0.00 0.11 0.02 6.04 0.00 Terms of trade -0.06 0.03 - 2.18 0.03 -0.08 0.02 - 4.57 0.00 -0.05 0.01 - 3.32 0.00 Rents (% GDP) 0.02 0.04 0.46 0.63 0.05 0.03 1.78 0.07 0.02 0.02 1.34 0.18 Rule of law 0.02 0.08 0.26 0.78 -0.17 0.04 - 3.98 0.00 -0.09 0.03 - 3.49 0.00 Debt (% GDP) 0.03 0.05 0.74 0.45 0.03 0.03 1.13 0.25 0.05 0.02 3.22 0.00 Note: The table shows the impact of determinants on sovereign yield spreads across all maturities (ranging from 1 to 10 years) and three quantiles of the spreads distribution (0.1, 0.5, and 0.9). All variables have been standardized to have zero mean and unit variance, facilitating the comparison of effects. Significant effects are denoted in bold and shaded. Notably, variables achieving significance in this updated model specification are highlighted in green, compared to the models in Table 3, which did not incorporate variables for natural disasters. Conversely, variables that lose significance in the new models are highlighted in red. Source: Authors’ elaboration. Table 6 reports the modeling outcomes of the country-idiosyncratic effects, which this time consist of the traditional country-specific effects, and the variations in the location of the spread distributions resulting from the incorporation of natural disasters in 𝑧12. As can be observed, by general rule, natural disasters more often than not are insignificant across quantiles and maturities. From the three variables, ! 44" Bulow, J., and K. Rogoff. 1989. A Constant Recontracting Model of Sovereign Debt. Journal of political Economy, 97(1): 155–178. Caggiano, G., and L. Greco. 2012. Fiscal and Financial Determinants of Eurozone Sovereign Spreads. Economics Letters, 117(3): 774–776. Cahan, E., J. Bai, and S. Ng. 2023. Factor-based Imputation of Missing Values and Covariances in Panel Data of Large Dimensions. Journal of Econometrics, 233(1): 113–131. Candelon, B., and R. Moura. 2023. Sovereign Yield Curves and the COVID-19 in Emerging Markets. Economic Modelling, 127, 106453. Céspedes, L. F., and A. Velasco. 2012. Macroeconomic Performance During Commodity Price Booms and Busts. IMF Economic Review, 60(4): 570–599. Chatterjee, S., and B. Eyigungor. 2019. Endogenous Political Turnover and Fluctuations in Sovereign Default Risk. Journal of International Economics, 117: 37–50. Chatterjee, S., D. Corbae, K. Dempsey, and J. V. Ríos-Rull. 2023. A Quantitative Theory of the Credit Score. Econometrica, 91(5): 1803–1840. Chudik, A., and M. H. Pesaran. 2016. Theory and Practice of GVAR Modelling. Journal of Economic Surveys, 30(1): 165–197. Cole, H. L., and P. J. Kehoe. 1995. The Role of Institutions in Reputation Models of Sovereign Debt. Journal of Monetary Economics, 35(1): 45–64. Dailami, M., P. R. Masson, and J. J. Padou. 2008. Global Monetary Conditions Versus Country-Specific Factors in the Determination of Emerging Market Debt Spreads. Journal of International Money and Finance, 27(8): 1325–1336. De Santis, R. A. 2020. Impact of the Asset Purchase Programme on Euro Area Government Bond Yields Using Market News. Economic Modelling, 86, 192–209. Delgado, R., H. Eguino, and A. Lopes. 2021. Fiscal Policy and Climate Change. Recent Experiences of Finance Ministers in Latin America and the Caribbean. Washington, D.C.: Inter-American Development Bank. Du, W., and J. Schreger. 2016. Local Currency Sovereign Risk. The Journal of Finance, 71(3): 1027–1070. Du, W., J. Im, and J. Schreger. 2018. The US Treasury Premium. Journal of International Economics, 112: 167–181. Eichler, S. 2014. The Political Determinants of Sovereign Bond Yield Spreads. Journal of International Money and Finance, 46, 82–103. Eichler, S., and D. Maltritz. 