Macroprudential Policy
What Instruments and How to Use them? Lessons From Country Experiences1

This paper provides the most comprehensive empirical study of the effectiveness of macroprudential instruments to date. Using data from 49 countries, the paper evaluates the effectiveness of macroprudential instruments in reducing systemic risk over time and across institutions and markets. The analysis suggests that many of the most frequently used instruments are effective in reducing pro-cyclicality and the effectiveness is sensitive to the type of shock facing the financial sector. Based on these findings, the paper identifies conditions under which macroprudential policy is most likely to be effective, as well as conditions under which it may have little impact.

Abstract

This paper provides the most comprehensive empirical study of the effectiveness of macroprudential instruments to date. Using data from 49 countries, the paper evaluates the effectiveness of macroprudential instruments in reducing systemic risk over time and across institutions and markets. The analysis suggests that many of the most frequently used instruments are effective in reducing pro-cyclicality and the effectiveness is sensitive to the type of shock facing the financial sector. Based on these findings, the paper identifies conditions under which macroprudential policy is most likely to be effective, as well as conditions under which it may have little impact.

I. Introduction

This paper is prepared at the request of the IMF Board. Macroprudential policy is quickly gaining traction in international circles as a useful tool to address system-wide risks in the financial sector.2 Yet, the analytical and operational underpinnings of a macroprudential framework are not fully understood and the effectiveness of the instruments is uncertain. In April 2011, the Board initiated a discussion of these issues in the context of the paper “Macroprudential Policy: An Organizing Framework” (SM/11/54). In concluding, the Board asked for further work on several fronts.3 This paper responds to the specific request for a review of country experiences to better understand the design and calibration of macroprudential instruments, their interaction with other policies, and their effectiveness.

While macroprudential policy is widely seen as a useful policy response to changes in the global financial environment, views on the contours of macroprudential policy can vary substantially among policymakers. The IMF—in conjunction with the Bank for International Settlements and the Financial Stability Board—has characterized macroprudential policy with reference to three defining elements:4

  • Its objective: to limit the risk of widespread disruptions to the provision of financial services and thereby minimize the impact of such disruptions on the economy as a whole. Systemic risk is largely driven by fluctuations in economic and financial cycles over time, and the degree of interconnectedness of financial institutions and markets.

  • Its analytical scope: the focus is on the financial system as a whole (including the interactions between the financial and real sectors) as opposed to individual components.

  • Its instruments and associated governance: it primarily uses prudential tools that have been designed and calibrated to target systemic risk. Any non-prudential tools that are part of the framework need to be specifically designated to target systemic risk through their governance arrangements.

Against this organizing framework, the objective of the paper is to identify conditions under which macroprudential policy is most effective. The assessment uses data provided by the 2010 IMF Survey on financial stability and macroprudential policy, as well as an internal survey of desk economists.5 Relative to previous studies, this approach has the advantage of examining a much broader range of instruments,6 risks, and countries, taking greater account of the implications of cyclical disturbances and interconnectedness. The goal is to help policymakers make more informed decisions about macroprudential policy and to guide the Fund’s policy advice and technical assistance in this area.

The paper is structured as follows. Section II reviews country experiences with macroprudential policy, focusing on the objectives, types of instruments and how they have been chosen and applied. Section III presents the empirical analysis based on case studies and panel regressions. Section IV draws common lessons and policy messages, noting the conditions under which the instruments appear to have been most effective. Section V concludes with next steps for further research and analysis.

II. Country Experiences with Macroprudential Instruments

A. What Instruments Are Used?

Country authorities have used a variety of policy tools to address systemic risks in the financial sector. The toolkit contains mostly prudential instruments, but also a few instruments typically considered to belong to other public policies, including fiscal, monetary, foreign exchange and even administrative measures. The IMF survey identified 10 instruments that have been most frequently applied to achieve macroprudential objectives. There are three types of measures:

  • Credit-related, i.e., caps on the loan-to-value (LTV) ratio, caps on the debt-to-income (DTI) ratio, caps on foreign currency lending and ceilings on credit or credit growth;

  • Liquidity-related, i.e., limits on net open currency positions/currency mismatch (NOP), limits on maturity mismatch and reserve requirements;7

  • Capital-related, i.e., countercyclical/time-varying capital requirements, time-varying/dynamic provisioning, and restrictions on profit distribution.

There is usually a clearly stated policy objective when the instruments are applied. Specifically, the instruments have been used to mitigate four broad categories of systemic risk (Figure 2):8

  • Risks generated by strong credit growth and credit-driven asset price inflation;

  • Risks arising from excessive leverage and the consequent deleveraging;

  • Systemic liquidity risk; and

  • Risks related to large and volatile capital flows, including foreign currency lending.

Figure 2.
Figure 2.

Objectives of Macroprudential Policy Instruments

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IMF Financial Stability and Macroprudential Policy Survey, 2010.

The recent financial crisis has prompted an increasing number of countries to use the instruments, and with greater frequency. According to the IMF survey, two-thirds of the respondents have used various instruments for macroprudential objectives since 2008. Emerging market economies have used the instruments more extensively than advanced economies, both before and after the recent financial crisis. Elements of a macroprudential framework existed in some emerging market economies in the past, when they started to use some of the instruments to address systemic risk following their own financial crises during the 1990s. For these countries, the instruments are part of a broader “macro-financial” stability framework that also includes the exchange rate and capital account management.9 The recent crisis has also led to an increase in the number of advanced countries that deploy the instruments within a more formal macroprudential framework. The work of the European Systemic Risk Board is an example (Box 1).

Macroprudential Instruments in the European Union10

Work on selecting and applying macroprudential instruments is a priority in the European Union (EU), both at a national and at a Union level. The European Systemic Risk Board (ESRB) was established as of January 1, 2011, in order to provide warnings of macroprudential risks and to foster the application of macroprudential instruments.

Macroprudential instruments have a particular relevance in the EU context, given the constraints on macroeconomic and microprudential policies and their coordination, including the absence of national monetary policies and policies to harmonize capital standards. The ESRB has an additional role to foster “reciprocity” through its “comply or explain” powers amongst the national authorities, so that all banks conducting a particular activity in a country will be subject to the same macroprudential instrument irrespective of the bank’s home country.

The European Commission has been focusing on countercyclical capital as the main macroprudential instrument. Other agencies, as well as some national authorities, propose casting the net much wider, to take account of regional, national, sub-national, or sectoral conditions. For instance, with real estate lending having been central to past financial crises, there is likely to be a focus on instruments such as the loan-to-value ratio.

B. Why Use Macroprudential Policy and What Affects the Choice of Instruments?

Macroprudential policy has several advantages compared with other public policies to address systemic risk in the financial sector. In their survey responses, country authorities indicate that macroprudential instruments are less blunt than monetary tools, and are more flexible (with smaller implementation lags) than most fiscal tools. Many instruments (e.g., caps on the LTV, DTI, foreign currency lending, and capital risk weights) can be tailored to risks of specific sectors or loan portfolios without causing a generalized reduction of economic activity, thus limiting the cost of policy intervention. Some countries have imposed caps on foreign currency lending, for example, because these target excessive lending in foreign currency directly in a way that no other policies can. These instruments are especially useful when a tightening of monetary policy is not desirable (e.g., when inflation is below target).

Country authorities indicate that they choose instruments that are simple, effective, and easy to implement with minimal market distortions. They consider it necessary that the choice of macroprudential instruments be consistent with other public policy objectives (fiscal, monetary, and prudential). They also believe it important to choose macroprudential instruments that minimize regulatory arbitrage, particularly in advanced economies with large nonbank financial sectors and complex and highly interconnected financial systems.

A number of factors seem to influence the choice of instruments. The stage of economic and financial development is one such factor (Figure 3). In general, emerging market economies have used macroprudential instruments more extensively than advanced economies. This may reflect a greater need to address market failures where financial markets are less developed and banks usually dominate relatively small financial sectors. Emerging market economies are more concerned about systemic liquidity risk and tend to use liquidity-related measures more often. Advanced economies tend to favor credit-related measures, although more of them are beginning to use liquidity-related measures after the recent crisis.11

Figure 3.
Figure 3.

