Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries

Working paper November 2022 Carlos Giraldo Iader Giraldo Jose E. Gomez-Gonzalez Jorge M. Uribe Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo Iader Giraldo Jose E. Gomez-Gonzalez Jorge M. Uribe November 2022 * ** *** **** Abstract Before the outbreak of the Global Financial Crisis, on May 6, 2007, the Colombian central bank imposed a cap on the Gross Leverage Position in Foreign Exchange Derivatives of financial intermediaries. It was the only country in the world in implementing this prudential policy. By leveraging insights from synthetic control literature we construct counterfactual scenarios and show that this policy intervention, while costly in financial stability terms in the pre-GFC period, was effective in reducing Colombia’s financial stability risks during the crisis. A trade-off between “calm” and “turbulent” periods emerges from our results, which should be taken into account when deciding on the right policy tools to use before a crisis breaks out. Keywords: PBA, synthetic-control, macroprudential policy. JEL: E58, E63, F38 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries * Latin American Reserve Fund | [email protected] ** Latin American Reserve Fund | [email protected] *** Lehman College, City University of New York; and, Visiting Professor (Summer School), Universidad de La Sabana [email protected].edu **** Open University of Catalonia | [email protected] Table of contents I. Introduction 4 II. Methodology 6 III. Data 8 IV. Results 8 V. Conclusions 14 References 16 4 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe I. Introduction1 We examine the effects of the implementation of a macroprudential policy in Colombia on May 6, 2007, consisting of a cap on the banks’ ratio of gross exchange rate derivatives to capital (PBA for its name in Spanish) of financial intermediaries. We analyze the evolution of Colombia’s financial stability index constructed by UNCTAD (Bouhia et al., 2022) after the implementation of the policy, and compare it with the evolution of a “synthetic Colombia” in which the PBA was not implemented. Comparing the results with those of a “placebo” test for Mexico, in the spirit of the synthetic control literature (Abadie et al., 2010), we conclude that the effects of the policy can be divided into two parts: i) between its implementation and before the outbreak of the Global Financial Crisis (GFC) in September 2008, in which the policy resulted in a worsening of the country’s financial stability conditions and, ii) after the crisis started, in which the policy acted as a shield that generated a notable increase in terms of stability compared to the counterfactual of no policy implementation. Inter-temporal aggregate gains from the policy are positive from the second quarter of 2007 to the last quarter of 2010. The GFC was a powerful reminder that economic fragility can develop beneath the surface of apparently stable macroeconomic conditions. External and financial imbalances can silently develop during periods of sustained growth dynamics, calling into question the adequacy of conventional tools to monitor economic risks and assess longer-term financial stability. While major global crises are infrequent, the severe recessions they generate entail significant costs in terms of foregone income and persistently high unemployment, justifying efforts for developing better tools and policies for assessing and guaranteeing economic resilience. One important source of macroeconomic vulnerability, especially in emerging market economies, relates to the volatility of large capital flows. While they can deliver substantial benefits for countries, they also have the potential to contribute to the buildup of large imbalances and systemic financial risk. The recent literature has shown that the benefits of capital flows are greater in countries in which institutional arrangements, and financial, macroeconomic, and fiscal policies enable them to intermediate capital flows safely (e.g., Igan et al., 2016; Schroth, 2016). Many countries have recently implemented reforms, including the development of macroprudential policies, that are helping to strengthen the resilience of financial systems to shocks, including those emerging from the volatility of international capital flows. Macroprudential policies, in support of sound macroeconomic policies and strong financial supervision and regulation, can play an important role in helping countries harness 1 The opinions in this paper are those of the authors and do not commit FLAR or its directory board. 