2013. The Term Structure of Sovereign Default Risk in EMU Member Countries and its Determinants. Journal of Banking & Finance, 37(6): 1810–1816. Freixas, X., and J. C. Rochet. 2008. Microeconomics of Banking. Cambridge, MA: MIT Press. Geraci, M., and M. Bottai. 2007. Quantile Regression For Longitudinal Data Using the Asymmetric Laplace Distribution. Biostatistics, 8(1): 140–154. ! 45" --------. 2014. Linear Quantile Mixed Models. Statistics and Computing, 24, 461-479. Gilchrist, S., B. Wei, V. Z. Yue, and E. Zakrajšek. 2022. Sovereign Risk and Financial Risk. Journal of International Economics, 136: 103603. Gomez-Gonzalez, J. E., J. M. Uribe, and O. M. Valencia. 2023a. Risk Spillovers Between Global Corporations and Latin American Sovereigns: Global Factors Matter. Applied Economics, 55(13): 1477–1496. --------. 203b. Does Economic Complexity Reduce the Probability of a Fiscal Crisis?. World Development, 168: 106250. Gu, S., B. Kelly, and D. Xiu. 2020. Empirical Asset Pricing Via Machine Learning. The Review of Financial Studies, 33(5): 2223–2273. Harding, M., C. Lamarche, and M. H. Pesaran. 2020. Common Correlated Effects Estimation of Heterogeneous Dynamic Panel Quantile Regression Models. Journal of Applied Econometrics, 35(3): 294–314. Hatchondo, J. C., and L. Martinez. 2009. Long-duration Bonds and Sovereign Defaults. Journal of International Economics, 79(1): 117–125. Hidalgo, C. A., and R. Hausmann. 2009. The Building Blocks of Economic Complexity. Proceedings of the National Academy of Sciences, 106(26): 10570– 10575. Hilscher, J., and Y. Nosbusch. 2010. Determinants of Sovereign Risk: Macroeconomic Fundamentals and the Pricing of Sovereign Debt. Review of Finance, 14(2): 235– 262. Igan, D., E. Kohlscheen, G. Nodari, and D. Rees. 2022. Commodity Market Disruptions, Growth and Inflation (No. 54). Basel, Switzerland: Bank for International Settlements. Kaufmann, D., A. Kraay, and M. Mastruzzi. 2010. Response to ‘What do the Worldwide Governance Indicators Measure?’. The European Journal of Development Research, 22, 55–58. Klusak, P., et al. 2023. Rising Temperatures, Falling Ratings: The Effect of Climate Change on Sovereign Creditworthiness. Management Science, 69(12): 7468– 7491. Koenker, R. 2004. Quantile Regression for Longitudinal Data. Journal of Multivariate Analysis, 91(1): 74–89. Koenker, R., and G. Bassett Jr. 1978. Regression Quantiles. Econometrica: Journal of the Econometric Society, 33–50. Kose, M. A., S. Kurlat, F. Ohnsorge, and N. Sugawara. 2022. A Cross-country Database of Fiscal Space. Journal of International Money and Finance 128, 102682. Krishnamurthy, A., S. Nagel, and A. Vissing-Jorgensen. 2018. ECB Policies Involving Government Bond Purchases: Impact and Channels. Review of Finance, 22(1): 1– 44. LeCun, Y., Y. Bengio, and G. Hinton. 2015. Deep Learning. Nature , 521(7553): 436–444. ! 46" Liu, Z., and P. Spencer. 2013. Modelling Sovereign Credit Spreads with International Macro-factors: The Case of Brazil 1998–2009. Journal of Banking & Finance , 37(2): 241–256. Longstaff, F. A., J. Pan, L. H. Pedersen, and K. J. Singleton. 2011. How Sovereign Is Sovereign Credit Risk? American Economic Journal : Macroeconomics, 3(2): 75– 103. Mallucci, E. 2022. Natural Disasters, Climate Change, and Sovereign Risk. Journal of International Economics , 139, 103672. Maltritz, D. 2012. Determinants of Sovereign Yield Spreads in the Eurozone: A Bayesian Approach. Journal of International Money and Finance , 31(3): 657–672. Maltritz, D., and A. Molchanov. 2014. Country Credit Risk Determinants with Model Uncertainty. International Review of Economics & Finance, 29, 224–234. Marino, M. F., and A. Farcomeni. 