Use of Macroprudential Policy Instruments

(% of countries in each group using each type of instruments)

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

1/ The ratio of credit/financial claims to GDP. Countries with the ratio at or above the medium are classified as “large,” otherwise “small.”2/ The ratio of net capital inflow to GDP. Countries with the ratio at or above the medium are classified as “large,” otherwise “small.”Sources: IMF Financial Stability and Macroprudential Policy Survey, 2010.

The exchange rate regime appears to play a role in the choice of instruments. Countries with fixed or managed exchange rates tend to use macroprudential instruments more since the exchange rate arrangement limits the room for interest rate policy. In these countries, credit growth tends to be associated with capital inflows as the implicit guarantee of the fixed exchange rate provides an incentive for financial institutions to expand credit through external funding.12 Credit-related measures (e.g., caps on the LTV and ceilings on credit growth) are often used by these countries to manage credit growth when the use of interest rates is constrained. They also tend to use liquidity-related measures (e.g., limits on NOP) to manage external funding risks.

The type of shocks is another factor that may influence the choice of instruments. Capital inflows are considered by many emerging market economies to be a shock with a large impact on the financial sector, given the small size of their domestic economy and their degree of openness. Some Eastern European countries have used credit-related measures (e.g., caps on foreign currency lending) to address excessive credit growth resulting from capital inflows. In Latin America, several countries (e.g., Argentina, Brazil, Colombia, Peru, and Uruguay) have also used liquidity-related measures (e.g., limits on NOP) to limit the impact of capital inflows. In the Middle East, some oil exporters with fixed exchange rates have also used credit-related measures to deal with the impact of volatile oil revenue on credit growth. Unlike other policy tools aimed at the volume or composition of the flows (e.g., taxes, minimum holding periods, etc.), macroprudential instruments are more directly aimed at the negative consequences of inflows, i.e., excessive leverage, credit growth and exchange rate induced credit risks that are systemic.

C. How Are Instruments Applied?13

Country experiences show that a combination of several instruments is often used to address the same risk. Caps on the LTV and DTI, for instance, are frequently applied together by country authorities to curb rapid credit growth in the real estate sector. Sometimes a range of measures are implemented (Figure 4). On the other hand, using a single instrument to address systemic risk is rare.14 The rationale for using multiple instruments seems simple—to provide a greater assurance of effectiveness by tackling a risk from various angles. While this may be true, there may be a higher regulatory and administrative burden of enforcing multiple instruments.

Figure 4.
Figure 4.

How Instruments Are Used

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Sources: IMF Financial Stability and Macroprudential Policy Survey, 2010.

Many instruments, particularly credit-related, are calibrated to target specific risks. Macroprudential instruments are generally more targeted than monetary and fiscal policy tools, and they are frequently further differentiated for specific types of transactions. Caps on the LTV and DTI, for example, have been applied according to the loan size, the location and the value of the property (Hong Kong SAR and Korea). Reserve requirements used for macroprudential purposes have been differentiated by currency, types of liabilities, and applied within a band or on a marginal basis, or if credit growth exceeds the official limit (Argentina, Chile, China, Indonesia, Peru, Russia, Serbia, and Turkey). Sometimes social and other developmental aspects are taken into account when the instruments are calibrated (Canada). Many countries apparently find it useful to take full advantage of the targeted nature of macroprudential instruments, but others also apply the instruments broadly with no further differentiation.

Making countercyclical adjustments of macroprudential instruments is a common practice. Instruments aimed at credit growth, such as caps on the LTV, the DTI and reserve requirements, are adjusted most frequently. The adjustments are usually made to give the instruments a progressively larger countercyclical impact, but in some cases they also reflect the need to proceed cautiously on a trial and error basis. Capital-related measures, such as countercyclical capital requirements and dynamic provisioning, are designed to work through the cycle by providing a buffer, but some countries have adjusted them at different phases of the cycle to give them a more potent countercyclical impact.15

The design and calibration of the instruments are usually based on discretion and judgment, as opposed to rules. The use of rules-based instruments has the advantage of less regulatory uncertainty, preventing political economy pressures and overcoming policy inertia when systemic risk is building up.16 However, most countries that participated in the IMF survey have used judgment almost entirely when designing and calibrating the instruments. The implementation of the instruments is a learning-by-doing process, in which judgment on how to calibrate an instrument is often formed by trial and error, depending on the type of shock the system is facing. A few exceptions include dynamic provisioning as used in Spain and several Latin American countries, where the amount of provisioning is based on a formula and varies with the economic cycle.

Macroprudential instruments are sometimes applied in conjunction with other macroeconomic policies. Some Asian and Latin American country authorities have used macroprudential instruments such as caps on the LTV with other policies, for example, monetary and fiscal policies.17 Some Eastern European countries have kept fiscal policy loose, but tightened monetary policy and attempted to contain banks’ foreign currency lending through various macroprudential measures. The combined use of policy tools typically occurs when the credit cycle coincides with the business cycle and there is a generalized risk of excessive credit growth and economic overheating. In such cases, macroprudential instruments are implemented as part of a larger policy action to curb excess demand and the build-up of systemic risk, so they play a complementary role to macroeconomic policies.18 Figure 5 summarizes the intensity of use of the instruments.

Figure 5.
Figure 5.

Intensity of Use

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Sources: IMF Financial Stability and Macroprudential Policy Survey, 2010.

III. Effectiveness of Macroprudential Instruments

Macroprudential instruments may be effectively applied to address specific risks if used appropriately. According to the IMF survey, most country authorities who have used macroprudential instruments believe that they are effective. To assess the effectiveness of macroprudential instruments more thoroughly, this paper uses three different approaches. The first is a case study, involving an examination of the use of instruments in a small number of countries to see if they have achieved the intended objectives. The second is a simple approach, involving an examination of the performance of the target (risk) variables before and after an instrument is introduced. The third is a more sophisticated approach, which uses panel regression to assess the effect of macroprudential instruments on various target risk variables by comparing the introduction of an instrument with a “counterfactual” scenario where no macroprudential instrument is implemented.

The usual caveats, of course, apply to the evaluation. First, data availability and quality present challenges. Firm level data are preferable since many of the macroprudential instruments are aimed at the balance sheet of financial institutions, but these are not readily available or consistent over time or across countries. Moreover, the number of countries that have used macroprudential instruments in a systematic way is small since macroprudential policy frameworks have been put in place only recently, limiting the degree of confidence in any statistical analysis. In addition, establishing causality is not straightforward, or even feasible in some cases, with a selection bias that favors high risk countries where policies are implemented in reaction to adverse economic or market developments. The empirical analysis also does not take into account issues such as costs and distortions, important factors to consider when using the instruments. These caveats notwithstanding, the evaluation still provides valuable insights into the effectiveness of macroprudential instruments.

A. The Case Study

Experiences of a few countries suggest some success in using the instruments to achieve their intended objectives. The case study covers a small but diverse group of countries, including China, Colombia, Korea, New Zealand, Spain, the United States and some Eastern European countries. While small, the sample seems representative. Some countries use the instruments singly while others in combination (and in coordination with other policies); instruments are both broad-based and targeted; some keep the instruments fixed while others make adjustment (both rules-based and discretionary). Their experience suggests that, to various degrees, the instruments may be considered effective in their respective country-specific circumstances, regardless of the size of their financial sector or exchange rate regime. Appendix II presents the case studies, which are summarized briefly below.

  • In China, the authorities managed to lower credit growth and housing price inflation by taking a series of steps in 2010 that also included fiscal and monetary measures.

  • In Colombia, the authorities took measures in 1999 to limit banks’ exposure to default risk. The measures seem to have been effective. Non-performing loans declined and remained low while credit to the private sector recovered after an initial reduction.

  • In Eastern Europe, the authorities adopted several measures to curb bank lending in foreign currency. The instruments appear to have been effective in slowing credit growth and building capital and liquidity buffers, although they were circumvented partly as lending activity migrated to nonbanks (leasing companies) and to direct cross-border lending by parent banks.

