WPS6793 Policy Research Working Paper 6793 Reserve Requirements in the Brave New Macroprudential World Tito Cordella Pablo Federico Carlos Vegh Guillermo Vuletin The World Bank Development Economics Vice Presidency Operations and Strategy Unit February 2014 Policy Research Working Paper 6793 Abstract Using a new, large data set on quarterly reserve business cycle in developing countries. This enormously requirements for the period 1970–2011, this paper complicates the use of interest rates as a countercyclical provides new evidence on the use of reserve requirements instrument (because of its effect on the exchange rate) as a countercyclical macroprudential tool in developing and calls for a second instrument. The paper suggests countries. The appeal of reserve requirements lies in that conflicts may arise between the microprudential and the pro-cyclical behavior of the exchange rate over the macroprudential policy stances. This paper is a product of the Operations and Strategy Unit, Development Economics Vice Presidency. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at tcordella@worldbank.org, cvegh1@jhu.edu, pablo.federico@blackrock.com and GVuletin@brookings.edu. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Reserve Requirements in the Brave New Macroprudential World* Tito Cordella, † Pablo Federico, ‡ Carlos Vegh § and Guillermo Vuletin ** JEL Classification Numbers: E44, E58, F31, F32, F38, G28 Keywords: Reserve Requirements, Macroprudential policies, Monetary Policy, Capital Flows, Exchange Rate, Business Cycle * This working paper summarizes the main policy conclusions of a regional study sponsored by the Chief Economist Office of the Latin American Region of the World Bank. The study was conducted by a team composed by Tito Cordella, Augusto de la Torre, Pablo Federico, Alain Ize, Samuel Pienknagura, Carlos Vegh, and Guillermo Vuletin. Excellent advice has been received from our peer reviewers Eva Gutierrez, Luis Jacome, and Andrew Powell. We would also like to thank participants at conferences/seminars at the Bank of Spain, Central Bank of Turkey, CEMLA, Central Bank of Uruguay, IDB, and IMF for very insightful comments and suggestions, and Gabriela Calderon Motta and James Trevino for excellent editorial assistance. Financial support from the Government of Spain through the SFLAC Trust Fund is gratefully acknowledged. The usual disclaimers apply. † Development Economics (DECOS), The World Bank ‡ BlackRock § Johns Hopkins University and NBER ** The Brookings Institution 1 Introduction In the aftermath of the global financial crisis of 2008-2009, macroprudential policy has become the new policy buzzword. Indeed, it seems hard to find any detailed statement about macroeconomic policy—particularly relating to emerging countries—that does not include, at some point or another, some reference to financial stability or systemic risk and the resulting need for “macroprudential policy.” Moreover, in the last few years the debate on the consequences of quantitative easing (and its tapering) by the industrial world (particularly the United States) and the ensuing “currency wars,” put macroprudential policy, in the form of changes in reserve requirements, at the forefront of policy discussions. For example, in a February 12, 2013 article, the Financial Times focuses on the macroeconomic problems that appreciating currencies have brought about in several Latin American countries and notices that while Chile, Colombia, and Peru have eschewed Brazilian-style capital controls, “they have turned to direct foreign currency intervention; paying down foreign debt; macroprudential measures such as increases in bank reserve requirements and interest rate cuts that reduce the ‘carry’, or interest rate differential, for yield hungry investors.” 6 In other words, changes in reserve requirements are very much part of the main policy menu at the disposal of emerging market policy makers concerned about how to respond to capital inflows or outflows. Indeed, when the appeal of emerging markets started to wane last summer and their currencies consequently began to plummet, opposite kinds of measures were undertaken. Brazil cut its tax on overseas investment in domestic bonds from 6 percent to zero, the Indian government relaxed overseas borrowing rules, and forex interventions to defend currencies became widespread. It is not hard to predict that other macroprudential measures such as reserve requirements will soon be adjusted to this new environment. However, as a French saying rightly reminds us, plus ça change, plus c’est la même chose (the more things change, the more they remain the same). In a widely-cited 1993 article in the IMF Staff Papers on capital inflows to Latin American in the early 1990s, Calvo, Leiderman, and Reinhart discuss in detail the policies deployed by policy makers in the region and conclude that “there are grounds to support a mix of policy intervention based on the imposition of a tax on short-term capital inflows, on enhancing the flexibility of exchange rates, and on raising marginal reserve requirements on short-term bank deposits.” And, even earlier, in his celebrated 1985 paper “Good-bye financial repression, hello financial crash”, Diaz-Alejandro had already entertained the possibility that “prudential regulatory machinery could be used to discourage volatile international financial flows [by] relying primarily on taxes or tax-like requirements, i.e., via special reserve requirements for certain types of unwanted international financial transactions.” In terms of the broader policy discussion on macroprudential policy, it is important to recognize that there seems to be two different reasons why macroprudential policies are so popular nowadays. The first, fashionable in academic circles and IFIs, focuses on the buildup of 6 “Currency fears spread in Latin America” by John Paul Rothbone, Financial Times, February 12, 2013. 2 systemic risk and is driven by financial stability concerns. The second—already mentioned above and more popular among policy makers in emerging markets—focuses on macroeconomic stabilization in the presence of large capital flows and is driven mostly by the desire to stabilize the exchange rate and the credit cycle. In practice, the distinction between the systemic risk- driven and the business-cycle driven use of macroprudential policy is not a stark one. Macroprudential policies aimed at dealing with systemic risk can contribute to macroeconomic stabilization by reducing the amplitude of the credit cycle (the dynamic provision measures implemented in Spain are a good example) and the use of macroprudential instruments to deal with the business cycle can certainly contribute to financial stability by preventing excessive (or “sub-optimal”) fluctuations in capital flows, which may reduce the probability of systemic-risk. While this paper will not be able to distinguish between the two different drivers of macroprudential policies, it will focus mainly on the use of business cycle-driven macroprudential policy and will provide arguments supporting the predominance of macroeconomic stabilization considerations. 7 Further, even though we can certainly not rule out a priori that systemic-risk macroprudential policy may also be related to the business cycle to the extent that the “risk cycle” (assuming such a concept can be well-defined and measured) may be correlated with the business cycle, we will take the position that the use of macroprudential policy linked to the business cycle is primarily driven by macroeconomic stabilization considerations. In addition, we will focus exclusively on reserve requirements (RR) for two reasons. First, RR are the most widely used prudential instrument to deal with business cycle concerns and have a long history in the region and elsewhere. 8 9 In fact, ever since the financial liberalization of the late 1970s in the Southern-Cone, each new wave of capital inflows into the region (either due to financial liberalization or following a period of outflows) has triggered a remarkably similar discussion regarding the pros and cons of higher reserve requirements. 10 7 Needless to say, by doing so, we do not mean to diminish in any way the importance of systemic risk- driven macroprudential policy. Our goal, however, was to approach the subject of macroprudential policy from a strictly macroeconomic stabilization point of view, an angle that has received relatively little attention in the academic literature. 8 See Federico, Vegh, and Vuletin (2013c) for an analysis of 40 years of macroprudential policy history in Brazil and Mexico. They document that while a wide variety of macroprudential instruments were used over time in both countries, RR were the most common. 9 One may question whether the use of RR should be viewed as macroprudential policy rather than monetary policy. In our view of the world, this would depend on policy makers’ intentions. If RR are used exclusively to regulate market liquidity, they should indeed be viewed as monetary policy. If, instead, they are used to influence the build-up of systemic risk over the business cycle, they should be viewed as macroprudential policy. While interesting, this distinction is not relevant for our empirically- based analysis since we do not observe policy makers’ intentions. We thus take the implicit position – consistent with common practice – that changes in RR indeed constitute as a “macroprudential policy.” 10 Of course, the use of reserve requirements as a macroeconomic stabilization tool should not be confused with their use as a tool of financial repression in the preceding decades, as discussed in detail in the classical analyses of McKinnon (1973) and Shaw (1973). In this case, high reserve requirements, together with interest rate ceilings, quantitative restrictions on credit allocation, multiple exchange rates, and other non-market mechanisms were part of a financial repression model, which by depressing savings and investment, led to dismal growth and overall economic performance. Of course, the financial liberalization of the 1960s and 1970s in Latin American brought its own set of financial problems, as masterfully discussed by Diaz-Alejandro (1985), which arguably endure until today. 3 Second—and to be as substantive and novel as possible—we wanted our analysis to be based on long and comparable time series for a large number of countries, both industrial and developing, and RR are the only instrument for which it was possible (although not without a lot of hard work) to put together such an extensive and novel data set. To this effect, we have put together a data set on quarterly legal reserve requirements for 52 countries (15 industrial and 37 developing countries) for the period 1970-2011. 11 12 This novel data set, together with more readily available data on credit, international reserves, GDP, and other macroeconomic series will allow us to identify a series of stylized facts regarding the use of RR as a macroeconomic stabilization tool. Based on these stylized facts, we can then proceed to provide a policy rationalization for what we have observed and for the relation between reserve requirement policies (RRP), monetary policy and foreign exchange market intervention. This working paper proceeds as follows. Based on Federico, Vegh, and Vuletin (2013a), section 2 documents the key stylized facts regarding the use of reserve requirements as a macroeconomic stabilization tool and how RRP relates to the level of credit, monetary policy, and foreign exchange market intervention. We first propose an operational definition to distinguish countries that make use of RR as a stabilization tool (which we dubbed “active countries”) from those that do not (“passive countries”). We show that developing countries are much more likely to pursue an active RRP than industrial countries. This is particularly true in the post-2004 period, which is consistent with the casual observation of the increasing reliance on macroprudential policy in emerging markets. We then show that, within active countries, the level of credit to GDP is an important determinant of how countercyclical is the use of RR. We then analyze the relation of RRP with monetary policy. Our starting observation is the fact that, unlike industrial countries, many emerging markets have not used interest rates countercyclically (i.e., to smooth out the business cycle). We will interpret this as reflecting the need to defend the currency in bad times (for fear of a sudden and perilous depreciation) and the fear of attracting yet more capital inflows in good times (by making domestic-currency assets more attractive). In order to understand the simultaneous choice of RRP and monetary policy, we then build what we call “policy mix matrices” to illustrate all the possible combinations of RRP and monetary policy in terms of their cyclical properties (pro-cyclical, a-cyclical, and countercyclical). 