2015. Linear Quantile Regression Models for Longitudinal Experiments: An Overview. Metron , 73(2): 229–247. Marino, M. F., N. Tzavidis, and M. Alfò. 2018. Mixed Hidden Markov Quantile Regression Models for Longitudinal Data with Possibly Incomplete Sequences. Statistical Methods in Medical Research, 27(7): 2231–2246. Mati, M. A., M. E. Baldacci, and M. S. Gupta. 2008. Is it (still) Mostly Fiscal? Determinants of Sovereign Spreads in Emerging Markets. Washington, D.C.: International Monetary Fund. Matsumura, M. S., and J. V. M. Vicente. 2010. The role of Macroeconomic Variables in Sovereign Risk. Emerging Markets Review, 11(3): 229–249. Mendoza, E. G. 1995. The Terms of Trade, the Real Exchange Rate, and Economic Fluctuations. International Economic Review, 101–137. --------. 1997. Terms-of-trade Uncertainty and Economic Growth. Journal of Development Economics, 54(2): 323–356. NGFS (Network for Greening the Financial System). 2019. Macroeconomic and Financial Stability Implications of Climate Change. Central Banks and Supervisors. https://www.ngfs.net/sites/default/files/medias/documents/ngfsreport-technical-supplement_final_v2.pdf Poghosyan, T. 2014. Long-run and Short-run Determinants of Sovereign Bond Yields in Advanced Economies. Economic Systems, 38(1): 100–114. Powell, A., and O. M. Valencia. 2023. Dealing with Debt: Less Risk for More Growth in Latin America and the Caribbean. In Dealing with debt: less risk for more growth in Latin America and the Caribbean. [Washington, DC]: Inter- American Development Bank. Sánchez, J. M., H. Sapriza, and E. Yurdagul. 2018. Sovereign Default and Maturity Choice. Journal of Monetary Economics, 95, 72–85. Schuler, P., L. E. Oliveira, G. Mele, and M. Antonio. 2019. Managing the Fiscal Risks Associated with Natural Disasters. In M. A. Pigato (Ed.), In Fiscal Policies for Development and climate action (pp. 133–153). Washington, DC: World Bank. ! 47" Stekhoven, D. J., and P. Bühlmann. 2012. MissForest—non-parametric Missing Value Imputation for Mixed-Type Data. Bioinformatics , 28(1): 112–118. Stock, J. H., and M. W. Watson. 2002. Forecasting using Principal Components From a Large Number of Predictors. Journal of the American Statistical Association , 97(460): 1167–1179. Stock, J. H., and M. W. Watson. 2011. in Clements, M. P., & Hendry, D. F. (Eds.). Dynamic Factor Models. The Oxford Handbook of Economic Forecasting. OUP USA. Van der Ploeg, F., and S. Poelhekke. 2009. Volatility and the Natural Resource Curse. Oxford Economic Papers , 61(4): 727–760. Volz, U., et al. 2020. Climate Change and Sovereign Risk. SOAS University of London, Asian Development Bank Institute, World Wide Fund for Nature Singapore, Four Twenty Seven. Xiong, R., and M. Pelger. 2023). Large Dimensional Latent Factor Modeling with Missing Observations and Applications to Causal Inference. Journal of Econometrics , 233(1): 271–301. Yu, K., and R. A. Moyeed. 2001. Bayesian Quantile Regression. Statistics & Probability Letters , 54(4): 437–447. ! 48" Appendix Table A1 Indicator Abreviation Source Mean Median Std.Dev Max. Min. Fiscal rule rule IMF; Schaechter et al. (2012) 0.27 0 0.44 1 0 population in millions pop World Bank 37.17 8.07 134.42 1433.78 0.04 Dummy variable that takes the value of 1 for a fiscal crisis year fiscal_crisis Medas et al. 2018 until 2015, from 2015 own elaboration 0.34 0 0.47 1 0 Dummy variable that takes the value of 1 for serial defaulters serial_defau lt Argentina and Greece 0.02 0 0.13 1 0 Dummy variable that takes the value of 1 forresource rich economies resource_ric h_imf Mlachila and Ouedraogo (2020) 0.39 0 0.49 1 0 Gross capital formation, % GDP gkf IMF 23.73 21.96 16.02 442.77 - 39.73 Gross fixed capital formation, % GDP gfkf IMF 22.28 20.88 12.6 