5 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe the benefits of capital flows, while reducing the negative effects related to their volatility. This is particularly meaningful in commodity-dependent economies in which these flows are strongly related to international commodity price cycles, such as Latin America and Caribbean countries. Central banks have played a special role in building macroprudential policy tools for enhancing country resilience, especially through the implementation of prudential international capital control and the accumulation of international reserves. However, while necessary, policies for “having the house in order” are clearly not enough for strengthening economic resilience and reducing the vulnerability of small open economies to large external shocks and high capital flow volatility. The increasing level of financial globalization, which comes in hand with a larger volatility of economic and financial cycles (Rodrik, 1997; Scheve and Slaughter, 2001), demands the development of better insurance mechanisms against international shocks. In the midst of a period of massive portfolio capital inflows, the central bank of Colombia implemented in May 2007 a heatedly debated prudential policy tool with the objective of reducing the country’s vulnerability to a capital flows sudden stop. The policy tool, called the Gross Leverage Position in Foreign Exchange Derivatives, imposed a cap on a bank’s ratio of gross exchange rate derivatives to capital. The Banking Association of the country protested the PBA, as they considered it an extreme measure that would not effectively control credit and counterparty risk in the banking sector and could instead limit the capacity banks had to offer firms and households reliable financial instruments to hedge against exchange rate risk. The central bank, arguing that the PBA would limit the credit risk exposure of Colombian banks safeguarding the stability of the financial system, imposed a 500% limit to the PBA effective on May 6, 2007. In brief, the intention of this policy was to moderate the credit growth rate in Colombia, which according to central bank officials was too high and could lead to the development of undesirable financial sector imbalances and house price bubbles. While economic authorities in Colombia had already imposed a tax to foreign debt and to portfolio investment in the country one year before, these measures were not unique to the country as Brazil and Chile had undertaken similar policies at roughly the same time. The PBA, however, was implemented by Colombia alone. In fact, such a policy had never been implemented before in any other country. Our results show, however, that similar prudential policies can be successfully implemented in other small open economies when sudden capital inflow upsurges may threaten the stability of the financial sector. While in various issues of the Financial Stability Report and in their report to Congress the central bank of Colombia assured this prudential policy tool fulfilled the purpose for which it was created, no formal policy evaluation was conducted to show the validity of their conclusion. To the best of our knowledge, there is only one paper formally studying the effectiveness of the macroprudential policy measures undertaken by the central 6 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe bank of Colombia between 2006 and 2007 (Gomez et al., 2020). However, they focus on two macroprudential policies, marginal reserve requirements and dynamic provisions. Their results suggest that dynamic provisions, the countercyclical reserve requirement, and an aggregate measure of the macroprudential policy stance had a negative effect on credit growth, which varies according to bank and debtor-specific characteristics. However, their results cannot be interpreted as causal because they do not use an adequate counterfactual analysis. We evaluate the causal effect of the PBA implementation on the stability of the financial system of a representative medium-sized emerging market economy, Colombia. We use synthetic control methods following the seminal papers of Abadie and Gardeazabal (2003) and Abadie et al. (2010). These methods, which have been recently used in the literature to evaluate the effects of several economic policies, such as right-to-carry laws (Donohue et al., 2019), immigration policy (Bohn et al., 2014), taxation (Kleven et al., 2013), and pay-equity laws (McDonald and Thornton, 2015), are useful when evaluating the causal effects of aggregate interventions. They are tools for disentangling causal effect when interventions are implemented at an aggregate level affecting a small number of large units (like countries), on some aggregate outcome of interest. II. Methodology Let Yit be the outcome that would be observed in the financial stability index for the country i at time t in the absence of PBA, for countries i = 1, …, J + 1, and time periods t = 1,… T. Let T0 denote the number of periods before the PBA intervention, which satisfies 1 ≤ T0 < T. Let Yit be the outcome that would be observed from country i at time t if country i is exposed to the intervention from T0 + 1 to T. We assume that the PBA did not have effects before March 31, 2007, which marks the last day of the quarter before the actual implementation of the policy on May 6, 2007. That is, before the first quarter we assumed there were not any “anticipation effects” on the policy. Let ait = Yit - Yit be the