  • In Spain, the authorities introduced dynamic provisioning as a macroprudential tool in 2000. The instrument appears to have been effective in helping to cover rising credit losses during the global financial crisis, but the coverage was less than full because of the severity of the actual losses.

  • In Korea, the authorities adopted measures after the financial crisis to deal with the build-up of vulnerabilities associated with capital flows. They appear effective in curbing banks’ short-term external borrowing, which remained some 30 percent below its pre-crisis levels as of 2010.

  • In New Zealand, the authorities introduced two liquidity mismatch ratios and a core funding ratio in 2010 to limit banks’ liquidity risk. The ratios had an effect even before they were formally implemented—banks began to lengthen their wholesale funding structure after the ratios’ announcement.

  • In the United States, the authorities adopted a minimum leverage ratio for banks in 1991. The requirement was not adjusted over time in response to changing circumstances, but a key weakness was the fact that it did not apply to investment banks after 2004. As result of the divergence in regulatory requirements, leverage rose noticeably at investment banks but remained lower at commercial banks.

B. The Simple Approach

Some targeted risk variables show a change of course after the instruments are introduced. An examination of the performance of the target risk variables during the periods before and after the implementation of an instrument indicates that a number of them may have had the intended effect. Some instruments, e.g., caps on the LTV, caps on the DTI, dynamic provisioning, and reserve requirements, seem to have an impact on credit growth (Figures 6), but the effect of other instruments is less obvious.19 Specifically,

  • Caps on the LTV: credit growth and asset price inflation decline after its implementation in more than half of the countries in the sample.

  • Caps on the DTI: credit growth decline but asset price inflation does not.

  • Dynamic provisioning: credit growth and asset price inflation, and to a lesser extent, leverage growth, decline.

  • Reserve requirements: both credit growth and asset price inflation decline.

Figure 6.
Figure 6.

Change in Credit Growth After the Introduction of Instruments

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Notes:1/ Average of sample countries’ y/y growth in credit (detrended).2/ t denotes the time of the introduction of instruments.3/ For details, see charts in Appendix III.Source: International Financial Statistics.

Macroprudential instruments seem to have been effective in reducing the correlation between credit and GDP growth. In countries that have introduced caps on the LTV, DTI and reserve requirements, the correlation is positive but much smaller than in countries without them, as shown by the flattening of the curve in Figure 7. In countries that have introduced ceilings on credit growth or dynamic provisioning, the correlation between credit growth and GDP growth becomes negative as shown by an inverted curve. The difference in the correlations is also statistically significant, except in the case of caps on foreign currency lending and restrictions on profit distribution. A more sophisticated analysis is described below to try to demonstrate causality and to disentangle the effects of other macroeconomic policies.20

Figure 7.
Figure 7.

Credit Growth and GDP Growth

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IFS.

C. The Panel Regression

A panel regression analysis suggests that macroprudential instruments may have an impact on four measures of systemic risk—credit growth, systemic liquidity, leverage, and capital flows.21 Specifically, eight instruments22 are estimated to see if they limit the procyclicality of credit and leverage—their tendency to amplify the business cycle. Procyclicality is captured in this case by the respective correlation of growth in credit and leverage with GDP growth. This specification has the advantage of showing the effect of the instruments in both the expansionary and recessionary phases of the cycle without “timing” the cycle. In addition, the effects of the other two instruments23 on common exposure are estimated, using proxies for risks related to liquidity and capital flows, although the scope is limited by data availability. Dummy variables for factors such as the degree of economic development, the type of exchange regimes and the size of the financial sector are used to see if the instruments are effective across countries. The regressions use data from 49 countries during a 10-year period from 2000 to 2010 collected in the IMF survey.

The specification of the panel regressions addresses several challenging issues, including:

  • How to disentangle the effect of macroprudential instruments from that of other policies. For monetary policy, an interest rate variable is introduced, and for fiscal policy, GDP growth is used as a proxy. Using fiscal deficit has the disadvantage of introducing multicollinearity given its high correlation with GDP growth, and there seems no direct linkage between fiscal policy and procyclicality of credit or leverage. Any indirect linkage would be captured by interest rates and GDP growth.24

  • How to infer the general effect of macroprudential instruments in the context of country-specific characteristics. This is addressed by introducing dummy variables to control for the type of exchange rate regime, the size of the financial sector and the degree of economic development. The panel regressions’ fixed effect takes into account other unobserved country-specific characteristics.

  • How to avoid estimation biases to ensure a correct quantification of the effect of macroprudential instruments.25 This is addressed by using the System Generalized Method of Moments,26 widely used to deal with panel data with endogenous explanatory variables.

Results of the panel regressions suggest that the majority of the 10 instruments may be effective. The empirical analysis finds no evidence to suggest that the degree of economic development, the type of exchange rate regimes or the size of the financial sector affects the effectiveness of the instruments—the estimated coefficients of their dummy variables are all statistically insignificant—even though these factors may influence their choice. The results also show that the instruments remain effective after controlling for macroeconomic policies. As indicated by an impulse response analysis of an open economy DSGE model, a combination of policies may have lower welfare costs than monetary or macroprudential policy used alone (Box 2). In addition, instruments that are rules-based have a larger effect, although there is not enough evidence to indicate whether individual or multiple instruments are more effective due to the lack of granular data. Results of the regressions are summarized as follows:

  • On credit growth (yoy change in inflation-adjusted claims on the private sector), the coefficients of five of the 10 instrument dummy variables (caps on the LTV, DTI, ceilings on credit growth, reserve requirements and time-varying/dynamic provisioning) are statistically significant (Table 1).27 This indicates that these instruments may reduce the correlation between credit growth and GDP growth. Caps on the LTV, for example, reduce the procyclicality of credit growth by 80 percent.28 This is in line with findings of previous studies that associate higher LTV ratios with higher house price and credit growth over time.29 The coefficient of the dummy variable for a subgroup of countries that have adjusted the LTV caps over time is also significant.

  • On systemic liquidity, credit expansion funded from sources other than deposits (credit/deposit) is used as a proxy for wholesale funding in the estimation of the effectiveness of limits on maturity mismatch. The estimation is intended to see if this instrument limits wholesale funding, considered a source of systemic risk with a cross-sectional dimension. The coefficient of the dummy variable for limits on maturity mismatch is statistically significant, and the credit/deposit ratio is 5 percent lower in countries with the instrument than in countries without it.

  • On leverage (assets/equity), the coefficients of six of the 10 instrument dummy variables (caps on the DTI, ceilings on credit growth, reserve requirements, caps on foreign currency lending, countercyclical/time-varying capital requirements30 and time-varying/dynamic provisioning) are statistically significant (Table 2). This indicates that, while capital-related measures are expected to reduce the procyclicality of leverage, other instruments aimed at limiting credit growth may also have an impact on leverage growth. Dynamic provisioning appears to reduce the procyclicality of both credit growth and leverage. The effect of other capital-related measures is not obvious probably because the number of observations available is limited as only a few countries have implemented them in the last two years.

  • On capital flows and currency fluctuation, external indebtedness (foreign liabilities/foreign assets) is used as a proxy for common exposure to risks associated with them. The only dummy variable that has a statistically significant coefficient is limits on NOP. The results suggest that for every dollar of foreign assets held, the foreign liabilities of countries with this instrument are 15 percent lower than those without it (Table 3).

Table 1.

Effectiveness of Macroprudential Instruments in Reducing the Pro-cyclicality of Credit

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variable is credit growth, the log change in the real level of credit. Credit is measured as claims on private sector from both bank and non-bank financial institutions (source: IFS). The interest rate is the nominal long-term interest rate on prime lending, from the IMF’s International Financial Statistics. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

Non-Significant Results w hen Interest Rate included.

The coefficient corresponds to the interaction term between GDP growth and a dummy for the respective macroprudential instrument.

Source: IMF staff estimates.
Table 2.

Effectiveness of Macroprudential Instruments in Reducing the Pro-cyclicality of Leverage

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variable is leverage growth, the log change in the level of leverage. Leverage is measured as assets over capital (source: IMF FSIs). The interest rate is the nominal long-term interest rate on prime lending, from the IMF’s International Financial Statistics. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

The coefficient corresponds to the interaction term between GDP growth and a dummy for the respective macroprudential instrument.