319.06 0 Human capital index hc Penn World Tables 2.28 2.25 0.72 4.35 1.01 Log of per capita real consumption ccon Penn World Tables 12.45 12.44 1.49 16.19 9.06 Real domestic absorption, at current PPPs (in mil. 2017US$) cda Penn World Tables 10.87 10.74 2.11 16.88 5.43 Expenditure-side real GDP at current PPPs (in mil. 2017US$) cgdpe Penn World Tables 10.82 10.7 2.17 16.85 5.19 Output-side real GDP at current PPPs (in mil. 2017US$) cgdpo Penn World Tables 10.83 10.7 2.17 16.84 5.21 Capital stock at current PPPs (in mil. 2017US$) cn Penn World Tables 11.97 11.91 2.38 18.44 5.49 Capital services levels at current PPPs (USA=1) ctfp Penn World Tables 0.67 0.67 0.26 1.9 0.05 Real internal rate of return irr Own estimates 0.11 0.09 0.08 1.1 0.01 Nominal exchange rate, end period trm_end Blomberg 444.6 5 6.47 2134.76 42000 0 Change nominal exchange rate, end period change_trm Own estimates 613.4 2 1.94 17530.56 1314185 - 4277 3 Exchange rate, national currency/USD (market+estimated) xr IMF 12572 .24 6.45 961340 763699 42 0 Trade openness index, (exports+imports)/GDP openness Own estimate, data from IMF 73.35 60.14 51.41 402.32 0.14 Financial openness, Chinn-Ito index kaopen Chinn and Ito Web page 0.06 -0.15 1.55 2.32 -1.92 Exports Diversification Index diversificati on UNCTAD 0.67 0.71 0.15 0.94 0.23 Exports Concentration Index concentrati on UNCTAD 0.34 0.29 0.22 0.99 0.04 Interest payment % GDP, primary balance - overall balance interest WEO (October 2019) 1.9 1.46 2.45 17.71 - 35.48 Implicit interest rate, Interest Payment / Debt interest_rat e2 Own estimate, data from IMF 3.19 3.13 3.36 11.5 - 34.82 Primary balance % of GDP, general government primary_bal ance IMF -0.56 -0.66 6.46 126.46 - 186.7 9 ! 49" Indicator Abreviation Source Mean Median Std.Dev Max. Min. Overall balance % of GDP, general government total_balan ce IMF -2.41 -2.47 6.52 125.14 - 151.31 Pop 65+/ Pop 15-65 ratio_old World Bank 11.09 7.83 7.16 48.64 0.8 GDP constant prices, domestic currency gdp_r IMF 12153 6 522.92 939712 1511298 6 0.06 General government revenue, % GDP revenue IMF 27.88 25.06 14.01 164.05 0.04 Domestic currency debt % total debt p_dd IMF 45.72 41.95 29.25 100 0 Foreign currency debt % total debt p_fd IMF 54.28 58.05 29.25 100 0 Oil rents (% of GDP) oil_rents World Bank 3.55 0 9.26 71.49 0 Coal rents (% of GDP) coal_rents World Bank 0.3 0 2.49 69.8 0 Forest rents (% of GDP) forest_rents World Bank 2.08 0.32 4.17 44.6 0 Mineral rents (% of GDP) mineral_ren ts World Bank 0.78 0.01 2.45 39.67 0 Natural gas rents (% of GDP) gas_rents World Bank 0.37 0 2.34 68.68 0 Fractionalization Index frac Drazanova (2019) 0.52 0.58 0.28 1 0 Polarization Index polariz The Polarization Index 0.4 0 0.77 2 0 Voice and Accountability, Estimate vae World Bank -0.05 -0.04 1.07 4.28 -5.78 Voice and Accountability, Percentile Rank (0-100) var World Bank 48.32 48.28 31.02 183.87 - 85.12 Political Stability and Absence of Violence/Terrorism, Estimate pve World Bank -0.08 0.05 1.35 6.5 -7.92 Political Stability and Absence of Violence/Terrorism, Percentile Rank (0-100) pvr World Bank 47.58 47.09 40.43 269.68 - 137.7 7 Government Effectiveness, Estimate gee World Bank -0.08 -0.17 1.11 3.92 -4.27 Government Effectiveness, Percentile Rank (0-100) ger World Bank 48.23 48.82 35.24 171.65 - 136.2 5 Regulatory Quality, Estimate rqe World Bank -0.09 -0.18 1.19 6.47 -5.78 Rule of Law, Percentile Rank (0- 100) rlr World Bank 48.12 45.54 33.32 235.96 - 64.6 8 Control of Corruption, Estimate cce Penn World Tables -0.05 -0.27 1.14 5.79 -6.04 Control of Corruption, Percentile Rank (0-100) ccr Penn World Tables 48.74 47.81 34.68 186.46 - 134.6 5 Regulatory Quality, Rank rqr Penn World Tables 