11 For convenience, we will use the term “countries” but notice that our list includes the Euro zone as a single economic unit and Ecuador as two distinct economic units (before and after full dollarization in the year 2000). The reader is referred to Federico, Vegh, and Vuletin (2013a) for the list of countries, sample periods, and data sources. 12 Notice that we are interested in the legal reserve requirement rate as opposed to the effective or average reserve requirements because the latter would be highly sensitive to the business cycle (as deposits correlate positively with the business cycle) and would thus not be a good indicator of the cyclical stance of reserve requirement policy. 4 We show how industrial countries typically make no use of RR and conduct countercyclical monetary policy. Emerging markets, in contrast, rely heavily on RRP and often use it as a substitute for monetary policy, in the sense of using RRP to smooth the business cycle and using interest rate policy either a-cyclically or pro-cyclically. We then examine the use of foreign exchange market intervention as an additional policy tool and how it interacts with RRP. We find that, by and large, foreign exchange market intervention is used to smooth out changes in the nominal exchange rate by buying foreign assets in good times (thus preventing “excessive” currency appreciation) and selling foreign assets in bad times (thus preventing “excessive” currency depreciation). In section 3, we use four LAC countries to simulate the policy response to output and exchange rate shocks. The picture that emerges is one in which in bad times developing countries choose a policy mix that consists on increasing interest rates (to defend the currency), selling international reserves (also to defend the currency), and lowering reserve requirements (to stimulate credit and, hence, economic activity). Conversely, in good times countries tend to accumulate international reserves (to fight off appreciation of the domestic currency), increase RR to cool off the credit cycle, but are reluctant to increase interest rate because of the fear of exacerbating capital inflows. Section 4 attempts to provide a policy rationale for the stylized facts we identified in the previous sections. In our view of the world, the distinctive feature of emerging markets’ business cycle is the fact that domestic currencies tend to appreciate in good times and depreciate (and often precipitously) in bad times. A sharp currency depreciation in bad times leaves policy makers with little choice but to increase interest rates (or at least not lower them) to make domestic-currency assets more attractive and prevent the currency from plummeting. This pro- cyclical monetary policy triggers the need for a second instrument that can be used to stimulate output. In this context, lowering RR should help in fostering credit expansion and thus steer the economy out of recession. A critical issue—often ignored in policy as well as academic circles—is the possible (often unintended) consequences of different macroprudential instruments on individual banks’ risk taking incentives. In section 5, we argue that if from a macroprudential point of view, different instruments—such as RR, capital requirements, and taxes on credit—may be equivalent, from a microprudential point of view, generally, they are not. In addition, depending on what the drivers of the business cycle are, the macro and microprudential stances may or may not go hand in hand. Section 6 elaborates on the possible tensions between different policy objectives and maintains that policy makers may disregard the microeconomic consequences of different macroprudential instruments only at their own peril. The main policy implications of the study are summarized in section 7. 5 2 Stylized Facts This section identifies the main stylized facts related to the use of reserve requirements as a macroeconomic stabilization tool. We first develop an operational definition that will allow us to establish which countries have indeed used reserve requirements as a macroeconomic stabilization tool. We then look at the relation between RRP and monetary policy. Finally, we look at the relation between RRP and foreign exchange market intervention. 2.1 Which countries have used reserve requirements as a macroeconomic stabilization tool? This is the first question that one would like to answer since, unlike monetary or fiscal policy, not all countries may have necessarily used reserve requirements to smooth out the business cycle. Answering this question, however, requires some operational definition of what do we mean by using reserve requirements as a macroeconomic stabilization tool. The main idea behind our operational definition will be that if the average duration between the changes in RR for any given country is shorter than the average duration of the business cycle in the same country, we will classify that country as having an active RRP. If not, we will classify that country as having a passive RRP. 13 For example, suppose that, on average, a country changes reserve requirements every ten years but the average duration of its business cycle is three years. Clearly, such a country is not using reserve requirements to influence the business cycle. Conversely, suppose that reserve requirements are changed every two quarters with the same average duration of the business cycle. In this case, it is very likely that the main purpose of changing reserve requirements is to smooth out the business cycle. 14 Figure 1 shows the quarterly frequency of changes in legal RR for each of the 52 countries in our sample for the period 1970–2011 (notice that duration is simply the inverse of frequency.) Yellow (light) bars indicate developing countries while black bars denote industrial countries. For example, if a country has a frequency of 0.5, this means that it changes RR once every two quarters. In other words, the average duration between changes in RR is two quarters (1/0.5). As the figure makes clear, developing countries exhibit the highest frequencies. In fact, the industrial country with the highest frequency, Sweden, has a frequency of 0.10 indicating that it changes RR once every 10 quarters (or 2.5 years). From Figure 1, we can already guess that, under our operational definition, many more developing than industrial countries will be classified as having an active RRP. 13 Formally—and to account for the high variability of the business cycle duration within countries—we will classify a country as pursuing an active RRP if the average duration between changes in RR is lower than the average duration of the business cycle plus one standard deviation. 14 It is, of course, possible that reserve requirements are being changed for reasons other than smoothing out the business cycle (i.e., for microprudential reasons). Casual observation, however, suggests that the frequency of changes for microprudential purposes is much lower. 6 Figure 1: Frequency of changes in reserve requirements (1970-2011) 0.75 Quarterly frequency of changes in reserve requirements India 0.5 Ecuador (pre-dollarization) Honduras Argentina Serbia Jamaica Trinidad and Tobago Costa Rica Nicaragua Ecuador (dollarization) Belarus 0.25 Malaysia China Colombia Croatia Latvia Romania Venezuela Turkey Sweden Hungary Philippines Poland United Kingdom Peru United States Macedonia Brazil Domincan Rep. Czech Rep. Germany Portugal Lithuania New Zealand Uruguay Japan Switzerland Israel France El Salvador Singapore Guatemala Panama Thailand Australia Denmark Mexico Spain Euro-17 Canada Norway Chile 0 Figure 2 shows the classification of all countries in the sample into active or passive when it comes to RRP. To this effect, we construct a scatter plot with the vertical axis indicating the average time (in years) between changes in RR and the horizontal axis indicating the average duration of the business cycle (also in years). We then draw a 45-degree line. Generally speaking, countries below the 45-degree line are “active” countries because the average time between changes is lower than the average duration of the business cycle. Conversely, countries above the 45-degree line are passive countries. 15 16 Figure 2: Active versus passive reserve requirement policy (1970-2011) Infinite Passive Time between changes in RR (in years) 25 Active 20 15 10 5 0 0 2 4 6 8 10 Duration of business cycle (in years) 15 Countries with zero frequency of change would have an “infinite” amount of time between changes and are plotted out of scale. 16 We say “generally” because, to keep it simple, the scatter plot does not allow for the one standard deviation of the business cycle duration embedded in our formal definition above. In the figure, a black dot (cross) denotes an active (passive) country. 7 Note: The dashed line is a 45 degree line. Countries located below (above) the 45 degree line are countries for which the change in reserve requirements takes place, on average, at least (less than) one time per business cycle. Countries are classified as active if the average duration of business cycle plus one standard deviation of business cycle is larger than the average time between changes in reserve requirements. Based on Figure 2, 62 percent of countries (32 of 52) are classified as active countries. As expected, there is a striking difference between developing and industrial countries: 68 percent (or 27 of 37) of developing countries are classified as active, whereas only 33 percent of industrial countries (five of 15) are classified as active. In the case of LAC countries, 65 percent (or 11 of 17) are classified as active. In fact, the difference between industrial and developing countries becomes even more striking if we divide the sample before and after 2004. For the post-2004 sample—and as Figure 3 illustrates—while the frequency of changes in RR for developing countries looks roughly similar to the whole sample (Figure 1), there is not a single industrial country that has changed legal reserve requirements in this period.17 Using our formal criterion, 57 percent of developing countries (or 21 of 37) are classified as active in the 2005–2011 period. Figure 3: Frequency of changes in reserve requirements (2005-2011) 0.75 Quarterly frequency of change of reserve requirement India China 0.5 Honduras Trinidad and Tobago Serbia Peru Venezuela 0.25 Belarus Croatia Colombia Latvia Turkey Malaysia Romania Nicaragua Philippines Argentina United Kingdom Jamaica Macedonia Costa Rica Poland United States Brazil New Zealand Ecuador Lithuania Switzerland Panama Czech Rep. El Salvador Uruguay Guatemala Hungary Singapore Australia Denmark Thailand Sweden Euro-17 Canada Norway Mexico Japan Israel Chile 0 17 Of the 15 industrial countries in this sub-sample, 8 have in fact no legal reserve requirement. 