The cyclicality of official bilateral lending: Which cycle do flows follow?

Avellán, Leopoldo; Galindo Andrade, Arturo José; Gómez, Tomás; Lotti, Giulia Working Paper The cyclicality of official bilateral lending: Which cycle do flows follow? IDB Working Paper Series, No. IDB-WP-1383 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Avellán, Leopoldo; Galindo Andrade, Arturo José; Gómez, Tomás; Lotti, Giulia (2022) : The cyclicality of official bilateral lending: Which cycle do flows follow?, IDB Working Paper Series, No. IDB-WP-1383, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0004637 This Version is available at: https://hdl.handle.net/10419/289982 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-nc-nd/3.0/igo/legalcode The Cyclicality of Official Bilateral Lending: Which Cycle Do Flows Follow? Leopoldo Avellán A rturo J. Galindo Tomás Gómez Giulia Lotti IDB WORKING PAPER SERIES Nº IDB-WP-1383 December 2022 Department of Research and Chief Economist Inter-American Development Bank December 2022 The Cyclicality of Official Bilateral Lending: Which Cycle Do Flows Follow? Leopoldo Avellán A rturo J. Galindo Tomás Gómez Giulia Lotti Inter-American Development Bank Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library The cyclicality of official bilateral lending: which cycle do flows follow? / Leopoldo A vellán, Arturo J. Galindo, Tomás Gómez, Giulia Lotti. p. cm. — (IDB Working Paper Series ; 1383) Includes bibliographic references. 1. Debtor and creditor-Econometric models. 2.Capital movements-Econometric models. 3. International finance-Econometric models. 4. Debts, Public-Econometric models. 5. Financial institutions, International-Econometric models. 6. Debt reliefEconometric models. I. Avellán Leopoldo. II. Galindo, Arturo. III. Gómez, Tomás. IV. Lotti, Giulia. V. Inter-American Development Bank. Department of Research and Chief Economist. VI. Series. IDB-WP-1383 Copyright © Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 AttributionNonCommercial-NoDerivatives (CC-IGO BY-NC-ND 3.0 IGO) license (http://creativecommons.org/licenses/by-nc-nd/3.0/igo/ legalcode) and may be reproduced with attribution to the IDB and for any non-commercial purpose, as provided below. No derivative work is allowed. 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 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 CC-IGO license. Following a peer review process, and with previous written consent by the Inter-American Development Bank (IDB), a revised version of this work may also be reproduced in any academic journal, including those indexed by the American Economic Association's EconLit, provided that the IDB is credited and that the author(s) receive no income from the publication. Therefore, the restriction to receive income from such publication shall only extend to the publication's author(s). With regard to such restriction, in case of any inconsistency between the Creative Commons IGO 3.0 Attribution-NonCommercial-NoDerivatives license and these statements, the latter shall prevail. Note that link provided above includes additional terms and conditions of the license. The opinions expressed in this publication 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. http://www.iadb.org 2022 Abstract1 Using a large panel of official bilateral loan data for 111 borrowing countries and 78 lending countries between 1980 and 2020, this paper shows that international government borrowing from bilateral sources is acyclical with respect to the economic cycle of the borrower, but procyclical with respect to the cycle of the lending country. This holds in the case of loans both from advanced economies and from China, currently the largest supplier of official bilateral lending to the average developing country. We find this form of procyclicality most often among middleincome recipient countries across most regions of the world. We also find that bilateral loans follow economic links captured through bilateral trade, and political ties measured by the alignment of votes in the United Nations. The results are consistent across a battery of robustness tests. JEL classifications: E60, F32, F34 Keywords: Bilateral debt, Cyclicality, Capital flows, International government debt 1 The views in this paper are exclusively those of the authors and do not represent those of the Inter-American Development Bank or its Board of Directors. E-mail addresses: [email protected] (L. Avellán), [email protected] (A. J. Galindo), [email protected] (T. Gómez), and [email protected] (G. Lotti). 2 1. Introduction Official (i.e., government-to-government) lending is a significant source of financing for many countries, particularly developing ones. Most official lending is directed to government counterparties and made up of loans provided by international financial institutions that include multilateral development banks (MDBs)2 and the International Monetary Fund, and bilateral loans from governments or government-owned institutions.3 This study focuses on bilateral loans, particularly in their role in smoothing procyclical private capital flows. As such, the study builds on the literature on the cyclicality of capital flows to emerging economies and also contributes to the smaller body of literature on the behavior of official bilateral flows. There is broad consensus on the procyclicality of capital flows in emerging markets and how such flows can harm economies on the receiving end by being additional sources of volatility in troubling times.4 This is particularly true for capital flows that come from the private sector.5 The main argument found in the literature is