Source: IMF staff estimates.
Table 3.

Effectiveness of Macroprudential Instruments in Reducing Cross-Sectional Risks

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variables are the ratio of financial system liabilities with foreign residents to claims on foreign residents (1) and the ratio of banking institutions claims to deposits (2), obtained from the IMF’s International Financial Statistics. The interest rate is the nominal long-term interest rate on prime lending, also from IFS. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

The coefficient corresponds to a dummy variable with a value of 1 for countries with limits on net open positions in foreign currency, and zero otherwise.

The coefficient corresponds to a dummy variable with a value of 1 for countries with limits on maturity mismatches, and zero otherwise.

Source: IMF’s staff estimates.

The regression results are independently confirmed by other studies. A separate study that focuses more on the structural determinants of credit growth corroborates the initial findings of the panel analysis. This study uses a different model and assumption on endogeneity, and the coefficients of caps on the DTI, caps on foreign currency lending, reserve requirements and time-varying/dynamic provisioning have a negative sign on credit to GDP and are statistically significant.31

This paper’s finding that the effectiveness of the instruments does not depend on the type of exchange rate regime is also independently confirmed by a structural model used in IMF (2011h), which shows that the impact of macroprudential instruments is virtually identical in economies with either fixed or floating exchange rates. The regression results need to be interpreted with caution. Statistically, the coefficients of the dummy variables for the instruments are averages of country performances. Their magnitude is affected by the number of countries in the sample that have used the instruments as well as the effectiveness in individual countries, and their statistical significance is not an indication that the instruments are equally effective in all countries. Country-specific circumstances, such as the quality of supervision, the phase of the credit cycle in which the instruments are implemented, the extent to which circumvention and arbitrage are possible, the ability of the authorities to take coordinated policy actions to limit circumvention and their responsiveness to changed conditions are among factors that determine whether an instrument is effective when applied in a particular country.

While the panel regression yields promising results, more work is needed to confirm its findings. The use of macroprudential instruments is still relatively new. The short experience with macroprudential policy limits the number of observations available for a more comprehensive evaluation of its effectiveness. Further research with longer time series and better quality data is therefore necessary to corroborate the initial assessment and to evaluate an instrument’s effectiveness in country-specific contexts. Factors such as the costs involved in using macroprudential instruments, the degree of calibration, and the potential for regulatory and cross-border arbitrage, which can easily circumscribe the effectiveness of macroprudential policy, should be taken into account in future analysis.

Monetary and Macroprudential Policy: Are They Mutually Reinforcing? 1/

Should macroprudential measures be used in conjunction with monetary policy to mitigate risks associated with large capital inflows? To address this question, an open-economy, New Keynesian DSGE model is used to assess whether a combination of the two policies is superior to stand-alone policies.

In the model, firms can finance their investment through retained earnings or borrowing from domestic or foreign sources. Macroprudential policy is assumed to impose a higher cost of borrowing for firms, defined as an additional “regulation premium” to the cost of borrowing. Monetary policy is assumed to follow a Taylor rule, with the central bank reacting to changes in inflation and output gaps. An initial shock, modeled as a decline in investors’ perception of risk, triggers capital inflows, leading to a decline in financing costs; firms borrow and invest more. Eventually, higher leverage triggers an increase in risk premium, and financial conditions normalize. But both monetary and macroprudential policies have a nontrivial role in mitigating the impact of the shock.

uA01fig01

Dynamic Responses to a Positive Financial Shock (percent deviations from steady state)

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IMF staff analysis.

The simulations suggest that macroprudential measures could be a useful complement to monetary policy in stabilizing the economy after the initial shock. When policymakers adopt macroprudential measures that directly counteract the increase in leverage and the easing of underwriting standards, the responses of domestic and foreign debt to the shock become more muted. Output and inflation therefore respond less, and the welfare loss, computed as the sum of inflation and output volatilities in percent of steady state consumption, decreases by almost half (1.3) compared with the simple Taylor rule (2.5), where only monetary policy is implemented. In the scenario where macroprudential measures alone are implemented and the policy interest rate is kept unchanged, output and inflation become more volatile, and the welfare loss is large (31.5).

In conclusion, the combination of monetary and macroprudential policies are superior to stand-alone policies.

1/ See Unsal (2011).

IV. Lessons and Policy Messages

A number of instruments may be effective in addressing systemic risks in the financial sector. The effectiveness does not seem to depend on the stage of economic development or type of exchange rate regime. Emerging market economies with fixed or managed exchange rates, where room for interest rate policy is limited, facing large capital inflows or having thin financial markets and a bank dominated financial system tend to use macroprudential instruments more extensively, but the instruments seem equally effective when used by countries with flexible exchange rate regimes and by advanced economies. However, there are costs involved in using macroprudential instruments, as is the case with regulation more generally, and the benefits of macroprudential policy should be weighed against these costs. Moreover, calibrating the instruments may be difficult, which could lower growth unnecessarily or generate unintended distortions if not done appropriately. These issues are not addressed in the paper but are important considerations to take into account when using macroprudential instruments.

Underpinning the assessment of effectiveness is the assumption of a sound regulatory framework and high quality supervision. These are the foundation for the effective application of macroprudential instruments.32 In addition, institutional arrangements for macroprudential policy need to ensure a policymaker’s ability and willingness to act—including clear mandates; control over instruments that are commensurate with those mandates; arrangements that safeguard operational independence; and provisions to ensure accountability, supported by transparency and clear communication of decisions and decision-making processes.33

While care is needed to avoid one-size-fits-all approaches, there are common lessons on what instruments should be used to address specific risks that are considered systemic:

  • To address systemic risks generated by credit growth or asset price inflation, credit-related instruments may be useful. Of these, LTV and DTI caps can be kept in place, adjusted counter cyclically or targeted at specific sources of risk. They may be supplemented by reserve requirements or capital-related instruments, such as dynamic provisioning, should the credit boom become more generalized; these in turn can be targeted by currency if foreign currency lending proves to be the source of risk.

  • To address systemic liquidity risk, liquidity-related instruments such as limits on liquidity mismatch may be used, or limits on the net foreign currency position if the liquidity risk stems from foreign currency funding. A core (or stable) funding ratio, or a levy on non-core liabilities, which are not examined by this paper, could also be good candidates if wholesale funding is a significant funding source. The ratio or levy can be kept in place to prevent the buildup of systemic liquidity risk, or adjusted in response to a sudden liquidity shock.

  • To address risks arising from excessive leverage, capital-related instruments may be a good choice. These measures provide a buffer that can be made countercyclical through adjustments in the capital requirement, the risk weights of assets or the provisioning requirement, and can thus help curtail excessive growth in leverage. If leverage growth stems from banks’ drive to expand credit, capital-related measures can be supplemented by credit-related instruments to go to the source of the risk.

  • If the above mentioned risks arise due to capital flows, all three types of instruments can be used. Liquidity-related instruments, like limits on net open positions in foreign currency, are shown to be effective in limiting the financial sector’s dependence on foreign sources of funding. These instruments can be supported by credit-related instruments if excessive credit growth is what drives banks to borrow abroad. In this context, capital-related instruments may also be useful by limiting credit growth and providing a buffer.

Several considerations are relevant for the successful design and calibration of instruments. Countries have tailored the design and calibration of the instruments to their specific circumstances, taking into account the type and source of risk, the ability of the financial system to circumvent the measure, or bear the cost of additional regulation, the quality of supervision and enforcement, and the governance and accountability arrangements regarding macroprudential policy.34 The following five considerations are important (Table 4):

  • Single versus multiple

  • Broad-based versus targeted35

  • Fixed versus time-varying

  • Rules versus discretion

  • Coordination with other policies

Table 4.

Use of Macroprudential Instruments Some Considerations

article image
Source: IMF Staff Analysis.