47.5 47.3 36.96 331.87 - 169.5 3 Interest payment % GDP, primary balance - overall balance interest IMF 1.9 1.46 2.45 17.71 - 35.48 Implicit interest rate, ((debt+primary_balance)*(1+gdp_g rowth)/l.debt-1) interest_rat e1 IMF 10.61 7.09 22.61 202.18 - 104.5 9 Implicit interest rate, Interest Payment / Debt interest_rat e2 IMF 3.19 3.13 3.36 11.5 - 34.82 Dummy variable that takes the value of 1 for year with negative real GDP growth crisis Own elaboration 0.16 0 0.36 1 0 Chicago Board Options Exchange Volatility Index vix Bloomberg 19.49 17.1 6.15 32.7 11.09 ! 50" Indicator Abreviation Source Mean Median Std.Dev Max. Min. Debt spike: 1 if the 5-year change is bigger than the 80th percentile spike Own elaboration 0.15 0 0.36 1 0 Real GDP per capita gdp_pc IMF 2026 463 47125 12147459 181845 616 10.81 Fitch rating, numeric rating_fitch _num Bloomberg 11.42 11 4.95 20 1 Moodys rating, numeric rating_moo dys_num Bloomberg 12.7 12 5.37 21 1 Sp rating, numeric rating_sp_n um Bloomberg 13.29 13 5.46 22 1 Fiscal rule quality, all rules quality_fr Own elaboration based on IMF; Schaechter et al. (2012) 0.21 0 0.61 5 0 Foreign/US govt bond yield spread, end year diff_3m_en d Du et al. 2016, 2018 2.15 1.52 3.55 19.72 -5.64 Foreign/US govt bond yield spread, end year diff_1y_end Du et al. 2016, 2019 2.24 1.55 3.56 18.43 -6.02 Foreign/US govt bond yield spread, end year diff_2y_end Du et al. 2016, 2020 2.16 1.41 3.52 16.8 -5.98 Foreign/US govt bond yield spread, end year diff_3y_end Du et al. 2016, 2021 2.11 1.32 3.46 17.04 -5.89 Foreign/US govt bond yield spread, end year diff_5y_end Du et al. 2016, 2022 1.99 1.19 3.34 16.29 -5.29 Foreign/US govt bond yield spread ,end year diff_7y_end Du et al. 2016, 2023 1.84 1.11 3.17 15.3 -5.15 Value sovereign spread with respect to the US 20 years maturity ValSpread_ 20Y Bloomberg, own elaboration 2.04 0.12 5.28 65.52 -4.03 Value sovereign spread with respect to the US 30 years maturity ValSpread_ 30Y Bloomberg, own elaboration 0.69 -0.02 2.88 31.78 -3.49 ! 51" Figure A1. Sovereign Spreads Missing Observations by Country Note: The figure shows the frequency of missing observations per country in our sample for 1 year, 2 years, 5 years and 10 years maturities. Countries or maturities with a larger amount of missing data were excluded from the analysis. The number of missing data per country represents years and the total number of years in the sample is 20. Source: Authors’ elaboration. UGA UKR VNM ZAF ZMB QAT ROU RUS SGP SVK SVN SWE THA TUR NLD NOR NZL PAK PAN PER PHL POL PRT LKA LTU LVA MAR MEX MUS MYS NAM NGA IRL ISL ISR ITA JPN KAZ KEN KOR LBN ESP FIN FRA GBR GRC HRV HUN IDN IND CHL CHN COL CRI CYP CZE DEU DNK EGY AUS AUT BEL BGD BGR BRA BWA CAN CHE 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y year ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y # Missing Variables ! 52" Figure A2. Sovereign spreads missing observations by year Note: The figure shows the frequency of missing observations per year in our sample for 1 year, 2 years, 5 years and 10 years maturities. Years and maturities with a larger amount of missing data were excluded from the analysis. The number of missing data per year represents countries and the total number of countries in our sample is 68. Source: Authors’ elaboration. 2015 2016 2017 2018 2019 2010 2011 2012 2013 2014 2005 2006 2007 2008 2009 2000 2001 2002 2003 2004 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 ISO3 ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y ISO3 ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y ISO3 ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y ISO3 ValSpread_5Y ValSpread_10Y ValSpread_2Y ValSpread_1Y # Missing Variables