that when emerging markets are hit by shocks that push them into negative growth territory, private capital flows retract, accelerating and increasing the deleterious effects of the initial shock. By contrast, when economies are booming, capital flows increase and the risk of creating asset price bubbles, among other problems, increases as well. Not all capital flows are procyclical, however. Evidence also suggests that certain types of capital flows are countercyclical and can partially counteract the procyclical nature of private flows. These countercyclical flows are most often provided as official development assistance through MDBs. Using a panel data set for more than 130 developing countries between 1980 and 2015, Galindo and Panizza (2018) find that MDB lending is countercyclical with respect to the receiving country’s gross domestic product (GDP) cycle, and present additional evidence of the high procyclicality of private flows.6 Their findings are supported by supply and demand considerations. When countries experience economic slowdowns, their governments typically face greater expenditure pressures but tighter financing constraints, leading them to demand more funds 2 MDBs include the World Bank, and regional development banks such as the African Development Bank, the Asian Development Bank, the Inter-American Development Bank, and the European Bank for Reconstruction and Development, among many others. 3 A typical example is a development finance institution in a high-income economy lending resources to the government of a lowor middle-income country to carry out a specific project. 4 See Kaminsky, Reinhart, and Végh (2004) for a discussion. 5 See Galindo and Panizza (2018); Alfaro, Kalemli-Ozcan, and Volosovych (2014); and Dasgupta and Ratha (2000) for discussions. 6 This finding corroborates previous discussions by Levy Yeyati (2009) and Humphrey and Michaelowa (2013). 3 from MDBs (or other official creditors). On the supply side, MDBs have the mandate to support countries in tight financial situations (see Humphrey and Michaelowa, 2013). When demand meets supply, MDBs can increase their lending at low rates given their de facto preferred creditor treatment.7 In addition, if the economic downturn becomes a crisis, countries can access emergency lending from the IMF, which usually will be accompanied by additional lending from MDBs to support the country in resolving the crisis (Avellán, Galindo, and Lotti, 2021). In this sense, MDB lending acts as an insurance instrument for countries since it allows them to access relatively cheap resources during times of trouble when private sources of funding have dried out. While this key characteristic of MDB lending has been deeply studied, to our knowledge there is no evidence about the cyclicality of official bilateral loans, another type of official lending. This paper helps fill that gap. Official bilateral lending was the dominant source of external sovereign funding after the Napoleonic wars of the nineteenth century and the World Wars of the twentieth century, and it was key for the recovery of the countries that suffered most during those episodes. It became important particularly during the ensuing economic and financial crises, and before the creation of the IMF and the World Bank in 1944.8 Since then, and with the more recent creation of multilateral regional development banks, official bilateral loans and MDB loans have coexisted.9 Literature has explored why, despite the efficiency gains of lending through MDBs,10 bilateral lenders have chosen to maintain a direct lending relationship with borrowers.11 Possible benefits include maintaining full control over the intended outcomes of the resources provided and reducing the risk of surrendering specific elements in the design of operations to the criteria used by the MDBs. Also, acting through an MDB dilutes any specific preferences or objectives a bilateral lender may have, particularly political ones.12 Regardless of the reasons, to date, bilateral lending accounts for over a third of official lending to developing countries. 7 See Cordella and Powell (2021), and Schegl, Trebesch, and Wright (2019) for discussions on the preferred creditor treatment of MDBs. 8 For a detailed history of official lending over the past two centuries, see Horn, Reinhart, and Trebesch (2020). 9 The Inter-American Development Bank was created in 1959, the African Development Bank in 1964, the Asian Development Bank in 1966, and the European Bank for Reconstruction and Development in 1991, among others. 10 Lending through MDBs rather than directly can be cost-effective and more efficient since it exploits the leverage and mobilization capacity of MDBs. When borrowing from an MDB instead of from multiple lenders, countries need to comply with the rules and procedures of only one counterpart rather than many. 11 For a discussion, see Bobba and Powell (2006). 12 Bobba and Powell (2006) show that aid from donors to countries that vote the same way in the UN General Council is relatively less effective in supporting development. 