effect of the PBA in country i at time t, and Dit denote and indicator that takes value one if country i is exposed to the PBA at time t, and zero otherwise. Thus, the observed outcome for country i is given by: (1) If only the first country is exposed to the intervention and only after period T0, it follows that: N N I I N Yit = Yit + ait Dit Dit = 1 if i=1 and t > T0 0 otherwise (2) 7 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe Our goal is to estimate (a1T + 1 ,…, a1T ). For t > T0 , (3) Due to the fact that Y1t is observed, to assess a1t we just need to estimate Y1t. Following Abadie et al. (2010) we can assume that Yit dynamics is driven by a factor model in the following way: (4) where dt is an unobservable common factor, with constant factor loadings across countries, Zi is a (r x 1) vector of observed covariates (unaffected by the PBA implementation), qt is a (r x 1) vector of unknown parameters, lt is a (F x 1) vector of unobserved factor loads, and eit are random noise at the country level. Abadie et al. (2010) show that, given a (J x 1) vector of weights W = (w2 ,…, wJ+1 )’ such that wj ≥ 0 for j=1,…,J + 1 and all weights sum up to unity, we have that each particular vector W represents a potential synthetic control, or in other words, a particular weighted average of control countries in the “donor pool”. In this case, the outcome of the financial stability indicator for each synthetic control indexed by W is given by: (5) Suppose that there exist (w2 ,…, wJ+1) which are optimal, in the sense that they are set as to minimize the distance between the outcome pre-treatment series and the synthetic control series before the policy intervention, and minimize ∑J+1 wj Zj - Z1. In cases where the number of pre-intervention periods is large compared to the scale of the country errors we can use a1t as an estimator of a1t. Such that: (6) for t ∈ {T0 +1,…,T}. These equations naturally do not hold exactly in practice, as none convex combination of the control countries can perfectly replicate the pre-intervention dynamics of the treated country. Thus, optimal weights of both control units and predictors are simultaneously set such that these equations hold approximately. We use the “synth” package in R, to estimate both sets of weights. 0 a1t = Y1t -Y1t = Y1t - Y1t NI N N N I Yit = dt + qt Zi + lt mi + eit , N ∑J+1 wj Yjt = dt + qt ∑J+1 wj Zj + lt ∑J+1 wj mj + ∑J+1 wj ejt j=2 j=2 j=2 j=2 * * j=2 * ˆ a1t = Y1t - ∑J+1 wj Yjt , * ˆj=2 8 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe III. Data We acknowledge the caveats of synthetic control methods. Namely that even if there exists a synthetic control that offers a good fit for Colombia (see Table 1), interpolation biases may be significant if the factor model represented by Equation 4 does not hold relatively well over the entire set of countries in our sample. In this case, researchers aiming to minimize the biases caused by interpolating across many heterogeneous countries with largely dissimilar characteristics should restrict the donor pool to countries with similar characteristics to the country exposed to the policy intervention. In our case, this means that we opted for keeping a reduced number of control units following two economic reasons: first, the macroeconomics literature emphasizes on the fundamental role of regional characteristics to explain international capital flows to countries (e.g. Kaminsky, 2005). Therefore, our control units must belong to Latin America. Second, there is a great heterogeneity of countries within Latin American countries, including dollarized economies (i.e., Ecuador, El Salvador), and economies that are perceived as likely defaulters from their international debt obligations (e.g., Argentina, Venezuela). Our selection of controls is restricted to those countries that possibly resemble Colombia most, economically and in terms of regulation. In Table 1 we present the summary statistics of our variables for Colombia and the control countries: Brazil, Chile, Mexico, Peru. In our set of predictor variables for the model in equation 4 we include: real output growth (real.growth), the investment to GDP ratio (investment.to.gdp), the unemployment rate (unemployment), inflation rate (inflation), the growth of real estate prices (real.estate.prices), the monetary base (monetary.base), the central bank policy rate (rate), foreign reserves as percentage of GDP (foreign.reserves), and the following balances: fiscal (fiscal. balance), trade (trade.balance), current account (current, balance), financial account (capital.balance), total direct investment (direct.investment.balance), total portfolio (portfolio.balance), the other investments (other. balance), all of them as percentage of GDP. We also include the index of multilateral real exchange rate (real. exchange.rate). The sample mean of these statistics is shown in the fourth column of Table 1. IV. Results 1. Policy evaluation in Colombia Table 1 presents summary statistics on relevant variables for Colombia (treated) and for the synthetic control. Notably, mean data for the synthetic are closer to observed mean data for Colombia than (unweighted) sample means for the countries included for building the control. This is especially true for variables which are usually considered as highly related to financial stability and still unaffected by the policy, like real estate prices. 