The use of multiple instruments has the advantage of tackling the same risk from various angles. A combination of instruments also reduces the scope for circumvention and provides a greater assurance of effectiveness by addressing different sources of the risk. Caps on the LTV and the DTI, for example, complement each other in dampening the cyclicality of collateralized lending, with the LTV addressing the wealth aspect, and the DTI the income aspect, of the same risk.36 In general, when credit-related instruments are used to address risks generated by excessive credit growth, it may also be useful to limit funding risks with liquidity-related instruments and to provide a cushion by using capital-related instruments. Nevertheless, the use of multiple instruments may impose a higher cost on banks and are harder to calibrate and communicate, so it is important to choose instruments that minimize the cost and plan the implementation carefully to avoid an unnecessary burden on the financial sector.

Some instruments can be used to target specific risks, although the targeted approach has its limits. Macroprudential policy is already more targeted than monetary policy, and the ability of macroprudential instruments to target specific types of activities is another advantage that makes them more precise and potentially more effective. A lower LTV cap on more expensive houses helps limit the risk to banks since such exposure tends to be riskier while a higher LTV cap on less expensive houses may be desirable from a social perspective as well. However, the targeted approach requires more granular data, has a higher administrative cost and may be more susceptible to circumvention. Excessive targeting may also result in micromanagement, which would increase the cost of policy actions. The additional benefit of targeting should be weighed against its cost.

It is useful to adjust macroprudential instruments at different phases of the cycle to smooth out cyclicality. Some macroprudential instruments counter the cyclicality in the financial system as an “automatic stabilizer.” Dynamic provisioning and the capital conservation buffer under Basel III fall into this category, whose buildup during the upturn and depletion during the downturn help limit the severity of the cycle. However, other instruments, such as caps on the DTI, ceilings on credit growth and reserve requirements, may need to be adjusted during different phases of the cycle to minimize cyclicality. In addition, adjustments in the LTV cap and capital requirements can make them more potent in smoothing out the cycle, as indicated in Section III. While necessary, the adjustments should be based on sound and transparent principles and ad hoc and frequent changes that are disruptive to financial activities should be avoided.

Instruments that vary through the cycle based on rules have clear advantages and should be used to the extent possible. Dynamic provisioning and the capital conservation buffer are two examples of such instruments. The use of rules-based instruments helps overcome policy inertia and provides greater predictability in the regulatory environment. However, these two instruments may be rare exceptions, most other instruments, such as caps on the LTV, DTI, ceilings on credit growth and reserve requirements, may need to be adjusted at the discretion of the policymaker because designing rules for their adjustment may be difficult or even impossible, especially when it is necessary to use multiple instruments in combination. When discretion is necessary, it is useful to make the adjustment on a trial and error basis in a learning-by-doing process. Still, even when discretionary action is necessary, macroprudential policymakers should base their decisions on formal methods of analysis, and explain the rationale behind their actions publicly to enhance policy transparency and effectiveness.

The need for the discretionary use of the instruments calls for a framework to guide the conduct of macroprudential policy. This framework should include a mechanism to identify and monitor systemic risk, procedures for using macroprudential instruments, and careful choice of specific objectives macroprudential policy actions are to achieve. The criteria for the choice of instruments and methodology for the evaluation of their effectiveness should also be important elements in the framework. In addition, since many fiscal and monetary tools may be used to address systemic risk, a clear communication strategy and a set of principles and rules regarding the use of other public policy tools for macroprudential objectives are essential for transparency and the credibility of the macroprudential authority.

Well coordinated policy actions are a necessary condition for a successful response to systemic risk. The combined use of macroprudential instruments with monetary and fiscal policy tools in addressing systemic risk tends to be more effective when financial sector risks intertwine with those in other sectors or the financial cycle coincides with the business cycle. In general, macroeconomic policies should always be the primary tool to use when the source of systemic risk is domestic demand imbalances. In particular, macroprudential policy should be used only as a complement to monetary policy, which is more blunt and potent in addressing excess demand. On the other hand, macroprudential policy is better suited to target specific sectors, and should be used primarily to increase the resilience of the financial system. In any event, mechanisms should be established to address coordination challenges and limit any potential policy conflicts.

V. Next Steps

This paper has examined the use of macroprudential instruments to mitigate systemic risk in the financial system. The analysis focuses on the factors affecting the choice of the instruments, the circumstances in which the instruments are used, and the effectiveness of the instruments in achieving their intended objectives by drawing on the experience of a sample of 49 countries that have actively applied macroprudential instruments in the past 10 years. Several common lessons and policy messages, on conditions for macroprudential policy to be effective and situations to avoid, are derived from country experiences and econometric analysis. The broad guidelines set out in this paper should contribute to the international debate on how to make macroprudential policy operational and help guide the Fund’s policy advice in surveillance and technical assistance.

The findings are preliminary and more work is needed in several areas. The paper has assessed mostly the time dimension of systemic risk, and largely with experiences from emerging market economies. The analysis of the cross-sectional dimension of systemic risk has been more limited, and data availability has been the main constraining factor. In analyzing the interconnectedness of global systemically important institutions, more granular data would be required. Filling the data gaps would also help to develop mechanisms to identify and monitor systemic risk, which is essential to make macroprudential policy operational.

A deeper understanding of design and calibration issues and how they shape effectiveness is needed. The paper has shown that some approaches have advantages over others, but whether instruments would be more effectively used strictly as a form of insurance against future crisis or as a tool to correct imbalances is unclear. Another issue not addressed in this paper but may warrant further research is whether price-based or quantity-based instruments are more effective. Effectiveness may also vary with the degree of complexity (e.g., as instruments become more targeted), or if the instrument is used to pursue more than one objective.

The cost of implementing macroprudential instruments is another issue that needs further exploration. Although these issues are beyond the scope of this paper, it will be important to consider costs related to the regulatory burden, distortions, or other unintended consequences when making macroprudential policy operational. Most notably, macroprudential instruments may cause a migration of systemic risk to other parts of the financial system, and care is needed to mitigate such “leakages.”

The relationship between macroprudential policy and microprudential regulation also needs to be further clarified. Many of the macroprudential instruments cited in this paper are traditional prudential regulation tools. These instruments are assumed to be “readily” available for use as macroprudential instruments. However, it is important to clarify when the prudential tools begin to serve macroprudential purposes so that the implementation of macroprudential policy can be well coordinated with microprudential objectives.

Appendix I. Macroprudential or Capital Flow Measures?

Many countries have recently undertaken measures which can be considered both macroprudential—in the sense that they seek to respond to rising systemic risk in the financial system—and capital flow management measures (CFMs)—in the sense that they are designed to affect capital inflows and hence the exchange rate. This box describes recent examples of such measures in Brazil, Korea, and Turkey. These measures are further described in IMF (2011f), Recent Experiences in Managing Capital Inflows—Cross-Cutting Themes and Possible Policy Framework. A common theme from these cases is that concerns of preserving financial stability and macroeconomic stability (exchange rate appreciation, overheating, etc.) are often intertwined.

Brazil. Managing large capital inflows has been one of the main policy issues in Brazil since the global financial crisis. In January 2011, Brazil imposed a 60 percent unremunerated reserve requirement on banks’ short foreign exchange (FX) positions in the spot market exceeding $3 billion or Tier 1 capital (whichever is lower). The measure was motivated by concerns that banks or the local currency market could face disruptions following a large shock to the exchange rate, given the banks’ large short FX spot positions. At the same time, the measure also complemented Brazil’s IOF (Imposto sobre Operações Financeiras) tax on bond and equity inflows as it was expected to reduce the attractiveness of non-residents’ long local currency positions. These forward positions in the onshore and offshore markets, a form of carry trade, were typically facilitated by local banks which took the other side of nonresident investors’ positions and hedged themselves by borrowing FX. By raising the cost of such short FX positions, the measure was expected to affect an important channel for carry trades that was left open in the original design of the IOF while reducing potential vulnerabilities in the banking sector.

Korea. In the aftermath of the global financial crisis, Korea experienced a pronounced sudden stop of short-term external bank debt. Such debt had grown rapidly prior to the crisis driven in part by demand for currency forward contracts by the corporate sector on expectations of won appreciation. In June 2010, and following other measures, Korea introduced ceilings on banks’ foreign derivatives positions to reduce the short-term external debt that resulted from banks’ provision of forward contracts to corporates. The ceilings were expressed as a ratio to bank capital and set at 50 percent for resident banks’ and 250 percent for foreign banks branches (due to the much smaller capital for foreign bank branches). In late 2010, the authorities announced a macroprudential stability levy on banks’ non-deposit foreign currency liabilities, with increasingly penal rates on shorter maturities. This measure, which became effective on August 1, 2011, is a CFM since it is designed to affect capital inflows.