4 The amounts lent and the counterparties chosen have been amply studied. The most common explanation of why a bilateral lender, typically a high-income economy, lends directly to another government is that it wants to support something that is valuable for its own economic and/or political stability. This suggests that lending is not conducted for altruistic motives, but rather to avoid negative spillovers of a political or economic crisis in a country that is economically exposed.13 In such situations, bilateral loans may be granted to avoid the collateral damage that economic distress or political instability in the receiving country could cause in the lending country. In this context, the more economically and politically exposed a country is to another, the greater its incentive to offer bilateral loans. Economic exposure has been quantified in various ways by researchers. The measures most often used are bilateral trade relationships and the exposure of the banking system of the lending country to the borrowing one (see Horn, Reinhart and Trebesch, 2020). Greater economic integration is associated with more bilateral lending. Political connections have also been analyzed in the literature as determinants of bilateral lending and bilateral aid. In this context, bilateral loans are provided to countries that are friendly to the political views of the country supplying resources. To proxy this, researchers commonly use the alignment of votes at the United Nations (see Horn, Reinhart and Trebesch, 2020; and Bobba and Powell, 2007). Authors have found that the provision of resources increases as the recipient votes closer to the lending country. Thus, bilateral lending may be viewed as rewarding political alignment. Finally, and in the spirit of gravity models frequently used in the trade literature, geographical and cultural ties are also explanatory factors and are proxied with indicators such as sharing a common language, having colonial links, and sharing a frontier. Again, where any of these ties are stronger, there is more bilateral official lending.14 In addition to the above determinants, this paper explores if bilateral lending also acts as insurance when developing economies face downturns. To our knowledge, this question has not been explored in depth. It has been considered to some degree by Horn, Reinhart and Trebesch (2020), who use a novel database of bilateral lending between 1790 and 2015 to explore the determinants of lending during crisis episodes or natural disasters. They use gravity models and 13 Discussions motivating this rationale can be found in Gourinchas, Martin, and Messer (2020), Farhi and Tirole (2018), and Tirole (2015). 14 See Horn, Reinhart and Trebesch (2020) for evidence on this link. 5 other econometric approaches to show that, during episodes of distress, bilateral lending is higher when there are greater economic and political affinities. The underlying assumption is that when a country is in distress, bilateral lenders will provide funds. The question of our paper is more general: does bilateral lending counteract capital flow cycles? We explore not only extreme episodes as in Horn, Reinhart, and Trebesch (2020), but also what happens at regular upturns and downturns of the cycle. Moreover, we explore if the provision of resources is also affected by the capital flows of the country where resources originate. If a potential lender faces fiscal constraints induced by a recession, it is less likely to offer bilateral loans. To tackle the question of the cyclicality of official bilateral loans, this paper uses a panel data set of 111 developing countries that received such loans and 78 countries that supplied them between 1980 and 2020. We find that official bilateral debt is procyclical with respect to the economic cycle of the lender and acyclical with respect to the economic cycle of the borrower. Where the flows originate matters. Our estimates suggest that flows from China are procyclical when China experiences a contraction, while those from high-income or “advanced” economies are procyclical when those countries experience an expansion. Regardless of the origin, our main result is replicated when the recipient is a middle-income country. That is, in middle-income countries, official bilateral flows are acyclical with respect to the recipient’s cycle and procyclical with respect to the origin’s cycle. For low-income countries, we find evidence of procyclicality in the recipient’s cycle when the flows originate in China, but no evidence of procyclicality in China’s own cycle. By contrast, if the origin is an advanced economy, we see evidence of procyclicality in the lender’s cycle. These results are stronger in middle-income countries than in lower-income ones and hold in most recipient regions of the world. They are also robust to changes in the specification, the inclusion of additional controls, and alternative measures of the economic cycle. In summary our results suggest that, in contrast to official flows supplied by multilateral development banks that tend to be countercyclical or acyclical, and hence work as an insurance mechanism when private capital flows retrench, bilateral loans do not offer this benefit, but on the contrary could bring an additional source of volatility linked to the economic cycle of the lender. These results highlight the importance of diversifying the pool of bilateral lenders, so that borrowers can mitigate volatility while reaping associated benefits. 