9 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe Table 2 shows the optimal weights used for constructing the synthetic. Both predictor weights (Panel A) and country weights (Panel B) are reported. Variables corresponding to the external sector, such as the current account balance and the capital balance, variables related to capital flows, such as the real exchange rate and foreign direct investment, are among the variables with highest weights in building the synthetic. Important to note, synthetic Colombia is basically a convex combination of Mexico and Chile. Figure 1 shows the effect of the PBA implementation on the Financial Stability Index. Both the factual and counterfactual values for this index are depicted. The continuous line represents the behavior over time of the index for Colombia (treated). The dotted line represents the behavior of the index for synthetic Colombia (control). Two moments in time are highlighted, namely the date in which the PBA was implemented (quarter corresponding to May 6, 2007) and the date of the Lehman Brothers default in September 2008. Recall that higher (lower) values of the index correspond to situations of higher (lower) financial stability. According to our results, the PBA had an impact on financial stability in Colombia. However, the impact was not homogeneous over time. Interestingly, when the policy measure was undertaken, before the beginning of the Global Financial Crisis, the PBA was detrimental to the overall financial stability of the Colombian financial sector. This apparently counterintuitive result may be explained by the fact that this macroprudential policy constrains the ability Table 1. Summary statistics for relevant variables: Colombia, synthetic Colombia, and sample averages Variable Treated Synthetic Sample.Mean real.growth 6.2 4.0 4.8 investment.to.gdp 22.1 22.2 20.2 unemployment 11.7 4.8 8.1 inflation 4.7 3.7 3.4 real.estate.prices 11.5 3.3 2.0 monetary.base 6.2 3.7 4.0 rate 6.7 7.3 8.5 foreign.reserves 10.1 9.2 11.6 fiscal.balance -4.0 1.2 0.9 trade.balance 0.6 3.0 5.2 current.balance -1.4 0.3 1.3 capital.balance 1.3 -0.2 -1.1 direct.investmente.balance 3.4 2.8 2.8 portfolio.balance -0.9 -1.7 -0.9 other.flows -0.1 -0.4 -1.6 real.exchange.rate 110,334 93,314 92,121 Note: the table shows the sample means of the treated Colombia (column 2), synthetic Colombia (column 3), and the unweighted sample mean of control units (colum 4). 16 Banks’ Leverage in Foreign Exchange Derivatives in Times of Crises: A Tale of Two Countries Carlos Giraldo | Iader Giraldo | Jose E. Gomez-Gonzalez | Jorge M. Uribe References Abadie, A., and Gardeazabal J. (2003). “The Economic Costs of Conflict: A Case Study of the Basque Country.” American Economic Review 93: 113-32. Abadie, A., Diamond, A., and Hainmueller, J. (2010). “Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program.” Journal of the American Statistical Association 105: 493-505. Abadie, A., Diamond, A., and Hainmueller, J. (2015). “Comparative Politics and the Synthetic Control Method.” American Journal of Political Science 59: 495-510. Bohn, S., Lofstrom, M., and Raphael, S. (2014). “Did the 2007 Legal Arizona Workers Act Reduce the State’s Unauthorized Immigrant Population?” Review of Economics and Statistics 96: 258-69. Bouhia, R., Blanc, G., Falciola, J., Kaczmarczyk, P., and Bodelet, J. (2022) “Between stress and strain: understanding, measuring, and analysing financial conditions in developing countries in times of Covid-19 and beyond”. DA-COVID 19 Project paper 06/22. United Nations Development Account Projects. Igan D., Kutan A.M., and Mirzaei A. (2016). “Real Effects of Capital Inflows in Emerging Markets,” IMF Working Paper WP/16/235, International Monetary Fund. Kaminsky, G. (2005) International Capital Flows, Financial Stability and Growth. DESA Working Paper No. 10 Kleven, H.J., Landais, C., and Saez, E. (2013). “Taxation and International Migration of Superstars: Evidence from the European Football Market.” American Economic Review 103: 1892-1924. McDonald, J.A., and Thornton, R. J. (2015). “Coercive Cooperation”? Ontario's Pay Equity Act of 1988 and The Gender Pay Gap,” Contemporary Economic Policy 33: 606-618. Rodrik D. (1997). “Has Globalization Gone Too Far?” Institute for International Economics. Scheve K.F., and Slaughter, M.J. (2001). “Globalization and the Perceptions of American Workers,” Institute for International Economics. Schroth, J. (2016). “Capital Flows to Developing Countries: Is There an Allocation Puzzle?” Staff Working Paper 2016-53, Bank of Canada. Fondo Latinoamericano de Reservas | FLAR Calle 84A No. 12-18 Piso 7 | Bogotá, Colombia Correo electrónico: [email protected] Tel: (571) 634 4360