Turkey. Facing rapidly rising capital inflows, the Central Bank of Turkey (CBT) implemented from the fourth quarter of 2010 a new policy mix intended to preserve macroeconomic and financial stability. Unremunerated required reserve ratios on all Turkish lira and FX liabilities of banks were raised in several steps to an average of 14 percent and 11.5 percent respectively (from their 5 percent and 9 percent troughs during the global crisis). Moreover, higher rates were applied to shorter-duration bank liabilities. In addition, the CBT’s interest rate corridor was widened significantly to facilitate increased volatility of short-term market interest rates. The use of reserve requirements served both macroprudential and capital flow management purposes by aiming to moderate inflows and lengthen their duration.

While the above are selected examples of recent measures that can be considered both CFMs and macroprudential, not all macroprudential measures are CFMs (and vice versa). In particular, macroprudential measures that are not designed to influence capital inflows—a matter of careful judgment based on the totality of circumstances, including whether the measures were introduced or intensified during an inflow surge—would not be considered CFMs. Examples could include capital adequacy requirements, loan-to-value ratios, limits on net open FX positions, and limits on foreign currency mortgages.

Appendix II. Selected Case Studies

Selected European Countries (Bulgaria, Croatia, Poland, Romania, and Serbia)

Background

Macroeconomic conditions in a number of Eastern European countries were buoyant in the mid-2000s. Optimism about the region’s prospects stemmed from its closer integration with the European Union (EU), with EU accession by Poland in 2004, and Bulgaria and Romania in 2007. GDP growth between 2003 and 2008 was strong, and current account balances showed large deficits (except Poland), financed by even larger net capital inflows (Table II.1). Credit growth boomed during this pre-crisis period, with credit/GDP increasing by 19 percentage points in Croatia and as much as 45 percentage points in Bulgaria. At the same time, the large capital inflows led to strong asset price growth and increasing household and corporate indebtedness.

Table II.1

Macroeconomic Indicators, average 2003–08 (in percent)

article image
Sources: WEO database, various central banks, and MCM exchange rate classification.

Such external imbalances called for fiscal restraint to avoid overheating and ensure sustainability. However, only Bulgaria accumulated fiscal surpluses during this period, in line with maintaining its currency board arrangement.

The other countries maintained fiscal deficits, and only Bulgaria and Serbia reduced their public debt to GDP ratios substantially during the pre-crisis period. Foreign banks dominate the financial systems in all these countries.

Action

The primary risk that needed to be addressed was systemic risk arising from currency-induced credit risk. Specifically, with the rapid expansion in credit (a significant portion of which was offered in foreign currency), rising asset prices, and increasing private indebtedness, the ability of unhedged borrowers to repay would be undermined in the event of a large depreciation.

How Instruments Were Used

All countries used multiple instruments as a package to tackle the systemic risk (Table II.2).

  • In Poland, measures were taken in 2006 to try to contain the risks of foreign currency (FX) lending, particularly for mortgages, and in 2008 higher risk weights for FX residential loans were introduced. In 2010, Poland adopted further measures aimed at FX mortgage and retail lending, including tighter LTV (e.g., based on loan maturity) and debt service to income ratios.

  • Croatia, Romania, and Serbia adopted several measures to curb FX lending. Romania focused on lending criteria and provisioning, and introduced a gross exposure limit, while Croatia and Serbia implemented higher risk weights; these reached 125 percent (Serbia) and 150 percent (Croatia) on lending to unhedged borrowers. Serbia also introduced an exposure limit for retail lending relative to Tier I capital.

  • Bulgaria targeted measures on overall credit growth and asset price growth, such as credit ceilings, differential risk weights based on LTV, and countercyclical provisioning requirements.

  • All countries imposed LTV ratios and all but Croatia restricted profit distribution, and several put in place debt service to income limits.

  • High reserve requirements (RR) were used extensively in all cases except Poland, in which a unified low reserve requirement was maintained, in line with EU practices. The RR were differentiated by currency, maturity and source of funding (Table II.3).

  • In Bulgaria and Croatia, marginal RR (MRR) were imposed on credit growth exceeding a threshold rate and additional external borrowing by banks. The required marginal rate was set very high in Bulgaria (200 percent in 2005).

Table II.2

Prudential Measures Imposed During the Boom Period, 2003-early 2008

article image
Sources: Central bank websites.
Table II.3

Reserve Requirement Features During the Boom Period, 2003-early 2008

article image

Taking into consideration adverse scenarios (significant devaluation of local currency) when assessing the creditworthiness of borrowers.

Sources: Survey responses, central bank websites.

Countercyclical adjustments were made in several instruments. Although RRs were tightened in the period leading up to the crisis, RR rates were subsequently lowered and in some cases removed altogether (Bulgaria and Croatia). In addition, FX liquidity requirements were relaxed (Croatia and Serbia), as well as some provisioning and capital rules.

In all cases, the instrument use can be characterized as discretionary, due to the relative frequency of adjustment. The authorities in many cases found that adjustments needed to be made, either because the measures were not as effective as expected or because of circumvention. For example, several countries first set higher reserve requirements for liabilities at shorter maturities, only to find that banks exceeded those maturities by small margins to get around the regulation, and the requirement had to be extended to all maturities. In Croatia and Serbia, frequent adjustments were needed to expand the RR base, mainly to deal with circumvention. Similarly, FX indexed loans had to be brought into the same umbrella as FX loans in Serbia.

The degree of cooperation with other policies (macroeconomic and microprudential) was mixed. On macroeconomic policies, monetary policy in all five countries was applied in the same direction as macroprudential policy. However, as noted in the IMF’s Article IV consultations, fiscal policy was insufficiently tight except for Bulgaria.37 With respect to microprudential policies, the consistency with macroprudential policy improved over time. Early on, banks evaded the measures by channeling funding through non-bank subsidiaries (including leasing companies), or through asset sales to avoid the macroprudential measures (Bulgaria, Croatia). In Poland, some banks took advantage of the EU “single passport rule” which enabled them to establish branches which were not subject to stricter prudential regulations. As these circumvention tactics became known, the authorities widened the perimeter of regulation and harmonized prudential rules, and this channel for regulatory arbitrage was closed.

Outcome

The instruments had been effective in slowing credit growth and building capital and liquidity buffers in these countries. The combination of measures created capital and liquidity buffers that helped most of these countries’ banking systems withstand the financial crisis fairly well even as credit quality deteriorated (except Romania).38 Together, the instruments appear to have altered the composition of external debt in some countries, as banks’ FX liabilities stopped growing in Croatia and Serbia (Figures II.1 and II.2).39

Figure II.1
Figure II.1

Croatia: Private External Debt/GDP

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Sources: Galac 2010 and Croatia National Bank.
Figure II.2
Figure II.2

Serbia: Private External Debt/GDP

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: National Bank of Serbia.

However, the measures were partly circumvented through direct cross-border borrowing by corporate borrowers, so indebtedness was still being built up.

Fine-tuning can be helpful, but may have been taken too far for RR. It appears that simpler RR were sufficient to create liquidity buffers in Bulgaria, without the need to resort to more complex measures with very high rates that are more difficult to administer and require frequent adjustments to address circumvention (e.g., in Croatia, Serbia and Romania).

Finally, a macroprudential approach consistent with the macroeconomic policy mix appears to have worked better. This was seen in Bulgaria, where fiscal policy was countercyclical and worked in concert with macroprudential policy. In the other cases, fiscal policy was too loose, and shifted the burden of adjustment to monetary and macroprudential policy.

New Zealand40

Background

New Zealand banks depend on short-term offshore funding to provide credit. Given low national saving, they have relied on external debt to fund private sector credit. Gross external debt exceeded 130 percent of GDP in 2009, and while New Zealand’s short-term external debt declined during 2009, it remained high at almost 60 percent of GDP at end-2009 (Figures II.3).