8 2.2 What have been the cyclical properties of RR as a macroeconomic stabilization tool? Now that we have identified a group of countries that have used RRP actively during the 1970–2011 period, we look into the cyclical properties of RR in these countries. In other words, we would like to know if RR have had a positive or negative correlation with the business cycle. To this effect, we compute the correlation between the cyclical components of RR and real GDP for each of the 32 countries classified as active. These correlations are illustrated in Figure 4. We can see that 72 percent (or 23 of 32) of the correlations are positive, indicating countercyclical RRP. 18 If we limit ourselves to positive correlations that are significantly different from zero, the figure accounts for 38 percent (or 12 of 32). Figure 4: Cyclicality of reserve requirement policy (1970-2011) Malaysia* 1 Croatia* Germany* Latvia* Correlation (RGDP, reserve requirement) Ecuador (dollarization) Venezuela* Lithuania* Hungary* China* Peru* Honduras 0.5 Brazil* France Belarus* Romania Sweden* Philippines Serbia Argentina Nicaragua Israel Japan India 0 Turkey Portugal Poland Uruguay Costa Rica Colombia Ecuador (pre-dollarization) Trinidad and Tobago Jamaica* -0.5 Av. industrial = 0.30* Av. developing = 0.18* 67% of industrial countries excluded 27% of developing countries excluded -1 Note: Average reserve requirement is used for calculations. Sample only includes active reserve requirement policy countries (5 of 15 industrial countries and 27 of 37 developing economies are active). * indicates that the correlation is statistically positive at five percent level. As we already mentioned, however, the figures for the whole sample period mask an important relative change. Dividing the sample again before and after 2004 (Figure 5, Panels a, and b, respectively), we can see that before 2005, the number of active countries with countercyclical RRP was 60 percent (or 18 out of 30) but after 2004 the number rises sharply to 90 percent (or 19 of 21). Particularly striking is the fact that in the post-2004 period, there is not a single industrial country that has pursued active RRP. 18 By definition, countercyclical (procyclical) RRP refers to a situation in which the correlation between RR and the business cycle is positive (negative). Naturally, the idea of countercyclical RRP would be to cool down the economy in good times by raising RR and to stimulate output in bad times by lowering RR. 9 Figure 5a: Cyclicality of reserve requirement policy (1970-2004) Malaysia* 1 Germany* Av. industrial = 0.25* Venezuela* Av. developing = 0.10* Correlation (RGDP, reserve requirement) Hungary* 67% of industrial countries excluded Peru* 32% of developing countries excluded France* Sweden* Croatia 0.5 Brazil* China* Latvia* Lithuania Czech Rep. Serbia Israel Argentina Belarus Japan 0 Trinidad and Tobago Portugal Uruguay Jamaica Poland Costa Rica Turkey Ecuador (pre-dollarization) Romania Colombia* Nicaragua India -0.5 -1 Note: Average reserve requirement is used for calculations. Sample only includes active reserve requirement policy countries (5 of 15 industrial countries and 25 of 37 developing economies are active). * indicates that the correlation is statistically positive at five percent level. Figure 5b: Cyclicality of reserve requirement policy (2005-2011) Argentina* Latvia* Lithuania* 1 China* Romania* Belarus* Croatia* Colombia* India* Malaysia* Venezuela* Turkey* Brazil* Correlation (RGDP, reserve requirement) Peru* Poland* Honduras 0.5 Philippines Nicaragua Serbia 0 Av. developing = 0.47* -0.5 100% of industrial countries excluded Trinidad and Tobago* 43% of developing countries excluded Jamaica* -1 Note: Average reserve requirement is used for calculations. Sample only includes active reserve requirement policy countries (21 of 37 developing economies are active). * indicates that the correlation is statistically positive at five percent level. 10 The evidence thus suggests that (i) reserve requirements have been used fairly frequently as a macroeconomic stabilization tool (i.e., countercyclically); (ii) their use has increased considerable since 2004, which is consistent with anecdotal evidence to this effect, and (iii) no industrial country has used RRP actively in the post-2004 period. 2.3 How is RRP related to the credit cycle? After all, we would think that the channel through which reserve requirements may smooth out the real GDP cycle is by dampening the amplitude of the real credit cycle. In this respect, the first observation is that, as expected and illustrated in Figure 6, real credit is highly pro-cyclical in both industrial and developing countries. Hence, the pro-cyclicality of credit per se cannot offer an explanation as to why some countries actively use RRP while others do not. In line with this, our policy rationale offered below relies instead on the behavior of the nominal exchange rate over the business cycle. Figure 6: Correlation of private credit with GDP (1970-2011) United States* Honduras* Switzerland* Venezuela* 1 Trinidad and Tobago* Macedonia* Germany* Argentina* Czech Rep.* Nicaragua* France* Panama* Croatia* Turkey* Spain* Mexico* Australia* Costa Rica* Correlation (RGDP, private credit) Domincan Rep. Thailand* United Kingdom Romania* Brazil* New Zealand Norway* Poland* 0.5 Colombia* Belarus* Jamaica Malaysia Ecuador Japan* Chile* Hungary Peru Latvia Lithuania India Guatemala Denmark Israel Portugal 0 Sweden Singapore Uruguay Canada* Philippines -0.5 Av. industrial = 0.30* China* Av. developing = 0.28* -1 We do find in the data, however, a relationship between the level of the credit to GDP ratio and the use of RR. 19 Specifically, Figure 7 shows that the credit to GDP ratio is considerably higher in developing countries making active use of RRP (38 percent) compared to developing 19 Not surprisingly—given that credit is highly procyclical in almost every country —we do not find a positive correlation between RRP and the ratio of the variance of credit to the variance of GDP. But neither do we find a positive relation between RRP and the variance of credit (the coefficient is not significantly different from zero). The problem here could be one of reverse causality since an effective RRP should reduce the variance of credit and thus possibly offset the positive correlation that one may have expected. To dig deeper into this issue, we would need to find valid instruments for the variance of credit. 11 countries that do not (29 percent). 20 Even more telling, Figure 8 shows that, for active countries, the correlation between RR and the cycle (a measure of the intensity of RRP as a macroeconomic stabilization tool) is an increasing function of the level of credit to GDP. In other words, the higher the level of credit in the economy, the more countercyclical the use of RR. 21 This might reflect the fact that emerging economies with higher levels of credit (relative to GDP) tend to have more pronounced fluctuations triggered by the capital flows cycle, as analyzed below. Figure 7: Private credit (% of GDP) for developing countries (1970-2011) 0.4 Active RRP 0.35 0.3 Passive RRP Private credit (% of GDP) 0.25 0.2 0.15 0.1 0.05 0 1 20 This difference is significant at the one percent level. 21 This relation continues to hold even if we control for the level of GDP per capita, suggesting that what matters is the credit to GDP ratio regardless of the level of development of the economy. 12 Figure 8: Cyclicality of RRP versus private credit (active countries, 1970-2011) 1 Correlation (RGDP, reserve requirement) 0.75 0.5 0.25 0 -0.25 -0.5 -0.75 Corr(RGDP, res. Req.) = -0.41 + 0.16* x ln(priv cred) R2 = 0.14 -1 2 2.5 3 3.5 4 4.5 5 5.5 ln(Private Credit (% of GDP)) 2.4 What is the relation between RRP and monetary policy? The fact that, since 2004, 90 percent of active developing countries have pursued countercyclical RRP is remarkable when contrasted with the countercyclical use of the interest rate, clearly the most ubiquitous and flexible policy tool. Indeed, as illustrated in Figure 9, while all industrial countries (black bars) exhibit a positive correlation between the cyclical components of the policy rate and real GDP (indicating countercyclical monetary policy), only 59 percent (or 22 of 37) of developing countries do so. In terms of correlations significantly different from zero, the figure for industrial countries is 73 percent (or 11 of 15) compared to 27 percent (or ten of 37) for developing countries. In fact, the average correlation for industrial countries is 0.40 (and significantly different from zero), compared to 0.07 (and not significantly different from zero) for emerging countries. In other words, the average industrial country pursues countercyclical monetary policy, whereas the average emerging country is a-cyclical. Again, the averages over such a long period of time (1970–2011) mask some important changes. As analyzed in detail in Vegh and Vuletin (2013), in the last decade or so we have observed a process of “graduation” whereby many developing countries have switched from pro- cyclical to countercyclical monetary policy. In terms of our sample, we see that for the pre-2004 period just 16 percent (or six of 37) of developing countries exhibit a positive (and significantly different from zero) correlation, as illustrated in Figure 10, Panel a. The corresponding figure increases to 41 percent (or 15 of 37) for the post-2004 period (Figure 10, Panel b). In fact, we can see that the average correlation for developing countries increases from 0.07 (and not significantly different from zero) to 0.27 (and significantly different from zero). As argued below, this graduation process can be explained by a fall in the fear of free falling (i.e., a reduction in the need to defend the currency in bad times). 13 Correlation (RGDP, central bank interest rate) Correlation (RGDP, central bank interest rate) -1.25 0.25 0.75 -1 -0.75 -0.5 -0.25 0 0.5 1 1.25 Uruguay* -1.25 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 1.25 Ecuador (pre-dollarization)* Domincan Rep.* Uruguay* Argentina* India* India* Argentina* Mexico* Domincan Rep.* Chile* Mexico* Costa Rica Ecuador* Macedonia Romania Romania Chile* Israel Israel Venezuela Thailand Thailand Venezuela Turkey Costa Rica Hungary Latvia Belarus Sweden Sweden Latvia Turkey Portugal Hungary Croatia Norway Brazil Denmark Jamaica 14 Panama Nicaragua Portugal Japan Belarus Norway Japan Philippines Brazil Serbia Malaysia Malaysia Trinidad and Tobago Denmark* Croatia Panama Serbia Ecuador (dollarization) United Kingdom* Colombia* New Zealand* Trinidad and Tobago Colombia* Lithuania* Czech Rep. New Zealand* Nicaragua Czech Rep.* El Salvador United States* Philippines United Kingdom* United States* Singapore* Jamaica Guatemala* Poland* Av. industrial = 0.40* Lithuania* Av. developing = 0.07 Peru* Guatemala Spain* Macedonia China* Av. industrial = 0.37* Poland* Av. developing = 0.07 Germany* Spain* Australia* Germany* Canada* Singapore* El Salvador* Figure 9: Cyclicality of reserve requirement policy (1970-2011) China* Figure 10a: Cyclicality of interest rate policy (1970-2004) France* Peru* Honduras* Australia* Switzerland* France* Euro-17* Canada* Switzerland* Euro-17* Figure 10b: Cyclicality of interest rate policy (2005-2011) United Kingdom* 1.5 Switzerland* Czech Rep.* United States* Euro-17* 1.25 Trinidad and Tobago* Correlation (RGDP, central bank interest rate) Denmark* Malaysia* El Salvador* Poland* Sweden* Honduras* New Zealand* Latvia* 1 Chile* Israel* Guatemala* Colombia* Canada* Japan* Norway* Peru* Australia* Venezuela* Singapore* 0.75 Romania India* Lithuania Thailand Philippines Mexico 0.5 Argentina China Panama Croatia Serbia 0.25 0 Hungary Brazil Jamaica Belarus Nicaragua -0.25 Turkey Ecuador (dollarization) Uruguay Macedonia -0.5 -0.75 Costa Rica* -1 Av. industrial = 0.73* Av. developing = 0.27* -1.25 -1.5 How are RRP and monetary policy related over the business cycle? To answer this question, we constructed a “policy mix matrix” (Table 1) that classifies countries according to the cyclical properties of RRP and monetary policy. Since countries may be pro-cyclical, a- cyclical, or countercyclical, there are nine possible cells. Notice that most industrial countries are on the second row because most of them are a-cyclical when it comes to RRP (i.e., they do not use RRP for macroeconomic stabilization purposes). In fact, most industrial countries (11 of 15 or 73 percent) fall in the [2,3] cell, colored orange, given that their monetary policy is countercyclical. In contrast, 10 developing countries (or 27 percent of all developing countries in our sample) fall in the third row because they exhibit countercyclical RRP. Of these 10 developing countries, six fall in the yellow cell [3,2] given that they have had a-cyclical monetary policy, while the remaining four fall in cell [3,3] because they have had countercyclical monetary policy. 15 Table 1: Policy mix matrix (1970-2011) Central bank interest rate policy Pro-cyclical (1) A-cyclical (2) Counter-cyclical (3) Pro-cyclical (1) Jamaica Australia, Canada, Colombia, Czech Rep., Reserve Israel, Japan, Denmark, Ecuador A-cyclical Argentina,Chile, Macedonia, Nicaragua, (dollarization), El Salvador, Costa Rica, Dominican Rep., requirement (2) Ecuador (pre-dollarization), Panama, Philippines, Euro-17, France, Guatemala, Portugal, Romania, Serbia, Honduras, New Zealand, policy India, Mexico, Uruguay Thailand, Turkey Norway, Poland, Singapore, Spain, Switzerland, Trinidad and Tobago, United Kingdom, United States Counter- Belarus, Brazil, China, Germany, cyclical (3) Croatia, Hungary, Latvia, Lithuania, Malaysia, Peru Sweden, Venezuela Substitutes Complements 16 As a matter of definition, we will refer to countries falling in cell [3,3] as using RRP as a complement to monetary policy in the sense that both policies reinforce each other. 22 These are cases in which both RRP and monetary policy are countercyclical, so that in good (bad) times both RR and the policy interest rate are increased (lowered). In contrast, countries that fall in cell [3,2] and, potentially, cell [3,1] are cases in which RRP acts as a substitute for interest rate policy because reserve requirements perform the function that the interest rate cannot, due to the need of either defending the currency in bad times or not attracting more capital inflows in good times. Further, we interpret cell [2,1] as a “bad place” to be because here countries feel compelled to use monetary policy pro-cyclically (for reasons to be discussed below) but are not taking advantage of RR as a second instrument. On the other hand, we can view [2,3] as the “promised land,” in terms of policy making. The reason is that here countries can use monetary policy for countercyclical purposes without fearing the effects of countercyclical interest rates on the exchange rate and/or capital inflows, presumably reflecting a large degree of policy and institutional credibility. 