12 between two countries at a determined time. A score equal to 1 indicates that two countries agree on all votes, while -1 indicates that two countries maximally disagree on all resolutions. Once all the data are merged, we have an unbalanced panel of 1,396 country pairs between 1981 and 2020. There are 111 recipient countries classified as non-advanced (according to the IMF definition) and 78 countries of origin. As robustness exercises, we also control for measures of total indebtedness of the recipient country; for this, we consider total external sovereign indebtedness, total multilateral debt, and total bilateral debt. The source of these data is the same as that of the dependent variable of our study: the IDS of the World Bank. Table 1 reports the key descriptive statistics of the main variables used throughout the empirical analysis. Table 1. Summary Statistics Obs. Mean SD Min p25 Median p75 Max Pair statistics Origin: All Bilateral disbursements (% trend GDP) 27,850 0.12 0.53 0.00 0.00 0.00 0.05 23.40 Log(Pair total trade [% trend GDP]) 27,850 -0.18 1.80 -11.13 -1.08 -0.05 0.94 4.94 Log(Pair UN votes agreement score) 27,850 -0.35 0.30 -2.21 -0.44 -0.31 -0.14 0.00 Origin: Advanced Bilateral disbursements (% trend GDP) 19,554 0.10 0.34 0.00 0.00 0.00 0.04 12.63 Log(Pair total trade [% trend GDP]) 19,554 -0.03 1.53 -7.08 -0.92 -0.03 0.92 4.37 Log(Pair UN votes agreement score) 19,554 -0.45 0.31 -2.21 -0.49 -0.37 -0.29 0.00 Origin: China Bilateral disbursements (% trend GDP) 1,979 0.37 0.87 0.00 0.00 0.04 0.32 10.82 Log(Pair total trade [% trend GDP]) 1,979 0.79 1.30 -5.36 0.02 0.97 1.61 4.34 Log(Pair UN votes agreement score) 1,979 -0.13 0.09 -0.69 -0.17 -0.11 -0.07 0.00 Recipient country statistics Recipient’s cycle 3,752 0.00 0.05 -0.54 -0.02 0.00 0.02 0.23 Recipient’s expansions 3,752 0.01 0.03 0.00 0.00 0.00 0.02 0.23 Recipient’s contractions 3,752 -0.02 0.03 -0.54 -0.02 0.00 0.00 0.00 Log(Total bilateral debt [% trend GDP]) 3,752 2.05 1.38 -5.54 1.33 2.17 2.95 5.54 Log(Total multilateral debt [% trend GDP]) 3,747 2.37 1.21 -7.78 1.84 2.53 3.11 5.41 Log(Total external public debt [% trend GDP]) 3,752 3.44 0.83 -0.74 2.94 3.46 3.98 6.09 Origin statistics Origin’s cycle (all) 1943 0.00 0.05 -0.65 -0.02 0.00 0.02 0.50 Origin’s cycle (advanced) 824 0.00 0.02 -0.09 -0.01 0.00 0.01 0.10 Origin’s cycle (China) 40 0.00 0.03 -0.07 -0.03 0.00 0.02 0.05 13 4. Baseline Results Table 2 reports the baseline results of estimating equation (1) with different structures of fixed effects, controls, and different origins. In column (1), recipient country fixed effects, origin country fixed effects, and year fixed effects are included, to control for global factors that vary in time, and for unobserved invariant characteristics of the recipient and origin countries separately. Bilateral flows do not exhibit a significant relationship with the economic cycle of the recipient country, pointing toward the acyclicality of official bilateral flows. Bilateral flows seem instead to be procyclical with respect to the origin country’s economic cycle, indicating that when a country is experiencing a positive (negative) cycle, its flows toward emerging economies increase (decrease). These results hold when separate recipient-country and origin-country fixed effects are substituted with country-pair fixed effects (column [2]). In column (3), we add controls at the country-pair level that vary in time: total trade between country j and i normalized by trend GDP of the recipient country and the UN voting agreement score, to capture commercial and political proximity. The main result regarding the procyclicality of bilateral flows with respect to the country of origin remains. In addition, we find that bilateral flows are also positively and significantly correlated with total trade between the two countries, indicating that they become larger as trade links become stronger. We do not find a statistically significant relationship between bilateral flows and political affinity, as measured by the UN votes agreement score. In columns (4) and (5), we then explore country heterogeneities by focusing our attention on origin countries that are advanced economies (column [4]) and on China (column [5]). In the case of flows originated in advanced economies, the results mirror those of column (3). Bilateral flows are acyclical with respect to the recipient’s business cycle and procyclical with respect to the origin’s. Bilateral flows are significantly correlated with trade flows. In the case of advanced economies, political ties are statistically significant; official bilateral lending flows increase as countries’ votes at the UN General Assembly align. These results are not only statistically significant but they are also economically relevant. A rise in one standard deviation in the cycle of the origin country increases official bilateral debt flows by 0.06 percentage points, nearly 20 percent of a standard deviation of bilateral flows for this sample (0.34). A one standard deviation increase in bilateral trade between advanced countries and recipients of bilateral debt flows increases debt flows by nearly 0.6 percentage points (almost two times a standard deviation). 14 Similarly, a one standard deviation rise in the proximity of votes increases bilateral debt flows by 0.2 percentage points. In the case of flows originating in China (column [5]), the specification is slightly different. In this specification, we drop the time fixed effects since they would fully correlate with the origin cycle variable. We find that bilateral debt flows are acyclical with respect to the recipient’s cycle, but more strongly correlated to China’s own cycle.24 A one standard deviation increase in China’s business cycle increases debt flows by 0.38 percentage points (over 40 percent of a standard deviation). Trade integration remains significant, both statistically as well as economically. A one standard deviation increase in the trade variable for this sample increases bilateral debt flows by 0.6 percentage points, around two-thirds of a standard deviation of bilateral debt flows in the same sample. We find no significant relationship between bilateral debt flows and the alignment of votes at the UN. Table 2. Baseline Regressions (1) (2) (3) (4) (5) Origin: All Origin: All