Figure II.3
Figure II.3

Total Short-term External Debt, 2009

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IMF Staff estimates.

Before the introduction of the liquidity rules, the share of non-resident funding had grown to 40 percent of total funding. The corresponding core funding as a share of bank loans was thus lower than in most other advanced countries (Figure II.4). At the onset of the financial crisis in 2007, about 60 percent of the non-resident funding had residual maturities of up to three months (Figure II.5).

Figure II.4
Figure II.4

Shares of Domestic and Non-resident Funding by New Zealand Banks

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: Reserve Bank of New Zealand Standard Statistical Return and RBNZ calculations.Note: Other resident funding includes interbank funding.
Figure II.5
Figure II.5

New Zealand Banks’ Non-resident Funding by Residual Maturity

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: Statistics New Zealand and RBNZ calculations.Note: Based on data from December 2007.

In late 2008, New Zealand banks experienced some difficulty rolling over their short-term debt when international markets were impaired after the collapse of Lehman Brothers. Banks came to the Reserve Bank of New Zealand (RBNZ) for liquidity support and used the government’s wholesale funding guarantee to gain access to international markets (Figures II.6 and II.7). Parent banks in Australia also provided funding to their subsidiaries in New Zealand. The four largest banks in New Zealand are Australian banks’ subsidiaries.

Figure II.6
Figure II.6

Central Bank Balance Sheet Sizes

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IMF Staff estimates.
Figure II.7
Figure II.7

New Zealand Banks’ Bond Issuance

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IMF staff estimates.

The RBNZ was concerned that liquidity risk was becoming systemic. Also, in the absence of strict liquidity coverage rules, the RBNZ felt that banks may rely excessively on central bank financing instead of managing their own liquidity more prudently.

Action

In October 2009 the RBNZ introduced new quantitative requirements to increase banks’ liquidity and reduce reliance on short-term offshore funding. The requirements became effective from April 2010. This step was preceded by a consultation process with banks that sought to refine the proposed ratios to ensure that the definitions were consistent with the stated objectives of the proposal. The RBNZ is the sole institution responsible for monitoring and enforcing the liquidity rules.

How the Instruments Were Used

The minimum liquidity and funding requirements in New Zealand were conceptually aligned with the respective Basel III’s new liquidity standards ratios:

  • Liquidity mismatch ratios set minimum ‘zero’ requirements for one-week and one-month mismatch ratios each business day. The mismatch ratios compare a bank’s liquid assets and likely cash inflows with its likely cash outflows, expressing the difference as a ratio of total funding.

  • A minimum core funding ratio (CFR) that requires banks to hold sufficient retail and longer-dated wholesale funding. The minimum CFR has been set at 65 percent of total loans and advances from April 2010, increasing to 70 percent from July 2011 and 75 percent from July 2012.

The required liquidity ratios and underlying assumptions have been fine-tuned over time. The ratios are based on a generic set of assumptions that provide a standard metric for the amount of required liquid assets. Assumptions about the share of funding withdrawn consider the financial sophistication of the providers and the size of their deposit, whereby larger deposits are subject to higher run-off rates. For committed lending facilities, the assumed drawdown rate (15 percent) is based on historical figures across a range of products. Similarly, the core funding ratio is based on assumptions about retention rates in determining available stable funding. The initial minimum rate of 65 percent was set with discretion in April 2010 and, (as of late 2009); all locally-incorporated banks were expected to meet that target.

The liquidity regulation can be considered a rules-based system. It comprises differentiated rates for assumed cash inflows and outflows that may not require discretionary adjustment along the cycle. Moreover, it specifies maximum exposures to individual providers of liquidity to avoid excessive concentration and defines eligible liquid securities to preserve quality holdings. The rates are not expected to change once the new system has been fully phased in.

In New Zealand, the instruments are stand-alone measures that are not used in conjunction with other policies. In fact, monetary and fiscal policies were not aligned with the instruments at the time of their introduction, but they were not seen to have had any adverse consequences on the usefulness of the instruments.

Outcome

The liquidity instruments had an effect even before they were formally implemented. In late 2008—upon publication of a consultation paper outlining the proposed measures—banks began to change the maturity structure of wholesale funding in favor of long-term funding. Other important reasons for banks to change include pressures from financial markets such as rating agencies and banks’ own funding difficulties experienced during the global crisis. As a result, New Zealand’s short-term debt dropped from 64 percent of GDP in December 2008 to 50 percent in December 2010. This shift corresponded to a 20 percentage point drop in New Zealand’s short-term external debt ratio to 50 percent of GDP. In the run-up to implementation, banks also started competing more strongly for retail deposits, which raised bank funding costs (estimated to correspond to a hike in the policy rate of 100-150 basis points) and led to an increase in lending rates. In the months following implementation, all banks met the liquidity and funding standards, with ratios at the system level in excess of the required minima by 7-10 percentage points (Figure II.8).

Figure II.8
Figure II.8

Liquidity Mismatch Ratios

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: RBZN.1/ Once-week and one-month liquidity mismatch ratios are defined as the mismatch dollar amount to total funding.

The instruments were arguably effective in achieving the stated goals. The consultation process ensured that banks were informed about the impending policy changes and able to comply with them by the time of their implementation.

Careful calibration of parameters—values of eligible assets, run-off rates, and the minimum core funding ratio—contributed to the relatively smooth transition. It appears, however, that the impact on average funding costs was higher than anticipated.

Spain

Background

Spain introduced dynamic provisioning (DP) as a macroprudential tool in 2000. This required banks to build reserves for eventual loan losses. Previously, banks would provision against loan portfolios in two ways. First, they would set aside 1 percent of their total lending as a “generic” provision in case of loan losses. Second, they would set aside a “specific” provision for potential losses on loans in the current period that would match realized loan losses in the most recent period. The new requirement was for banks to constitute a reserve fund periodically according to a formula that took account of average loan losses over a full economic cycle, average specific provisions, as well as specific provisions in the most recent period. This approach was called “dynamic” provisioning as the contribution to the countercyclical fund varied with the economic cycle. DP was introduced soon after Spain joined the euro zone in 1999. During the 1990s, the nation’s convergence to the euro zone entailed a focus on reducing the inflation differential with Germany and tightening fiscal policy. Following convergence, the nation’s banks benefited from a significant reduction in inflation, currency, and credit risk premia, and from significant declines in long-term interest rates to near zero (from levels near 4-5 percent in the mid-1990s), allowing access to much cheaper funding than before. These developments allowed banks to lend more freely to households and companies, resulting in rapid credit growth (Figure II.9.)

Figure II.9
Figure II.9

Credit and Deposit

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: Banco de España.

Much of the lending was directed towards the acquisition or development of real estate, causing home prices to rise sharply at a rate of more than 10 percent per year in the first few years after euro zone entry and eventually reaching an annual rate of 20 percent by 2004–2005 (Figure II.10).

Figure II.10
Figure II.10

House Prices

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Sources: National Authorities, Eurostat, Bloomberg, and IMF Staff Estimates.
Action

The primary objective behind the introduction of DP was to cope with a sharp increase in credit risk on Spanish banks’ balance sheets following a period of significant credit growth during the late-1990s. A secondary objective was to give banks an incentive for more prudent loan origination that would curb credit growth, as moral suasion had proved to be ineffective and heightened competition among banks had resulted in some types of loans being too cheap (i.e., risk premia were perceived to be too low for certain operations). In addition, there had been a significant reduction in non-performing loans in the second half of the 1990s, which meant that specific provisions were quite low.41

How the Instrument Was Used

The instrument was used as a stand-alone measure, as the authorities did not apply other macroprudential tools to meet the objective of protecting against credit losses. The authorities viewed dynamic provisions as being less volatile compared with “normal” provisions comprised of generic and specific provisions. The latter typically rise sharply in a credit cycle downturn when non-performing loans (NPLs) and corresponding loan losses are on the upswing. With dynamic provisions expected to reduce the amplitude of swings in “normal” provisions, the authorities believed that it could help reduce earnings volatility.