23 24 As before, we break the sample into before and after 2004 and construct the policy matrices for each period, as illustrated in Table 2. Comparing the pre and post-2004 policy mix matrices in Tables 2, Panels a and b, three important changes are worth noting: (i) the number of developing countries pursuing pro-cyclical interest rate policy falls from seven in the early period to just one in the more recent period; (ii) the number of developing countries engaging in countercyclical RRP increases from 7 to 15, and (iii) developing countries using monetary policy and RRP as substitutes (as captured in cell [2,2]) increases from 5 to 8, whereas developing countries using monetary policy and RRP as complements (as captured in cell [3,3]) increases from 2 to 7. 22 Notice that this definition of complements and substitutes only focuses on the comovement of the two policy instrument and thus should not be confused with the game theoretical concepts of strategic complementarity and substitutability à la Bulow, Geanakoplos, and Klemperer (1985). 23 Vegh and Vuletin (2013) formally link graduation from monetary procyclicality to institutional improvement. 24 Of course, countries may still want to use RR as systemic risk-driven macroprudential policy. In theory, as well, countries might want to use both monetary policy and RR countercyclically if they feel that the monetary transmission channel of monetary policy is not strong enough. Revealed preference, however, suggests that this is not the case and that industrial countries (many of which in fact do not even have legal RR) do not feel the need to supplement monetary policy with RR. Hence, our label of “promised land” is meant to convey an observed state of affairs rather than an optimal prescription regarding the mix of monetary policy and RR in mature countries. In fact, this is clearly an area that warrants further analytical work. 17 Table 2a: Policy mix matrix (1970-2004) Central bank interest rate policy Pro-cyclical (1) A-cyclical (2) Counter-cyclical (3) Pro-cyclical Colombia (1) Belarus, Costa Rica, Croatia, Reserve Czech Rep., Denmark, El Australia, Canada, Euro-17, requirement Salvador, Guatemala, Israel, Argentina, Chile, Dominican Lithuania, New Zealand, Poland, policy A-cyclical Jamaica, Japan, Macedonia, Rep., Ecuador, India, Mexico, Singapore, Spain, Switzerland, (2) Nicaragua, Norway, Panama, Uruguay United Kingdom, United Philippines, Portugal, States, Romania, Serbia, Thailand, Trinidad and Tobago, Turkey Brazil, Hungary, Latvia, Counter- Malaysia, Sweden, China, France, Germany, Peru cyclical (3) Venezuela Substitutes Complements 18 Table 2b: Policy mix matrix (2005-2011) Central bank interest rate policy Pro-cyclical (1) A-cyclical (2) Counter-cyclical (3) Pro-cyclical Jamaica Trinidad and Tobago (1) Australia, Canada, Chile, Czech Rep., Denmark, El Reserve Ecuador (dollarization), Salvador, Euro-17, Guatemala, requirement Hungary, Macedonia, Mexico, Honduras, Israel, Japan, New A-cyclical policy Costa Rica Nicaragua, Panama, Zealand, Norway, Singapore, (2) Philippines, Serbia, Thailand, Sweden, Switzerland, United Uruguay, Kingdom, United States Argentina, Belarus, Brazil, Colombia, India, Latvia, Counter- China, Croatia, Lithuania, Malaysia, Peru, Poland, cyclical (3) Romania, Turkey Venezuela Substitutes Complements 19 In this context, one can imagine different “routes” that countries may take in the policy journey from cell [2,1], a “bad place to be,” to cell [2,3], the “promised land.” One route would be the “direct route” that Chile took. Notice that Chile went from cell [2,1] in the pre-2005 period to cell [2,3] in the post-2004 period. This is remarkable because, in our view, it can only happen in the context of a notable improvement in policy/institutional credibility, as discussed below. In such a case, a country like Chile may not need to resort to countercyclical RRP, which is presumably the situation of a typical industrial country. 25 A more common route (involving several stages) would be to go from cell [2,1] to cell [3,2], then to cell [3,3], and finally to cell [2,3]. One could even add an additional stage in which countries would go from cell [2,1] to cell [2,2] and only then to cell [3,2]. Mexico and Uruguay, for instance, have gone from cell [2,1] in the pre-2005 period to cell [2,2] in the post-2004 period. This is presumably an improvement in the policy mix because they have graduated from monetary policy pro-cyclicality, but the fact that they are in cell [2,2] implies that they have no countercyclical tool. 26 In principle, they would benefit from moving to cell [3,2] where they would be using RR as a countercyclical tool. 2.5 How does foreign exchange market intervention fit into the picture? As is well known, developing countries often intervene in foreign exchange markets, buying (selling) foreign currency in good (bad) times to prevent an appreciation (depreciation) of the domestic currency. The extent of foreign exchange market intervention, however, will depend on the particular exchange rate regime/policy mix that policy makers choose in order to deal with the capital flow cycle. Clearly, under a fixed or predetermined exchange rate regime, foreign exchange market intervention will be geared towards maintaining a particular value of the exchange rate while the monetary policy should play a minor role, if any. 27 Under a relatively clean floating exchange rate, foreign exchange market intervention would be minimal with monetary policy playing a much larger role. The role of RRP under different exchange rate regimes has been much less studied both theoretically and empirically. In any event—and since over long periods of time, the exchange rate regime is a policy choice—we would like to let the data speak and tell us the extent of foreign exchange market intervention and how it relates to both monetary policy and RRP. Figure 11 plots the correlation between (the cyclical components of) international reserves and real GDP for our sample. It turns out that 54 percent (or 20 of 37) of developing countries exhibit a positive (and significantly different from zero) correlation, clearly suggesting 25 This would be a “one-step” graduation process, which is the one (implicitly) highlighted by Vegh, and Vuletin (2013) for the case of monetary policy and Frankel, Vegh, and Vuletin (2013) for the case of fiscal policy. For a more detailed analysis of institutional changes that may have allowed Chile to graduate, see Frankel (2011). 26 Not even fiscal policy because, according to Frankel, Vegh, and Vuletin (2013), both Mexico and Uruguay continue to be procyclical even in the more recent period. 27 Of course, under perfect capital mobility, there is no role for monetary policy. Under imperfect capital mobility, there may be a role for monetary policy even under fixed exchange rates; see Lahiri and Vegh (2003) and Flood and Jeanne (2005) for models along these lines. 20 a considerable degree of foreign exchange market intervention. 28 The corresponding figure for industrial countries is 38 (or five of 13). In terms of the before and after (not shown), the figure for developing countries has increased from 33 percent (or 11 of 33 countries) to 48 percent (or 16 of 33). Figure 11: Cyclicality of international reserves (1970-2011) 1.25 Argentina* 1 Macedonia* United Kingdom* Domincan Rep.* Uruguay* Peru* Croatia* Thailand* Correlation (RGDP, foreign reserve) Honduras* Philippines* 0.75 Costa Rica* Czech Rep.* New Zealand* Trinidad and Tobago Lithuania* Portugal Venezuela* Chile* Poland* Colombia* Turkey* Latvia* Norway* Spain Romania Nicaragua 0.5 Switzerland France* Mexico* Malaysia Ecuador* Singapore India China Japan Australia 0.25 Brazil Canada Serbia 0 Israel Jamaica Sweden Panama Germany Denmark -0.25 Guatemala El Salvador Hungary* United States* -0.5 -0.75 Av. industrial = 0.11 -1 Av. developing = 0.26* -1.25 How are monetary policy and foreign exchange market intervention related? In order to investigate this, Figure 12 plots the correlation between (the cyclical components of) international reserves and real GDP on the vertical axis against the correlation between (the cyclical components of) the policy interest rate and real GDP. We can see that the regression line is negatively sloped (and significantly so). This means that countries that pursue a pro-cyclical monetary policy (i.e., have a negative correlation between the policy rate and the business cycle) intervene more heavily in foreign exchange markets. As discussed below, such a pattern would be consistent with the idea that developing countries might feel the need to defend their currencies in bad times by both raising interest rates and selling foreign assets and that they need to fight appreciation in good times by buying foreign assets and lowering (or at least not raising) interest rates. 28 Of course, foreign exchange market intervention is automatic under any predetermined exchange rates regime and the higher correlation could reflect a preponderance of such regimes in emerging markets. This, however, is irrelevant for our point because the exchange rate regime is itself a policy choice and hence must reflect the policy makers’ preferences in terms of how much the nominal exchange rate is allowed to fluctuate. 21 Figure 12: Monetary versus foreign exchange market intervention policy (1970- 2011) 1 0.75 Correlation (RGDP, foreign reserves) 0.5 0.25 0 -0.25 -0.5 -0.75 Corr(RGDP, foreign res) = 0.26*** -0.28** Corr(RGDP, i) [6.8] [-2.6] 2 R = 0.11 -1 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 Correlation (RGDP, central bank interest rate) How are RRP and foreign exchange market intervention related? Figure 13 plots the correlation between (the cyclical components of) international reserves and real GDP on the vertical axis against the correlation between (the cyclical components of) reserve requirements and real GDP on the horizontal axis. The corresponding regression line is positive (and significantly so). This implies that countries that pursue a countercyclical RRP tend to intervene more heavily in foreign exchange markets. Hence, the picture that emerges is one where developing countries concerned about currency depreciation in bad times raise interest rates and intervene in foreign exchange markets to prevent further depreciation, yet at the same time lower reserve requirements in an attempt to stimulate credit expansion and, hence, economic activity. Figure 13: RRP versus foreign exchange market intervention (1970-2011) 1 0.75 Correlation (RGDP, foreign reserves) 0.5 0.25 0 -0.25 -0.5 Corr(RGDP, foreign reserves) = 0.26*** +0.26* Corr(RGDP, reserve requirement) -0.75 [4.9] [1.8] R2 = 0.15 -1 -1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 Correlation (RGDP, reserve requirement) 22 3. An Illustration of Policy Responses for Four LAC Countries To further illustrate the policy mix that developing countries may typically use in response to different shocks, we estimated an econometric model for four LAC countries (Argentina, Brazil, Colombia, and Uruguay) and simulate the policy response to a shock to GDP and a shock to the rate of depreciation. 29 Figure 14 illustrates the policy response to an increase in real GDP. We can see that while the policy rate is lowered (Panel A), both reserve requirements (Panel B) and international reserves (Panel C) increase. 30 Figure 14: Policy response to a real GDP shock (One standard-deviation shock) Panel A. Monetary policy response 4 3 2 1 0 ∆i -1 -2 -3 -4 -5 -6 0 1 2 3 4 5 6 7 8 9 10 29 Technically, we estimated a panel vector-autoregression model and used a narrative approach to distinguish between exogenous (to the business cycle) and endogenous changes in reserve requirements (see Box 1). The analysis covers the period 1995–2010 for Brazil and 1992–2011 for Argentina, Colombia, and Uruguay. Since our sample period covers Argentina’s Convertibility plan—and to test the robustness of our results—we estimated all panel VAR regressions excluding each country one-at-a-time and obtained the same results. For details, see the background paper by Federico, Vegh, and Vuletin (2013b). 30 We broke the sample into positive and negative increases to GDP and results were the same. 23 Panel B. Reserve requirement (endogenous) policy response 0.5 0.4 0.3 ∆ endog rr 0.2 0.1 0 -0.1 -0.2 0 1 2 3 4 5 6 7 8 9 10 Panel C. Foreign reserves response. 