Origin: All Origin: Adv. Origin: China Recipient’s cycle (t-1) -0.424 0.277 -0.075 -0.598 1.523 (0.509) (0.505) (0.504) (0.583) (0.980) Origin’s cycle (t-1) 3.109*** 2.734*** 2.317** 2.937* 12.535*** (0.948) (0.994) (0.988) (1.776) (3.115) Log(Pair total trade [% trend GDP]) 0.505*** 0.385*** 0.463*** (0.063) (0.072) (0.080) Log(Pair UN votes agreement score) 0.334 0.564** -0.571 (0.220) (0.229) (0.688) Constant -1.199*** 0.818*** 1.368*** 1.337*** -1.182*** (0.049) (0.004) (0.106) (0.133) (0.174) Number of observations 27,850 27,850 27,850 19,554 1,979 Number of pairs 1,396 1,396 1,396 867 95 Pseudo-R-squared 0.28 0.40 0.41 0.37 0.32 Log likelihood -8025.32 -6733.72 -6565.21 -3946.59 -1136.27 Fixed effects Recipient, origin & time Pair & time Pair & time Pair & time Pair Note: This table presents a set of PPML regressions where the dependent variable is bilateral disbursements as a percentage of the recipient’s GDP trend. The cycles are computed as the percent deviation between GDP and trend GDP, where the trend is computed with the Hodrick-Prescott filter. Clustered standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. 24 Given the large estimated coefficient indicating the procyclicality of Chinese bilateral flows with respect to the origin cycle, we also estimate equation (2) where we add an interaction between the origin economic cycle and a dummy with a unit value when the origin country is an advanced economy, or when it is China. When testing whether procyclicality with respect to the origin cycle differs between the two options, we find no significant difference. 15 Table 3 differentiates between episodes of expansion and contraction, defined as the output gap being above or below trend, respectively. Each of the cycle variables (recipient and origin) ise broken into two variables: one capturing expansions, the other capturing contractions. The expansion (contraction) variable takes the value of the cycle variable when the cycle is positive (negative) and zero otherwise. Each column reports results for different origin groups: Column (1) for the whole sample, column (2) for advanced economies, and column (3) for China. Table 3. Expansions and Contractions (1) (2) (3) Origin: All Origin: Advanced Origin: China Recipient’s expansions (t-1) -0.049 -0.295 0.981 (1.188) (0.949) (2.104) Recipient’s contractions (t-1) -0.098 -0.875 1.711 (1.219) (1.084) (2.305) Origin’s expansions (t-1) 1.981 8.919** 4.852 (1.795) (3.542) (4.723) Origin’s contractions (t-1) 2.716 -2.164 23.166*** (1.924) (4.350) (8.560) Log(Pair total trade [% trend GDP]) 0.505*** 0.386*** 0.422*** (0.063) (0.072) (0.074) Log(Pair UN votes agreement score) 0.338 0.565** 0.033 (0.216) (0.229) (0.760) Constant -1.360*** -1.437*** -0.886*** (0.114) (0.129) (0.192) Number of observations 27,850 19,554 1,979 Number of pairs 1,396 867 95 Pseudo-R-squared 0.41 0.37 0.32 Log likelihood -6565.16 -3944.63 -1133.60 Note: This table presents a set of PPML regressions where the dependent variable is bilateral disbursements as a percentage of the recipient’s GDP trend. The cycles are computed as the percent deviation between GDP and trend GDP, where the trend is computed with the Hodrick-Prescott filter. “Expansions” refer to the positive components of those cycles while “Contractions” refer to the negatives. Regressions with all and advanced economy origins include pair and time fixed effects, while those with the origin of China include only pair effects. Clustered standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. When considering the whole sample and separating the cycle into expansions and contractions in column (1), we lose the significance of the origin cycle variable reported in Table 2. However, when focusing on the two origin categories (i.e., advanced economy or China) in 16 columns (2) and (3), the significance of the origin cycle is maintained but the source of this significance varies. Column (2) suggests that the procyclicality of the origin cycle is more notable during expansions in advanced economies, suggesting that the more these economies grow, the greater the volume of official bilateral loans supplied. During contractions, however, flows do not fall. The opposite happens with flows of Chinese origin. During contractions in China, flows supplied by China to emerging and developing countries fall significantly, but when China grows above trend, its supply of bilateral loans does not increase proportionally. While in both cases, the volatility of the economic cycle in the origin countries can be transmitted to the recipient through bilateral lending, how it does so will vary, depending on the origin country and where it is in its own business cycle. 5. Heterogeneity We explore two relevant sources of heterogeneity among recipient countries. First, by income levels and, second, by geographical regions of the world. For the first exercise, we estimate equation (1) for different subsamples defined by the income level of the recipient country, according to the World Bank’s income classification.25 In order to simplify the presentation of the results in Figure 4 we report the coefficients estimated and their 90 percent confidence intervals for each of the estimated subsamples.26 As above, we also estimate each equation for the whole sample of origin countries (Panel A), for advanced economy origin (Panel B), and for Chinese origin (Panel C). As can be seen in Panel A, the procyclicality of bilateral flows with respect to the origin cycle occurs when recipients are middle-income countries, whereas in low-income countries acyclicality seems to prevail. When bilateral flows originate from advanced economies, they still seem to be procyclical with respect to their own cycle when financing both lowand middleincome recipient countries, and acyclical with respect to the recipient economic conditions (Panel B). 