4 3.5 3 2.5 ∆% foreign reserves 2 1.5 1 0.5 0 -0.5 -1 -1.5 0 1 2 3 4 5 6 7 8 9 10 To interpret such policy mix, think of a recession (a fall in real GDP). Then, Figure 14 indicates that policy makers would increase the policy rate (to defend the currency), lower reserve requirements (to stimulate the economy), and sell foreign exchange (also to defend the currency). This is fully consistent with the results obtained for the whole sample above that suggested that monetary policy and RRP have acted as substitutes while RRP and foreign exchange market interventions have been complements. Figure 15, captures the response of the three policy instruments to an increase in the rate of depreciation. The policy rate increases (Panel A), there is no significant change in RR (Panel B), and international reserves fall (Panel C). This policy response is entirely consistent with the idea that, when faced with a depreciating currency, policy makers in emerging markets increase interest rates and sell foreign assets to try to stabilize the value of the currency. 24 Figure 15: Policy response to a nominal exchange rate depreciation shock Panel A. Monetary policy response 6 4 2 ∆i 0 -2 -4 -6 0 1 2 3 4 5 6 7 8 9 10 Panel B. Reserve requirement (endogenous) policy response 0.2 0.15 0.1 0.05 ∆ endog rr 0 -0.05 -0.1 -0.15 -0.2 0 1 2 3 4 5 6 7 8 9 10 Panel C. Foreign reserves response 2 1 ∆% foreign reserves 0 -1 -2 -3 -4 0 1 2 3 4 5 6 7 8 9 10 25 Box 1: The narrative approach to identification Following Blanchard and Perotti (2002), the traditional strategy used to identify fiscal, monetary, and (more recently) reserve requirement shocks has relied on vector autoregression analysis (VAR) together with timing assumptions regarding the dynamics between policy and output. On the fiscal side—using quarterly data—it has been assumed that government spending requires at least one quarter to respond to news about the state of the economy and that active fiscal policy may affect output in the same quarter (i.e., government spending is more “exogenous” or slow-moving than output). The opposite has been assumed on the monetary front (using monthly data). While changes in policy rates are not allowed to have contemporaneous (i.e., within the month) effects on macroeconomic variables, developments in the stance of the economy are allowed to affect monetary policy. Mimicking this monetary identification strategy, several recent papers (e.g., Glocker and Towbin, 2012; Tovar, Garcia- Escribano, and Vera Martin, 2012) have analyzed the impact of reserve requirement changes on credit and economic activity obtaining—in some circumstances—some puzzling results. For example, Tovar, Garcia-Escribano, and Vera Martin (2012) find that an increase in reserve requirements increases private credit. Federico, Vegh, and Vuletin (2013b), rely on the use of Romer-Romer type narratives to identify shocks to reserve requirements. 31 To the best of our knowledge, this is the first instance in this literature in which such an approach has been followed. Using historical documents, including IMF and central banks reports, we classify changes in reserve requirements into (i) endogenous changes, which were mainly motivated by current or projected output fluctuations (i.e., when output growth differs from normal) and (ii) exogenous changes, which were triggered by reasons exogenous to the business cycle, including microprudential factors and financial liberalization. When we then incorporate the exogenous changes in reserve requirements into our VAR analysis, we find that, as expected, higher reserve requirements reduce private credit and output. 32 Our analysis also confirms that endogenous changes in reserve requirements, in turn, respond positively and strongly to output shocks. This striking difference clearly shows the importance of following an identification strategy that is able to isolate policy innovations that are exogenous to the business cycle. 31 See, for example, Romer and Romer (2010) and Riera-Crichton, Vegh, and Vuletin (2012) for a discussion of the use of narratives to evaluate the effect of taxation policy, and Romer and Romer (2004) and Coibion (2012) for the case of monetary policy. 32 Moreover, past output fluctuations are poor predictors of changes in exogenous reserve requirements changes, supporting our narrative categorization. 26 4. Policy Rationale The previous section has illustrated the fact that developing countries tend to pursue an active RRP when they are pursuing an either a-cyclical or pro-cyclical monetary policy. This suggests that these countries may have been unable to lower (increase) interest rates in bad (good) times for fear of letting their currency depreciate (appreciate) too fast. In other words, these countries may have resorted to reserve requirements as a second policy instrument in their quest to hit two different targets: real GDP and the exchange rate. This section develops this argument and examines the country characteristics that may affect the choice of such a policy mix. 4.1 The need for a second instrument In our view of the world, the critical feature of emerging markets’ business cycle that may trigger the need for a second policy instrument is the fact that there is a negative correlation between (the cyclical components of) the nominal exchange rate and real GDP. 33 This is illustrated in Figure 16 where we can see that 89 percent of developing countries (i.e., 33 out of 37) exhibit a negative correlation. By and large, then, the typical business cycle in an emerging country entails an exchange rate that is below (above) trend in good (bad) times. In other words, in good (bad) times the currency is relatively appreciated (depreciated). Further, the depreciation of the currency in bad times is often sharp and triggers large capital outflows which, in turn, exacerbate the currency depreciation leading to a vicious spiral of capital outflows, depreciation, and recession. More often than not, policy makers feel that they have little choice but to defend the currency because if the currency goes down so does the macroeconomy. 34 The need to defend the domestic currency in bad times is best exemplified by IMF policy advice during the 1997 Asian crisis. 35 33 Following standard practice in open economy macroeconomics, we are defining the nominal exchange rate as units of domestic currency per unit of foreign currency and hence an increase (decrease) entails a nominal depreciation (appreciation) of the domestic currency. 34 While we will be thinking mainly of exchange rates as driven by the capital inflows cycle, we should note that another factor behind the negative correlation of the nominal exchange rate and real GDP could be fluctuations in commodity prices for commodity-producing emerging markets. In this scenario, a higher (lower) price of the commodity would lead to an appreciation (depreciation) of the currency. Cordella and Gupta (2014) discuss the cyclical properties of different currencies and their main determinants. 35 See Fischer (1998). Of course, many prominent economists at the time, most notably Stiglitz, raised serious questions about such a policy (see Furman and Stiglitz, 1998). More generally, the effectiveness of raising interest rates to defend the currency has been the subject of an intense debate (see Montiel, 2003, for a detailed analysis). At a theoretical level, Hnatkovska, Lahiri, and Vegh (2008) show a model in which small increases in interest rates appreciate the currency but larger ones can actually depreciate the currency. 27 Figure 16: Cyclicality of nominal exchange rates (1970-2011) 1 Correlation (RGDP, nominal exchange rate) 0.75 Trinidad and Tobago 0.5 China* United States Switzerland Germany Brazil* 0.25 Euro-17 Turkey Japan 0 Norway Philippines Honduras Latvia Canada Belarus Uruguay Lithuania Jamaica Costa Rica Nicaragua Poland -0.25 Denmark* El Salvador Chile* Hungary Australia* Guatemala Singapore* Domincan Rep. Czech Rep.* Colombia* Israel* United Kingdom* Venezuela* Croatia* Sweden* India* France* Serbia* Portugal* -0.5 Macedonia* Thailand* Malaysia* Romania* Mexico* Argentina* Ecuador* Peru* New Zealand* Av. industrial = -0.18* -0.75 Spain* Av. developing = -0.25* -1 Figure 17 illustrates the defense of the currency in bad times by plotting along the horizontal axis the correlation between the nominal exchange rate and GDP and along the vertical axis the correlation between monetary policy and GDP. We can see that the regression line is significantly positive suggesting that the more the currency depreciates in bad times (i.e., the more to the left we are along the horizontal axis) the more pro-cyclical monetary policy is (i.e., the more pronounced is the interest rate defense of the currency). Figure 17: Monetary versus “exchange defense” policy (1970-2011) 1 Correlation (RGDP, central bank interest rate) 0.75 0.5 0.25 0 -0.25 -0.5 -0.75 Corr(RGDP, i) = 0.27*** +0.45** Corr(RGDP, e) [4.7] [2.5] R2 = 0.11 -1 -1 -0.75 -0.5 -0.25 0 0.25 0.5 Correlation (RGDP, nominal exchange rate) 28 The need to increase interest rates during bad times (or at least not lower them) presents a dilemma for emerging market policy makers because, in and of itself, higher interest rates will only exacerbate the fall in output. The opposite situation (good times) presents an analogous dilemma: in good times—when capital is flowing in, the currency is appreciating, the economy is overheating, and inflation is going up—a countercyclical monetary policy of higher interest rates to cool down the economy will only exacerbate capital inflows and currency appreciation because it makes domestic-currency assets more attractive for foreign investors. This “fear of capital inflows” may induce policy makers not to raise interest rates in good times and use higher reserve requirements to cool down the economy. In contrast, industrial countries are typically much less concerned with exchange rate fluctuations either because (i) the exchange rate correlates positively with the business cycle (as is the case for the Euro-area, Japan, U.S., and Germany in Figure 16) or, (ii) if the exchange rate does correlate negatively, it is no cause for concern. In other words, even if the currency depreciates in bad times in an industrial country, markets would not interpret this as a signal that policy makers are losing control of the macroeconomy and/or capital outflows will accelerate. In the United States, for example, fluctuations in the value of the dollar have normally little direct impact on Fed’s policy. The need for a second instrument may, of course, be thought of in terms of the so-called “impossible trinity,” which holds that a country cannot simultaneously achieve a fixed exchange rate, independent monetary policy, and high capital mobility but can only choose two of the three. In general, industrial countries do not care about the exchange rate and are thus content with having an independent monetary policy and high capital mobility. In contrast, emerging countries are typically quite concerned about the value of the domestic currency (the so-called “fear of floating”) and use the policy rate to smooth out fluctuations in the exchange rate, thus giving up in principle independent monetary policy. In this light, resorting to a second instrument (i.e., using RR to do the work that the policy rate would do) may be viewed as an attempt to return to an independent monetary policy and achieve the impossible trinity. 4.2 Why do RR often serve as the second instrument? Having offered a plausible rationale for a second instrument, the next question is: what should that instrument be? Clearly, foreign exchange market intervention cannot provide this second instrument simply because it does not have a direct effect on output. In other words, if interest rates are raised in bad times to defend the currency (thus worsening the accompanying recession), there is little that foreign exchange market intervention can do to stimulate output. It is thus not surprising that, as illustrated in Figure 12, there is a negative correlation between monetary policy and foreign exchange market intervention, which suggests that foreign exchange market intervention is pursuing the same objective of preventing the currency from depreciating (appreciating) in bad (good) times. In contrast, RR seem a natural choice as a second instrument because, in principle, both RR and policy interest rates should have a similar effect on the economy. To illustrate this, Figure 18 shows the effects of shocks to both instruments in the context of the 4 LAC countries mentioned above. We can see (first two rows) that increases in both RR and the policy interest rate lower real GDP and increase the interest rate spread. While higher RR also reduce private 29 credit, however, higher interest rate do not. 36 37 Leaving the latter anomaly aside, theory would also suggest that tighter monetary policy and higher RR should have similar effects. In Bengui, Vegh, and Vuletin (2013), for instance, an increase in reserve requirements has the standard effect of directly increasing the interest rate spread and reducing firms’ demand for credit. A higher policy rate has a similar effect by raising the lending rate that banks charge to firms on account of the higher interest rate on public debt. Figure 18: Relative effect of reserve requirement versus monetary policy Panel A. Effect of reserve req. policy on real Panel B. Effect of monetary policy on real GDP. GDP. 0.4 0.4 0.2 0.2 ∆% real GDP ∆% real GDP 0 0 0 1 2 3 4 5 6 7 8 9 10 -0.2 -0.2 -0.4 -0.4 -0.6 -0.6 0 1 2 3 4 5 6 7 8 9 Panel C. Effect of reserve req. policy on interest Panel D. Effect of monetary policy on rate spread. interest rate spread. 