25 Appendix D, Table D.1, shows the World Bank’s income classification of each recipient country in 2020. 26 Complete regression results are reported in Appendix C. 17 Figure 4. By Recipient Income Level (a) Origin: All (b) Origin: Advanced (c) Origin: China Note: These plots show the recipient and origin countries’ cycle coefficients from a set of PPML regressions detailed in Table D.1, Appendix D. The dependent variable is bilateral disbursements as a percentage of the recipient’s GDP trend. The cycles are computed as the percent deviation between GDP and trend GDP, where the trend is calculated with the Hodrick-Prescott filter. The recipient’s income group is according to the World Bank classification. The lines represent the 90 percent confidence intervals estimated from clustered standard errors. Regressions of Panels A and B include pair and time fixed effects, while those of Panel C include only pair effects. In Panel C, we observe that Chinese bilateral flows are procyclical with respect to China’s own cycle only when directed toward middle-income countries, but they are procyclical with respect to the economic cycle of recipient countries when these are low income. Given the recent increase in the relevance of Chinese flows for emerging and developing countries, the behavior of Chinese flows might pose a further threat precisely to the most vulnerable low-income countries, since when the latter are expanding, they will receive more flows from China, but when they are 18 experiencing economic downturns, they will confront capital reversals and more limited access to financing, at least as far as Chinese flows are concerned. For a deeper dive into country heterogeneities, we consider whether the patterns previously identified align with geographical regions. To do this, we estimate a modified version of equation (1) that includes interaction terms between the economic cycles of origin and recipient countries and dummy variables with a unit value when bilateral flows originate from different regions. Namely, we estimate: 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑖𝑖,𝑗𝑗,𝑡𝑡=𝛼𝛼+𝛽𝛽𝑐𝑐𝑐𝑐𝑐𝑐𝑏𝑏𝑏𝑏𝑗𝑗,𝑡𝑡−1 +𝛾𝛾𝑐𝑐𝑐𝑐𝑐𝑐𝑏𝑏𝑏𝑏𝑖𝑖,𝑡𝑡−1 +𝜆𝜆𝑐𝑐𝑐𝑐𝑐𝑐𝑏𝑏𝑏𝑏𝑗𝑗,𝑡𝑡−1 ∗𝑏𝑏𝑎𝑎𝑎𝑎𝑏𝑏𝑎𝑎𝑐𝑐𝑏𝑏𝑎𝑎 𝑜𝑜𝑏𝑏𝑏𝑏𝑜𝑜𝑏𝑏𝑎𝑎𝑖𝑖+ 𝜇𝜇𝑐𝑐𝑐𝑐𝑐𝑐𝑏𝑏𝑏𝑏𝑖𝑖,𝑡𝑡−1 ∗𝑏𝑏𝑎𝑎𝑎𝑎𝑏𝑏𝑎𝑎𝑐𝑐𝑏𝑏𝑎𝑎 𝑜𝑜𝑏𝑏𝑏𝑏𝑜𝑜𝑏𝑏𝑎𝑎𝑖𝑖+𝛿𝛿𝑐𝑐𝑐𝑐𝑐𝑐𝑏𝑏𝑏𝑏𝑗𝑗,𝑡𝑡−1 ∗ 𝐶𝐶ℎ𝑏𝑏𝑎𝑎𝑏𝑏 𝑜𝑜𝑏𝑏𝑏𝑏𝑜𝑜𝑏𝑏𝑎𝑎𝑖𝑖+𝜃𝜃𝑐𝑐𝑐𝑐𝑐𝑐𝑏𝑏𝑏𝑏𝑖𝑖,𝑡𝑡−1 ∗ 𝐶𝐶ℎ𝑏𝑏𝑎𝑎𝑏𝑏 𝑜𝑜𝑏𝑏𝑏𝑏𝑜𝑜𝑏𝑏𝑎𝑎𝑖𝑖+𝛿𝛿𝑋𝑋𝑖𝑖,𝑗𝑗,𝑡𝑡+𝜂𝜂𝑖𝑖,𝑗𝑗+𝜗𝜗𝑡𝑡+𝜀𝜀𝑖𝑖,𝑗𝑗,𝑡𝑡 (2) where 𝑏𝑏𝑎𝑎𝑎𝑎𝑏𝑏𝑎𝑎𝑐𝑐𝑏𝑏𝑎𝑎 𝑜𝑜𝑏𝑏𝑏𝑏𝑜𝑜𝑏𝑏𝑎𝑎𝑖𝑖 is an indicator variable equal to 1 when the country i from which flows originate is an advanced economy, and 𝐶𝐶ℎ𝑏𝑏𝑎𝑎𝑏𝑏 𝑜𝑜𝑏𝑏𝑏𝑏𝑜𝑜𝑏𝑏𝑎𝑎𝑖𝑖 is the equivalent when the origin country is China. Thus, 𝛽𝛽+𝜆𝜆 and 𝛾𝛾+𝜇𝜇 will capture the cyclicality of advanced economies’ bilateral flows with respect to the recipient country’s and the origin country’s cycle, respectively, and 𝛽𝛽+𝛿𝛿 and 𝛾𝛾+𝜃𝜃 will capture the cyclicality of bilateral flows with Chinese origin. As above, to simplify the presentation of the results, we report them graphically in Figure 5, focusing on these relevant parameters. Panel A reports the coefficient in equation (2) associated with the recipient’s cycle for each subregion of the world and each set of origin countries, and Panel B shows the results for the coefficient associated with the origin country’s cycle. In the case of flows from advanced countries, Panel A of Figure 5 suggests that the result of acyclicality with respect to the recipient’s cycle is confirmed in all regions. In the case of bilateral loans from China, the evidence is mixed and there is evidence of procyclicality with respect to the recipient’s cycle in Europe and Central Asia, South Asia, and Sub-Saharan Africa. 19 Figure 5. By Geographical Region A. Recipient’s Cycle B. Origin Country’s Cycle Note: These plots show the overall recipient and origin’s cycle joint coefficients from a set of PPML regression following equation (2) and reported in Table D.2, Appendix D. The dependent variable is bilateral disbursements as a percentage of the recipient’s GDP trend. The cycles are computed as the percent deviation between GDP and trend GDP, where the trend is calculated with the Hodrick-Prescott filter. The recipient’s geographical group is according to the World Bank’s classification: East Asia and Pacific (EAP), Europe and Central Asia (ECA), Latin America and the Caribbean (LAC), Middle East and North Africa (MENA), South Asia (SA), and Sub-Saharan Africa (SSA). The lines represent the 90 percent confidence intervals estimated from clustered standard errors. All regressions include pair and time fixed effects. 20 With respect to the cycle of the origin country, Panel B of Figure 5 shows that procyclicality dominates the picture. When considering advanced economy origins, we estimate procyclicality for recipients in East Asia and Pacific, the Middle East and North Africa, and SubSaharan Africa. In the case of flows originating in China, we find evidence of procyclicality in all cases except Latin America and the Caribbean and the Middle East and North Africa. In sum, flows tend to be acyclical with respect to the recipient country, except for bilateral flows from China toward Europe and Central Asia, South Asia, and Sub-Saharan Africa. In the cases of East Asia and Pacific and Sub-Saharan Africa, flows are always procyclical regardless of their origin, and, in the case of Latin America and the Caribbean, acyclical. In other regions of the world, the evidence is mixed, but some degree of procyclicality is consistently present. 