2.5 2.5 2 2 1.5 1.5 ∆ interest rate spread ∆ interest rate spread 1 1 0.5 0.5 0 0 -0.5 -0.5 -1 -1 -1.5 -1.5 -2 -2 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 36 This latter result is, of course, surprising. We speculate that this result could reflect the response of both credit demand and credit supply to policy interest rates actions (Kashyap and Stein, 2000). By raising interest rate spreads, an increase in policy interest rates should reduce credit demand. However, it could also increase the bank incentives to relax other non-price credit conditions. This later channel should, in principle, be more prevalent in economies such as the ones analyzed in this study with underdeveloped and small credit markets. 37 Credit is measured as the percentage change in real credit. 30 Panel E. Effect of reserve req. policy on private Panel F. Effect of monetary policy on credit private credit 2.5 2.5 2 2 1.5 1.5 1 1 ∆% private credit ∆% private credit 0.5 0.5 0 0 -0.5 -0.5 -1 -1 -1.5 -1.5 -2 -2 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Note: Panels A and B use a four-variable panel VAR. The four variables are exogenous changes in reserve requirement, changes in central bank interest rate, real GDP growth rate, and inflation; in that order. Panels C and D use a five-variable panel VAR. The five variables are exogenous changes in reserve requirement, changes in central bank interest rate, real GDP growth rate, inflation, and interest rate spread; in that order. Interest rate spread is defined as lending minus deposit interest rates. Panels E and F use a five-variable panel VAR. The five variables are exogenous changes in reserve requirement, changes in central bank interest rate, real GDP growth rate, inflation, and private credit growth rate; in that order. Both the reserve requirement and monetary (i.e., central bank interest rate) shocks involve a one standard-deviation shock. Changing the order of changes in central bank interest rate does not affect qualitatively our main (cumulative) results. Dashed lines refer to 95 percent confidence intervals constructed using Monte Carlo simulations. 4.3 Which country characteristics explain different policy mixes? An interesting question to ask is what country characteristics are associated with the adoption of different policy mixes. Table 3 addresses this particular question. This table comprises all developing countries in our sample and asks the following question: What percentage of countries that follow a particular policy mix has suffered a currency crisis during our sample period? 31 Table 3: Currency crises and policy mix Central bank rate policy Counter- Pro- cyclical A-cyclical Total cyclical Pro- cyclical Reserve Requirement A-cyclical 0.36 0.26 0.19 0.28 Policy Counter-cyclical 0.41 0.19 0.30 Total 0.36 0.31 0.19 32 We can see that 36 percent of countries that follow pro-cyclical monetary policy have had at least one currency crisis during our sample period. In contrast, just 19 percent of countries following countercyclical monetary policy have had a currency crisis. This is fully consistent with the idea that pro-cyclical monetary policy is more likely to be present in situations in which the need to defend the currency is more pressing. On the other hand, the table also indicates that there is not a noticeable difference when it comes to the use of RR. Table 4 sheds light on the relation between the level of credit and the use of RR. Specifically, it asks the question: suppose that a country follows a countercyclical RRP, what is its corresponding ratio of credit to GDP ratio? We can see that if a country follows a countercyclical RRP, its average credit to GDP ratio is 41 percent compared to just 29 percent if its RRP is a-cyclical. In a similar vein, Table 5 addresses the relationship between capital account openness and the policy mix using the Chinn-Ito index of capital account openness. It says that if a country is following countercyclical RRP, its capital account openness is considerably higher than countries that are following a-cyclical RRP (0.22 versus 0.09, respectively). 5. Microprudential Effects of Business Cycle Management The previous sections have offered novel evidence about the increasing use of RR as an additional tool in policy makers’ hands to cope with macroeconomic volatility. In particular, and not surprisingly, RR are more likely to be actively used when an open capital account and a pro- cyclical currency limit the effectiveness of monetary policy in controlling credit volumes (and/or economic activity), at least for politically “acceptable” levels of the exchange rate. In other words, RR are viewed as an useful additional tool in the presence of “impossible-trinity” dilemmas; not surprisingly, countries, such as Turkey or Brazil, have used RR aggressively when they could not, or did not want to, use interest rates in a forceful manner, see Box 2. Few would disagree with the fact that, by smoothing the business cycle, the countercyclical use of RR has the potential to smooth out the business cycle and reduce the build-up of systemic risk in an economy. However, less is known about how RR, or other prudential tools such as capital requirements, limits on LTV, and so forth affect individual institutions’ risk taking incentives. In other words, what are the microprudential properties of the different macroprudential tools? 33 Table 4: Credit to GDP ratio and policy mix Central bank rate policy Pro- Counter- A-cyclical Total cyclical cyclical Pro- cyclical Reserve A-cyclical 0.24 0.41 0.25 0.29 Requirement policy Counter- 0.29 0.50 0.41 cyclical Total 0.24 0.36 0.33 34 Table 5: Capital account openness and policy mix Central bank rate policy Pro- Counter- A-cyclical Total cyclical cyclical Pro- cyclical Reserve A-cyclical -0.14 0.14 0.30 0.09 Requirement policy Counter- -0.05 0.48 0.22 cyclical Total -0.14 0.09 0.35 Note: The Chinn-Ito index of capital account openness is a standardized index of de jure capital account openness. The index is based on the binary dummy variables that codify the tabulation of restrictions on cross-border financial transactions reported in the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER). The index varies between 2.44 (most financially open) and -1.86 (least financial open). See http://web.pdx.edu/~ito/Chinn- Ito_website.htm for a detailed explanation regarding the construction of such index. 35 Box 2: Macroprudential policy in emerging markets: The cases of Brazil and Turkey During late 2010through early 2011, large capital inflows into Brazil fueled by the real carry trade hampered the effectiveness of monetary policy set in response to rising inflation expectations and a credit boom. The Brazilian Central Bank came to the conclusion that monetary policy alone could not address internal and external “imbalances” and decided to introduce a number of macroprudential measures to help monetary policy. In particular, it increased unremunerated RR on term deposits from 15 to 20 percent and additional remunerated RR (on demand and term deposits) from 8 to 12 percent. These measures reversed the loosening of RR that took place in late 2008 to create liquidity during the financial crisis. In addition, to curb credit growth, the tax on financial operations (IOF) was raised from 1.5 to 3.0 percent (annualized), and applied to credit operations for individuals, on a daily basis, for a maximum of one year. Capital requirements for consumer loan were also increased through a change in risk weights: for the fast-growing segment of vehicle financing, the weight was increased from 75 to 150 percent, which was equivalent to an increase in capital requirement of 8 to 16.5 percent. At the same time, in order to avoid the build-up of excessive maturity mismatches (in Brazil, bank liabilities have very short maturity and/or duration) the loan to value ratio (LTV) on vehicle loans was increased, penalizing longer maturities—the maximum LTV was set to 80 percent for loans between 24 and 36 months, to 70 for loans between 36 and 48 months, and to 60 percent for loans between 48 and 60. Finally, in order to curb the carry trade, the IOF on nonresident portfolio investments was increased from 2 to 4 percent (and ultimately to 6) and a 60 percent unremunerated RR on banks’ short position in the forex spot market was imposed; additional measures were also adopted to limit banks’ exposure in forex derivative markets. In 2012, faced with less favorable global conditions, the central bank adjusted the macroprudential stance reducing the additional RR on demand deposits from 12 to 6 percent first and then to zero, and the additional RR on time deposits from 12 to 11 percent. Finally, in the summer of 2013, confronted with strong currency depreciation pressures triggered by the Federal Reserve’s “tapering talks,” the IOF on forex operation was abolished; in December 2013, the IOF on cash withdrawals in foreign countries was increased from 0.38 to 6.38 percent. Faced with policy dilemmas similar to Brazil’s, Turkey also relied heavily on macroprudential measures. RR changed very frequently. For instance, the RR on short term foreign currency deposits was reduced from 6 to 5 percent during the financial crisis and then was progressively increased up to 16 percent, only to be reduced again to the current 11.5 percent. RR for domestic currency deposits followed a similar pattern. In addition to the active use of RR—which on occasions moved in the opposite direction to the policy interest rate to contain the appreciation of the exchange rate and the cost of sterilized foreign exchange interventions—the Central Bank of Turkey allowed banks to satisfy a certain share of their RR for domestic currency deposits with foreign exchange or gold; interestingly, such a share has become an additional policy tool in the hands of the monetary authorities. Finally, during 2011, the Banking Regulation and Supervision Agency took a number of measures meant to prevent the build-up of excessive risk in the expansionary phase of the cycle. In addition to an (implicit) 25 percent nominal credit growth ceiling communicated to banks (relying on moral suasion), LTV ceilings on real estate loans were implemented; risk weights on consumer loans were increased (from 100 to 150 percent on maturities below two years, and from 100 to 200 percent on longer maturities); larger provisions were required for such loans and restrictions were imposed on the increase in credit cards limits for consumers with large outstanding credit. Sources: Pereira da Silva and Harris (2013), Central Bank of Brazil, IMF (2012) and Central bank of Turkey. 36 5.1 Trade-offs over the business cycle In order to properly assess how the use of different macroprudential tools may affect individual banks’ risk taking incentives we need to move away from the standard modelling of macroprudential policies in the DSGE literature that, due to just focusing on externalities, ends up treating all prudential instruments as alternative forms of Pigouvian taxation. Different prudential instruments do affect financial intermediaries’ incentives to take on risk differently and, in certain instances, the micro and macro prudential properties of specific policies may not go hand in hand. To be able to discuss some of the trade-offs that regulators may face along the business 38 cycle, it is important to focus on a few critical distortions that justify the presence of an active prudential regulation. More precisely, we should keep in mind that: (1) because of limited liability, individual banks take on excessive risk; (2) because of deposit insurance, they opt for cheap, short term, and unstable sources of funds, transferring to the public the cost of liquidity insurance; and (3) because of externalities, they do not internalize the effect of (i) their lending policies on the stability of the financial system (and tend to lend too much in good times and too little in bad times), (ii) they do not internalize the effects of their funding strategy on other banks’ liquidity risk. In the current jargon, the first two distortions call for microprudential policies, the third for macroprudential ones. Of course, this categorization, like all categorizations, is somehow arbitrary. However, it could be helpful in understanding different rationales for the countercyclical use of prudential measures. We will start by taking the macroprudential policy stance as given and will discuss in more depth the possible trade-offs at the end of the section. Building on Cordella and Pienknagura (2013), we now discuss whether macroprudential policies implemented either through RR (as discussed in previous sections), through countercyclical capital requirements as in Spain, Colombia, and Peru, 39 or through other forms of taxation may affect individual banks’ risk taking incentives. We start our analysis by focusing on the microprudential dimension in order to understand how individual banks’ risk taking incentives may vary across the credit cycle. When the credit cycle is driven by the project cycle, that is, if it is driven by the relative abundance or scarcity of quality projects, then banks tend to be naturally prudent on the upside not to lose their profit opportunities. This suggests that tensions may arise between the stance of macro and microprudential policies and that the regulator should carefully assess the nature of the financial frictions (principal agents versus externalities) in order to properly set the policy stance. This finding is not as surprising as one may think, and there is plenty of evidence that in periods of economic boom (that is, when there is abundance of good projects, or at least of projects perceived as good) supervisors are usually satisfied with the levels of banks’ capital and 38 The focus of this section is to discuss the policy trade-offs that may arise between different instruments when prudential regulation has to be adjusted along the business cycle. In other words, we focus on changes rather than on levels. 39 For a comprehensive discussion of these experiences, see Fernández de Lis and García-Herrero (2013). 37 provisions while, in period of downturns, they become more worried and may ask for remedial actions which, from a macroprudential perspective, may make things worse. When the cycle is driven by the cost of financing, instead, risk taking tends to be pro- cyclical. The reason is that, as Dell’Ariccia et al. (2014) show in a recent paper, easy financing conditions (low interest rates) reduce the benefits of holding capital (to reduce the agency problems associated with limited liabilities) and thus foster higher leverage and additional risk taking. While this is true when banks optimally choose their capital level, 40 if banks are undercapitalized, the risk shifting channel becomes predominant and the classical Stiglitz and Weiss’ (1981) result that high interest rates lead to riskier behavior may still hold true. While macro and microprudential policies may go hand in hand when the cycle is driven by the interest rate, this “synchronicity” is more likely if banks are well capitalized. In situations in which leverage is substantial, for instance because of implicit bailout guarantees, conflicts between the macro and micro policy stance may instead still arise. The next step is to look at the effects of different macroprudential instruments on banks’ risk taking incentives. Of course, the tightening of macroprudential regulation—through an increase in minimum capital requirements, RR, or taxes on credit—leads to a reduction in credit volumes. However, while an increase in capital requirements decreases the riskiness of the banks’ loan portfolio, an increase in RR or taxes on credit is likely to increase it. How does this happen? The reason is that an increase in the banks’ external cost of funding, driven by an increase in RR, affects the banks’ bottom line only in those states of the world in which the bank actually repays its debt, that is, when it does not fail. This means that higher RR reduce the banks’ returns in the case of success and thus make banks less willing to put an additional effort to improve the quality of their portfolio. In other words, higher RR tend to exacerbate moral hazard and to induce banks to behave in a less prudent way. The idea that RR have a negative effect on the risk taking incentives of banks is not new to policy makers. Indeed, Aldo Mendes, the Brazil Central Bank’s director responsible for monetary policy, recently stated that “lowering reserve requirements helps financial stability.” 41 The practical argument, highlighted in a recent IMF (2012) report, is that “unlike an increase in capital requirements, an increase in RRs has no impact on the resilience of the banking system to loan losses [...] When RRs squeeze profitability, this can lead banks to shift into higher margin, but higher risk segments, in an effort to restore return on equity” (p. 15). As an example, the report mentions the unwanted consequences (on risk taking) of Turkey’s aggressive RR increase in early 2011. Notice that even with the absence of deposit insurance, that is, when banks have an incentive to raise additional capital to reduce their cost of funding, they will nonetheless be prone to take on excessive risk (as compared to the social optimal levels) as long as they cannot credibly commit to a given level of portfolio risk. Of course, if the regulator has to act on behalf of less sophisticated (small) depositors and/or the deposit insurer, then it will have to further tighten prudential norms and raise capital requirements to re-equilibrate the incentives that were 40 This effect is present independently of the presence of a deposit guarantee, but is magnified if the latter is in place. 41 Bloomberg, September 20, 2012. 38 additionally perturbed by the presence of public guarantees. What we are arguing here is not that RR are bad instruments per se. They are powerful macrostabilization tools that help monetary policies in presence of volatile capital flows. Also, RR together with other prudential measures targeting the liability structure of the banking sector, such as the net stable funding ratio introduced in Basel III, could strengthen the resilience of the financial system to liquidity shocks. Instead, what we are trying to stress here is that there can be trade-offs between instruments. What stabilizes liquidity or reduces aggregate risk may have negative consequences on the risk taking incentives of individual banks. Looking at the trade-offs over the business cycle, our view is that the cyclical use of taxes to dampen the credit cycle may exacerbate the tensions between micro and macro prudential policies when the cycle is driven by interest rates. When interest rates are low, banks already have an incentive to move to riskier projects and incentives are strengthened if liabilities are heavily taxed. When, instead, the cycle is driven by a great abundance of good projects, banks tend to be more prudent and the adverse effect of RR (or other form of “taxes”) on banks’ incentives is partially offset by the prudent behavior fostered by the stronger demand for credit. If, instead of taxes, capital requirements are the instruments used to deal with the business cycle, there is no major structural conflict between micro and macroprudential measures. A tightening of capital requirements reduces both aggregate risk and individual banks’ risk taking incentives. However, there can still be cyclical tensions between micro and macro objectives, when the cycle is driven by the quality of good projects. In this case, capital requirements have to be countercyclical for macroprudential reasons but not for microprudential reasons. When we look at the interaction between monetary policy and capital requirements, the picture is a much more nuanced one. In particular, capital requirements may make monetary policy more microprudential friendly. Indeed, the effect of monetary policy tightening on banks’ risk taking may depend on the level of capital requirements. For low levels of capital requirements, an interest rate hike increases the riskiness of bank’s loan portfolio, while the opposite is true if capital requirements are sufficiently high. The intuition is simple: monetary policy affects the incentives of banks in two ways: (i) by decreasing the returns in the case of success (since the increase in the cost of funding is only partially passed on to borrowers), and (ii) by increasing the banks’ skin in the game. When capital requirements are low, the first effect dominates, and this leads risk taking to follow the interest rate. Instead, when capital requirements are sufficiently high, the latter effect dominates, and interest rate hikes make banks behave more prudently. This is the same argument as in Dell’Ariccia et al. (2014) who, as we discussed above, show that if the capital structure of the bank is fixed the effect of a monetary easing on risk taking depends on banks’ leverage. Finally, in order to have a more comprehensive understanding of some additional possible trade-offs between macro and macroprudential policies it is useful to look at the how banks’ incentives are affected by the amplitude and the persistence of the business cycle. From a macroprudential perspective, a persistent positive aggregate shock is expected to relax collateral constraints, increase leverage and, via externalities, induce excessive euphoria and risk taking. 39 On the other hand, negative and prolonged negative shocks will put a dent on financial intermediaries’ net worth, tighten collateral constraints, and induce downward spirals marked by massive fire sales and deleveraging. The avoidance of such boom and bust cycles fueled by externalities is indeed the main objective of macroprudential policies. From a pure microprudential perspective the effects of the persistence and the amplitude of the business cycle are more complex and may deserve additional investigation. Cordella and Pienknagura (2014) argue that the volatility of credit demand increases monitoring effort and that this effect is stronger when shocks’ persistence is low. The fact that an increase in the amplitude of demand fluctuations increases monitoring may at first look surprising, but it should not. It is the consequence of a well-established result, proved by Oi (1961), that demand fluctuations increase firms’ profits. In their framework, this implies that expected profits in case of success increase with the volatility of demand and so do the returns associated with monitoring. This is what makes banks behave more prudently at equilibrium. These results provide additional evidence to our hint that, from a microprudential perspective, there is no reason to assume that higher volatility in the economic environment necessary leads to more risk taking. Furthermore when the business cycle is driven by the project cycle conflicts between the micro and macroprudential stance are to be expected. 6. Policy Tensions and Trade-offs The idea that the overall design of macroprudential policies should start from a careful analysis of the role that different financial frictions play in different environments (or in different moments of the business cycle) is discussed in great depth by de la Torre and Ize (2013). 42 In particular, the authors focus their attention on how the introduction of asymmetric information, externalities, bounded rationality, and mood swings in a standard Arrow-Debreu model lead to different policy implications. In particular, they emphasize that when focusing on the different transmission channels it becomes clear that not all financial fluctuations, however large and pro- cyclical, and not all financial crises, however painful, justify preventive macroprudential policy. Given that similar symptoms can reflect very different underlying forces, suitable policy responses require a reasonable sense of what is behind the observed financial turbulence, whether the inefficiencies are mainly driven by policy failures or market failures and, in the latter case, whether the relevant market failures reflect mainly public moral hazard, substantial externalities that rational players do not internalize, or irrational mood swings driven by noise traders. Finding a proper balance in macroprudential policy is further complicated by tensions and trade-offs in policy impacts when different kind of financial frictions occur simultaneously. For example, penalizing short-term wholesale funding may be good to deal with collective action failures (it induces the internalization of externalities associated with systemic runs and liquidity risk) but can exacerbate agency failures (it weakens the ability of principals to discipline agents 42 This section draws very heavily on section 6 of this paper. 