6. Robustness Checks We perform three sets of robustness sets. First, we control for potential omitted variable bias and include additional regressors in our baseline estimations. Second, we use alternative methodologies in the estimation of the business cycles, the key feature in our research. Finally, we replace the country-pair fixed effects with time-invariant country-pair regressors frequently used in the trade literature. 6.1 Omitted Variables Our results could suffer from omitted variable bias. The fixed effects in our estimated equations control for global shocks that affect all countries and for everything that is invariant in time but constant at the country-pair level. Hence, if there was an omitted variable correlated with both our dependent variables and our independent variables of interest that varied in time at the country level, our estimated coefficients would be biased. To mitigate this concern, we test whether our baseline results hold to the introduction of other controls that vary at the country/year level in Table 4. We test if results are robust when controlling for the lagged values of accumulated stock of bilateral debt, multilateral debt, and the total external debt stock, public and publicly guaranteed, as the debt of the recipient country might influence the decision of other countries to finance or not.27 All of these variables are normalized 27 We extract these variables from the IDS. Public and publicly guaranteed debt comprises long-term external obligations of public debtors, including national governments, public corporations, state-owned enterprises, 21 by the trend GDP of the recipient country and expressed in logs. As shown in columns (1)–(3), results are not affected by these controls as bilateral flows are still procyclical with respect to the origin cycle and acyclical with respect to the recipient country. Interestingly, for the complete sample of countries of origin, the debt stocks are not statistically significant. Table 4. Controlling for Debt Stocks (1) (2) (3) Debt: Bilateral Debt: MDB Debt: Total External Recipient’s cycle (t-1) -0.255 -0.285 -0.267 (0.513) (0.517) (0.510) Origin’s cycle (t-1) 2.006** 2.109** 2.003** (0.953) (0.978) (0.956) Log(Pair total trade [% trend GDP]) 0.481*** 0.487*** 0.483*** (0.060) (0.060) (0.059) Log(Pair UN votes agreement score) 0.289 0.290 0.305 (0.215) (0.207) (0.209) Log(Debt stock [% trend GDP]) (t-1) 0.035 0.048 -0.000 (0.074) (0.063) (0.089) Constant -1.457*** -1.500*** -1.355*** (0.221) (0.168) (0.318) Number of observations 27,832 27,786 27,835 Number of pairs 1,396 1,393 1,396 Pseudo-R-squared 0.41 0.41 0.41 Log likelihood -6533.75 -6497.09 -6534.96 Note: This table presents a set of PPML regressions where the dependent variable is bilateral disbursements as a percentage of the GDP trend. The cycles are computed as the percent deviation between GDP and trend GDP, where the trend is computed with the Hodrick-Prescott filter. The debt variable changes from one column to another, as is indicated in each column title. All regressions include pair and time fixed effects. Clustered standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. In Table 5, we report the same exercises, but we split the samples between loans originating in advanced economies and those originating in China. The main baseline results regarding cyclicality hold. Results hold also for bilateral flows originating from advanced economies (columns [1]–[3]) and China (columns [4]–[6]). As opposed to the results in Table 4, here we find that the official bilateral flows are correlated with the initial stock of different types of debt that development banks and other mixed enterprises, political subdivisions, autonomous public bodies, and external obligations of private debtors that are guaranteed for repayment by a public entity. 28 Appendix A. List of Recipient Countries, by Income Level Low Income Benin Eritrea Madagascar Senegal Burkina Faso Ethiopia Malawi Sierra Leone Burundi Gambia, The Mali Tanzania Central African Republic Guinea Mozambique Togo Chad Guinea-Bissau Nepal Uganda Comoros Haiti Niger Zimbabwe Congo, Dem. Rep. Liberia Rwanda Middle Income Albania Dominica Lebanon Russian Federation Algeria Dominican Republic Lesotho Samoa Argentina Ecuador Maldives Sao Tome and Principe Armenia Egypt, Arab Rep. Mauritania Solomon Islands Azerbaijan El Salvador Mauritius South Africa Bangladesh Eswatini Mexico Sri Lanka Belize Fiji Moldova St. Lucia Bhutan Gabon Mongolia St. Vincent and the Grenadines Bolivia Ghana Montenegro Sudan Bosnia and Herzegovina Grenada Morocco Tajikistan Botswana Guatemala Myanmar Thailand Brazil Guyana Nicaragua Timor-Leste Bulgaria Honduras Nigeria Tonga Cabo Verde India North Macedonia Tunisia Cambodia Indonesia Pakistan Turkey Cameroon Jamaica Panama Uzbekistan China Jordan Papua New Guinea Vanuatu Colombia Kazakhstan Paraguay Venezuela, RB Congo, Rep. Kenya Peru Vietnam Costa Rica Kyrgyz Republic Philippines Yemen, Rep. Cote d’Ivoire Lao PDR Romania Zambia 29 Appendix B. List of Lending Countries, by Category Country Name Category Country Name Category Country Name Category Australia Advanced China China Libya Other Austria Advanced Algeria Other Malaysia Other Belgium Advanced Angola Other Mauritius Other Canada Advanced Argentina Other Mexico Other Czech Republic Advanced Barbados Other Morocco Other Denmark Advanced Belarus Other Nigeria Other Finland Advanced Bosnia-Herzegovina Other Oman Other France Advanced Brazil Other Pakistan Other Germany Advanced Brunei Other Panama Other Greece