40 by holding them on the “tight leash”). Mark-to-market requirements can have similarly conflicting impacts—they help markets cope with principal-agent frictions but can magnify contagious runs or spur irrational mood swings. Moreover, macroprudential policies aimed at addressing systemic risks in a collective action domain can act as a common factor (or as an aggregate shock) inasmuch as they affect financial intermediaries and financial contracts across the board. Hence, they may require countervailing macroprudential policies under an agency domain to offset the incentives distortions that the former can produce. For example, as we have discussed in the previous section, a Pigouvian tax or liquidity requirement that penalizes short-term funding under a collective-action rationale of macroprudential policy may undermine individual banks’ incentives, thereby requiring an offsetting boost in capital requirements because of microprudential concerns. More generally, the policy balance will also depend on the type of error one wishes to minimize. Policy makers that perceive credit dynamics and market forces to be strongly self- stabilizing, might prefer to minimize type I error—that is, avoid over-regulating markets or undermining market discipline, even if that implies an occasional systemic crisis. However, policy makers that perceive a large scope for collective action and cognition failures might prefer to minimize type II error—that is, prevent devastating systemic crises, even if that implies sacrificing financial market discipline and efficiency. Given these constraints, two broad macroprudential policy options can be envisaged. One option is to assemble an all-terrain regulatory framework. However, this may lead to policy inconsistencies and regulatory arbitrage, and end up being too inflexible to deal with large and relatively rapid changes in systemic risk buildup. The alternative is to develop a state-contingent (bi-modal) regulatory framework that focuses in normal times on market discipline and the classic agency frictions but shifts in exceptional times (of bubble formation or bubble bursts) to a focus on systemic risk and the destabilizing role of collective action and cognition frictions. This option presupposes, however, that the normal-times oversight is effective enough to detect fault lines in the financial system before it is too late, and that the regulator is sufficiently independent, objective, and agile to switch to exceptional-times prudential oversight when warranted. Progress towards bridging the gap between theory and practice will therefore require better identifying the main frictions and failures at work, formally incorporating them in theoretical models, assessing their welfare impact, and sorting out constrained efficiencies from constrained inefficiencies. This effort will need to be accompanied by further empirical efforts to estimate and calibrate the net impact of regulations, while at the same time gauging their unintended side effects. 41 7. Policy Conclusions Several important conclusions and policy lessons follow from our analysis: • We find a very different behavior in industrial and emerging countries regarding the use of macroprudential policy (at least in terms of RR). Since 2004, in particular, no industrial country has resorted to active RR policy, whereas close to half of developing countries have, of which 90 percent have used RR countercyclically. • RR seem to be an important component of a trio of policy instruments (together with short-term interest rates and foreign exchange market intervention) that developing countries have relied on for several decades now as they go through boom-bust cycles mainly induced by international capital flows. Despite all the buzz about systemic risk-driven macroprudential policy, we found no evidence of such use of RR in industrial countries. • The genesis for resorting to RR lies essentially on the behavior of the exchange rate over the business cycle in developing countries (with the currency depreciating in bad times and appreciating in good times). This complicates enormously the use of interest rates as a countercyclical instrument because doing so would appreciate (depreciate) even more the currency in good (bad) times. • The evidence suggests that RR are an effective instrument (i.e., a rise in RR increases the interest rate spread and reduces credit and GDP) that can well be used countercyclically when concerns about the effects of interest rates on the exchange rate become paramount. • It may well be the case that— and this is what we observe in countries such as Chile where policy institutions have improved steadily over time—developing countries may reach a point in time where it may no longer be necessary to use RR as a business cycle-driven macroprudential policy. Until then, however, RR seem a natural and effective instrument to complement monetary policy. • Even if and when a given developing country may reach a point where RR are no longer necessary as a part of the policy mix, RR may still be optimal to use as systemic risk-driven macroprudential policy. Our report, however, does not speak to the effectiveness of RR as a risk-reducing prudential instrument and therefore future research is called for in this regard. • While from a macroprudential point of view, the most common prudential instruments are essentially equivalent (for instance, RR, capital requirements, and taxes on credit), from a microprudential point of view they may elicit very different responses regarding banks’ risk taking behavior over the business cycle. • Depending on the nature and the drivers of the business cycle, conflicts may arise between the micro- and macro-prudential policy stances. 42 • The overall design of macroprudential policies should follow a careful analysis of the role that different financial frictions play in different environments since similar symptoms can reflect very different underlying forces. • More research is needed to embed banks’ risk taking incentives in macroeconomic models in order to properly assess and quantify the tensions that may arise between micro and macroprudential policies and to design a coherent prudential framework for the financial system. 43 References Bengui, Julien, Carlos Vegh, and Guillermo Vuletin, 2013, “Fixed costs in banking and the optimal mix of reserve requirement and interest rate policy,” unpublished manuscript. Blanchard Olivier, and Roberto Perotti, 2002, “An empirical characterization of the dynamic effects of changes in government spending and taxes on output,” The Quarterly Journal of Economics, Vol. 117, pp. 1329-1368. Bulow, Jeremy, John Geanakoplos, and Paul Klemperer, 1985, “Multimarket oligopoly: strategic substitutes and strategic complements,” Journal of Political Economy, Vol. 93, pp. 488-511. Calvo, Guillermo, Leonardo Leiderman, and Carmen M. Reinhart, 1993, “Capital inflows and real exchange rate appreciation in Latin America: The role of external factors,” IMF Staff Papers, Vol 40, pp. 108-151. Coibion Olivier, 2012, “Are the effects of monetary policy shocks big or small?” American Economic Journal: Macroeconomics, Vol. 4, pp 1-32. Cordella, Tito, and Poonam Gupta, 2014, “What makes a currency procyclical: and empirical investigation,” unpublished manuscript Cordella, Tito, and Samuel Pienknagura (2013), Macro prudential policies from a micro prudential angle, World Policy Research Working Paper No 6721 Cordella, Tito, and Samuel Pienknagura (2014), “Bank risk taking over the business cycle,” unpublished manuscript Coulibaly, Brahima, 2012, “Monetary policy in emerging market economies: What lessons from the global financial crisis,” International Finance Discussion Paper No. 1042, Board of Governors of the Federal Reserve System. de la Torre, Augusto & Ize, Alain, 2013. “The foundations of macroprudential regulation: a conceptual roadmap,” Policy Research Working Paper Series No 6575, The World Bank. Fernandez de Lis, Santiago, and Alicia Garcia-Herrero, 2013, “Dynamic Provisioning: Some Lessons from Existing Experiences,”Economia, Vol. 13, pp. 35-60 Dell’Ariccia, Giovanni, Luc Leaven and Robert Marquez, 2014, “Monetary Policy, Leverage, and Bank Risk Taking,” Journal of Economic Theory, Vol. 149, pp. 65-99. Diaz-Alejandro, Carlos, 1985, “Good-bye financial repression, hello financial crash,” Journal of Development Economics, Vol. 19, pp. 1-24. 44 Federico, Pablo, Carlos Vegh, and Guillermo Vuletin, 2013a. “Reserve requirement policy over the business cycle,” unpublished manuscript. Federico, Pablo, Carlos Vegh, and Guillermo Vuletin, 2013b. “Effects and role of macroprudential policy: Evidence from reserve requirements based on a narrative approach,” unpublished manuscript. Federico, Pablo, Carlos Vegh, and Guillermo Vuletin, 2013c. “Macroprudential policies in Brazil and Mexico from 1970 to 2010,” unpublished manuscript. Frankel, Jeffrey, 2011, “A solution to fiscal procyclicality: The structural budget institutions pioneered by Chile,” NBER Working Paper no. 16945 (2011). Frankel, Jeffrey, Carlos Vegh, and Guillermo Vuletin, 2013. “On graduation from fiscal policy procyclicality,” Journal of Development Economics, Vol. 100, pp. 32-47. Fernández de Lis, Santiago, and Alicia García-Herrero (2013), “Dynamic Provisioning: A Buffer Rather Than a Countercyclical Tool?” Economia, Vol. 13 No 2, pp. 35-60. Fischer, Stanley, 1998, “The IMF and the Asian crisis,” Forum Funds Lecture at UCLA (1998). Flood, Robert, and Olivier Jeanne, 2005. “An interest rate defense of a fixed exchange rate?” Journal of International Economics, Vol. 66, pp. 471-484. Furman, Jason and Joseph Stiglitz, 1998, “Economic crises: Evidence and insights from East Asia, Brookings Papers on Economic Activity, Vol. 2, pp. 1-135. Christian Glocker and Pascal Towbin, 2012, “The macroeconomic effects of reserve requirements, WIFO Working Paper No. 420. Hnatkovska, Viktoria, Amartya Lahiri, and Carlos Vegh, 2008, “Interest rates and the exchange rate: A non-monotonic tale,” NBER Working Paper No. 13925. IMF, 2012, “The Interaction of Monetary and Macroprudential Policies,” IMF Policy Paper Kashyap, Anil, and Jeremy C. Stein, 2000, “What do a million observations on banks say about the transmission of monetary policy?”American Economic Review, Vol. 90, pp. 407-28. Lahiri, Amartya, and Carlos Vegh, 2003, “Delaying the inevitable: Interest rate defense and balance of payments crises,” Journal of Political Economy, vol. 111, pp. 404-24. McKinnon, Ronald I, 1973, Money and capital in economic development, Washington, DC: Brookings Institution. Montiel, Peter, 2003, “Tight money in a post-crisis defense of the exchange rate: What have we learned?” The World Bank Research Observer, Vol. 18, pp. 1-23. 45 Oi, Walter, 1961, “The Desirability of Price Instability Under Perfect Competition,” Econometrica, 29: 58-64. Pereira da Silva, Luiz, and Ricardo Harris (2013), “Sailing through the Global Financial Storm,” in Dealing with the Challenges of Macro Financial Linkages in Emerging Markets, Canuto, Otaviano and Swati Ghosh eds., World Bank Study, Washington, DC: World Bank Riera-Crichton, Daniel, Carlos Vegh, and Guillermo Vuletin, 2012. “Tax multipliers: Pitfalls in measurement and identification,” NBER working paper No 18497. Romer, Christina, and David Romer, 2004, “A new measure of monetary shocks: Derivation and implications,” American Economic Review, Vol. 94, pp. 1055-1084. Romer Christina, and David Romer, 2010, “The macroeconomic effects of tax changes: Estimates based on a new measure of fiscal shocks.” American Economic Review 100, pp. 763- 801. Shaw, Edward, 1973, Financial deepening in economic development, New York: Oxford University Press. Stiglitz, Joseph, and Andrew Weiss, 1981, “Credit Rationing in Markets with Imperfect Information,” The American Economic Review, Vol. 71, pp. 393-410 Tovar Camilo, Mercedes Garcia-Escribano, and Mercedes Vera Martin, 2012, “Credit growth and the effectiveness of reserve requirements and other macroprudential instruments in Latin America,” IMF Working Paper No. 142. Vegh, Carlos, and Guillermo Vuletin, 2013. “Overcoming the fear of free falling: Monetary policy graduation in emerging markets,” in Douglas Evanoff, Cornelia Holthausen, George Kaufman, and Manfred Kremer (eds), The Role of Central Banks in Financial Stability: How Has It Changed?, pp. 105-131, Federal Reserve Bank of Chicago and European Central Bank. 46