Advanced Bulgaria Other Peru Other Ireland Advanced Burundi Other Poland Other Israel Advanced Colombia Other Qatar Other Italy Advanced Congo, Rep. Other Romania Other Japan Advanced Costa Rica Other Russian Federation Other Korea, Republic of Advanced Cote d’Ivoire Other Saudi Arabia Other Netherlands Advanced Cuba Other South Africa Other New Zealand Advanced Egypt Other Tanzania Other Norway Advanced Gambia, The Other Thailand Other Portugal Advanced Guatemala Other Togo Other Singapore Advanced Hungary Other Trinidad and Tobago Other Slovak Republic Advanced India Other Tunisia Other Spain Advanced Indonesia Other Turkey Other Sweden Advanced Iran, Islamic Republic Other United Arab Emirates Other Switzerland Advanced Iraq Other Uzbekistan Other United Kingdom Advanced Kuwait Other Venezuela Other United States Advanced Kyrgyz Republic Other Vietnam Other 30 Appendix C. Additional Figures Figure C.1 Disaggregation of Total Public External Debt by Geographical Region (Average Country) Source: Authors’ compilation based on the IDS of the World Bank. Figure C.2 Disaggregation of Total Bilateral Public Debt by Origin and Geographical Region (Average Country) Source: Authors’ compilation elaboration based on the IDS of the World Bank. 31 Appendix D. Additional Results Table D.1 Results by Recipient Income Level (1) (2) (3) (4) (5) (6) Origin: All Origin: All Origin: Advanced Origin: Advanced Origin: China Origin: China Recipient: Low Income Recipient: Middle Income Recipient: Low Income Recipient: Low Income Recipient: Middle Income Recipient: Low Income Recipient’s cycle (t-1) 0.403 -0.198 1.193 -1.032 2.579** 0.751 (0.896) (0.611) (1.219) (0.662) (1.310) (1.383) Origin’s cycle (t-1) 2.023 2.422** 8.713*** 3.229* 10.240 13.386*** (2.142) (1.086) (3.346) (1.872) (7.245) (3.399) Log(Pair total trade [% trend GDP]) 0.470*** 0.509*** 0.489*** 0.323*** 0.457*** 0.467*** (0.073) (0.081) (0.133) (0.084) (0.125) (0.100) Log(Pair UN votes agreement score) -1.158** 0.597** 0.228 0.601** -1.865 0.075 (0.567) (0.242) (0.508) (0.262) (1.160) (0.847) Constant -1.750*** -1.298*** -1.394*** -1.278*** -1.449*** -1.052*** (0.193) (0.145) (0.280) (0.151) (0.235) (0.235) Number of observations 5,013 22,837 3,129 16,425 517 1,462 Number of pairs 324 1072 179 688 26 69 Pseudo-R-squared 0.33 0.43 0.38 0.38 0.19 0.37 Log likelihood -1366.74 -5157.78 -747.96 -3164.93 -324.70 -809.91 Note: This table presents a set of PPML regressions where the dependent variable is bilateral disbursements as a percentage of the recipient’s GDP trend. The cycles are computed as percent deviation between GDP and trend GDP, where the trend is computed with the Hodrick-Prescott filter. The recipient’s income group is according to the World Bank classification. Regressions with all and advanced economy origins include pair and time fixed effects, while those with the origin of China include only pair effects. Clustered standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01. 32 Table D.2 Results by Region (1) (2) (3) (4) (5) (6) EAP ECA LAC MENA SA SSA Recipient’s cycle (t-1) -10.108 3.143** 0.580 0.623 9.263* -3.634** (7.014) (1.505) (1.985) (2.967) (5.120) (1.726) Origin’s cycle (t-1) 4.274** 3.410 6.031** 3.232 -6.957* -1.011 (1.795) (4.343) (2.984) (2.149) (3.648) (1.419) Log(Pair total trade [% trend GDP]) 0.481*** 0.595*** 0.367** 0.470*** 0.509*** 0.507*** (0.117) (0.149) (0.186) (0.134) (0.168) (0.054) Log(Pair UN votes agreement score) 0.278 5.123*** 0.301 2.045*** -0.618 -0.279 (0.622) (1.310) (0.357) (0.499) (0.606) (0.328) Recipient’s cycle (t-1)* Adv dummy 10.421 -1.794 -2.122 -3.375 -8.475 4.363** (7.039) (2.055) (2.418) (2.948) (5.686) (1.948) Origin’s cycle (t-1)* Adv dummy 2.162 -5.648 -5.000 7.109 5.703 6.902** (3.675) (6.491) (3.453) (5.309) (5.742) (2.853) Recipient’s cycle (t-1)* China dummy 7.186 5.348 -0.606 -4.158 -1.353 5.594*** (7.728) (4.016) (4.867) (6.059) (6.221) (1.925) Origin’s cycle (t-1)* China dummy 10.529 6.670 -0.202 -2.217 14.757*** 3.418 (8.539) (12.202) (7.050) (7.382) (5.642) (4.282) Constant -1.278*** -0.474* -1.374*** -0.756*** -1.028** -1.598*** (0.295) (0.287) (0.293) (0.279) (0.444) (0.125) Sum recipient’s cycle adv. 0.313 1.348 -1.542 -2.752 0.788 0.730 (1.280) (1.384) (1.097) (1.700) (3.330) (0.910) Sum origin’s cycle adv. 6.436* -2.238 1.031 10.341** -1.254 5.891** (3.359) (5.434) (2.876) (4.886) (4.692) (2.476) Sum recipient’s cycle China -2.922 8.491** -0.026 -3.535 7.910** 1.960** (4.066) (3.677) (4.427) (5.431) (3.160) (0.898) Sum origin’s cycle China 14.803* 10.080 5.829 1.015 7.800* 2.407 (8.453) (10.120) (5.751) (7.077) (4.385) (4.137) Number of observations 3,799 2,956 6,552 3,169 2,723 8,651 Number of pairs 174 187 293 124 116 502 Pseudo-R-squared 0.51 0.51 0.37 0.33 0.64 0.35 Log likelihood -808.81 -525.63 -1404.00 -651.27 -526.54 -2418.49 Note: This table presents a set of PPML regressions where the dependent variable is bilateral disbursements as a percentage of the GDP trend. The cycles are computed as the percent deviation between GDP and trend GDP, where the trend is computed with the Hodrick-Prescott filter. The “Sum” rows refer to the joint coefficients of each cycle plus its corresponding interaction with the advanced or Chinese origin dummy. The recipient’s geographical group is according to the World Bank’s classification: East Asia and the Pacific (EAP), Europe and Central Asia (ECA), Latin America and the Caribbean (LAC), Middle East and North Africa (MENA), South Asia (SA) and Sub-Saharan Africa (SSA). All regressions include pair and time fixed effects. Clustered standard errors in parentheses. * p<0.1, ** p<0.05, *** p<0.01.