Fine-tuning of DP: Dynamic provisions were applied across several categories of loans including mortgages (differentiated by high/low LTV), corporates, automobiles and credit cards. The formula calibrated on the basis of historical experience prior to 2000 suggested expected loss estimates ranging from 0.6 percent to 2.5 percent on these categories of loans, while average specific provisions varied from 0.1 percent to 1.6 percent.

There was a one-off adjustment of DP rates: the authorities lowered the provisioning rates in 2005, as the coverage of bad loans had risen above 300 percent in the wake of low NPLs and strong credit growth. This step resulted in a significant drop in provision coverage (Figure II.11). The liberated provisions were kept as “other reserves” in banks’ balance sheets.

Figure II.11
Figure II.11

Coverage Ratio

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: Banco de España.

The use of the instrument can be characterized as rules-based. The contribution to the fund of dynamic reserves was predicated on the difference between the average provision through the cycle and the current specific provision. Thus, dynamic provisions varied with the cycle, as specific provisions were low in the upswing and high in the subsequent downturn. While the authorities changed the DP rates once, this adjustment was not countercyclical.

The degree of cooperation with other policies was low. Monetary conditions set by the European Central Bank turned out to be too loose for Spain. Only in 2008, when the downturn had already begun, did Spain introduce more stringent treatment for commercial and residential real estate exposures than that envisaged in the Capital Requirements Directive of the EU. This was done in order to penalize non-traditional riskier mortgages requiring higher capital requirements.

Outcome

The instrument was largely effective in covering rising credit losses in Spain during the financial crisis. As credit growth declined sharply and house prices fell, banks experienced a significant pickup in NPLs, particularly in real estate exposures. These credit losses were partially absorbed by dynamic provisions. The increase in total provisioning cost (in percent of total loans) was lower than that of specific provisions as banks tapped into their dynamic reserve buffers (Figure II.12).

Figure II.12
Figure II.12

Total Provisions

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: Banco de España.

The coverage was less than full because the loan losses turned out to be much higher than expected losses at the time when the DP formula was calibrated. This conforms to the notion that loan loss reserves should cover expected losses, whereas bank capital should cover unexpected losses. It is probable that, with the benign adjustment in 2005, DP rates no longer reflected a prudent estimate of expected loan losses. On the other hand, market dynamics are inherently difficult to gauge in terms of their impact on the build-up of credit risk. It is clear, though, that capital needs would have been much higher still in the absence of dynamic provisions.

In addition, some banks were not fully covered, because DP rates were not differentiated enough. Figure II.13 shows the distribution of buffer size across banks (DP funds as a percentage of total loans as of June 2009). By this date, a significant share of banks had already run down their buffers, while fewer banks retained larger cushions. This skewed distribution is the result of the DP formula not fully capturing banks’ individual risk profiles. While the formula distinguishes between high and low risk loan segments for allbanks alike, it does not reflect that loan portfolios in a given segment differ in risk (i.e., the consumer loan portfolio of one bank has a higher expected loss than that of others). The banks that exhausted their buffers early on likely had riskier loan portfolios. To guard against this underprovisioning, rates could be differentiated by loan category and by bank.

Figure II.13
Figure II.13

Size of DP Funds (% of loans)

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: Banco de España.

DP was arguably less effective with regard to the secondary objective of curbing credit growth. After the introduction of DP in 2000, credit growth declined briefly, but it is difficult to disassociate this from the credit contraction that took place following the collapse of the tech stock bubble in 2000. Thereafter, credit grew by as much as 25 percent annually. It could be that credit would have been even more dynamic without DP, but the growth rates were high in absolute terms.

China

Background

The massive stimulus in 2008 and a delay in its exit helped fuel a domestic credit boom in 2009 and 2010. Credit growth was driven in part by lending to local government financing platforms (LGFPs), vehicles set up to make infrastructure investment, and to the real estate sector, including loans to developers and residential mortgages. Signs of overheating in the real estate sector began to emerge after mid-2009, and housing prices were rising at an annual rate of 15-20 percent by early 2010.

Action

The authorities have adopted a series of measures to curb credit growth and housing price inflation since 2010. The measures have been introduced in packages, fine-tuned with a differentiation between mortgages on first and second homes, adjusted over time at the discretion of policymakers, and include fiscal, interest rate, and administrative measures.

  • Caps on the LTV were lowered from 80 percent to 70 percent for primary homes and to 50 percent for second homes (April 2010); mortgages for third homes were suspended (September 2010); the LTV cap on second home mortgages was subsequently lowered to 40 percent (January 2011);

  • Interest rates on mortgages for second homes were raised to 1.1 times the officially administered benchmark lending rate (April 2010);

  • A capital conservation buffer, a countercyclical buffer, and a systemic capital buffer were introduced, raising the minimum capital adequacy ratio to 11.5 percent from 8 percent for large banks (2010); the provision coverage ratio was raised from 100 percent to 150 percent, and provisions were required to cover the higher of 150 percent of NPLs or 2.5 percent of total loans (2010);

  • Taxes were increased on the resale of properties within five years of purchase (January 2010); the exemptions of home purchases from stamp duties and of home sales from income taxes were abolished for all transactions except for cases involving a family’s only home (September 2010);

  • In cities with high house prices, rapid price increases, and low housing supply, local governments would limit the number of houses each family could buy, and non-local mortgage applicants were required to present proof of local tax payments for at least a year (September 2010); and

  • The official benchmark lending rate was raised five times between October 2010 and July 2011 for a total of 125 basis points, and the reserve requirement nine times for a total of 450 basis points (Figure II.14).

Figure II.14
Figure II.14

The Use of Reserve Requirements

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: People’s Bank of China.
Outcome

Overall, the measures have been effective in lowering credit growth and housing price inflation. Bank lending growth slowed to 16.9 percent (yoy) in June 2011 from 31.7 percent in December 2009. Home sales rose only 6 percent (yoy) in the first four months of 2011 compared with 30 percent (yoy) in the same period of 2010, and sales declined sharply in major cities. Home prices are leveling off, but a much anticipated house price correction has not materialized.

Colombia

Background

A combination of unsustainable fiscal positions and external shocks tipped the economy into recession in the late-1990s. The peso came under heavy pressure and was allowed to float. Currency depreciation, high unemployment and a rising current account deficit put considerable stress on the financial sector. Mortgage write-offs increased along with rising non-performing loans.

Action

The authorities implemented three prudential measures during the final months of 1999 to limit banks’ exposure to households’ default risk. The measures were introduced in a package, were broad-based, not adjusted subsequently, and not accompanied by fiscal or monetary policy actions.

  • A loan-to-value ratio limiting the loan amount to 70 percent of the value of the collateral,

  • A debt-to-income ratio limiting the borrower’s monthly debt service payments to 30 percent of disposable income,

  • A requirement limiting a bank’s global net open position in foreign currency to 20 percent of its capital; and a rule limiting the spot net open position to 50 percent of its capital.

Outcome

The implementation of the instruments was followed by a reduction in non-performing loans in subsequent years. Banks’ foreign liabilities also declined slowly while their foreign assets expanded. Credit to the private sector decreased initially and then recovered over the next few years, but non-performing loans remained subdued for some time (Figures II.15 & 16).

Figure II.15
Figure II.15

Non-Resident Assets/Liabilities

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IFS.
Figure II.16
Figure II.16

NPL Growth

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IFS.

Korea

Background

In the years leading up to the financial crisis in 2008, the Korean banking sector experienced a large buildup in short-term external debt. The external borrowing was associated with foreign exchange forward transactions, in which banks (mostly branches of foreign banks) bought dollar forwards from exporters wishing to hedge future export receipts. Banks covered their long dollar positions by borrowing from offshore banks, converting the proceeds into won and investing in local currency securities to offset the maturity and currency mismatch emerging from forward contracts with exporters. As the financial crisis hit, Korean banks were unable to roll over their maturing short-term external liabilities as global liquidity conditions worsened.

Action

The Korean authorities introduced a series of measures in the aftermath of the financial crisis to deal with large and volatile capital flows. These measures have been introduced in packages, fine-tuned with a differentiation between domestic banks and foreign bank branches, adjusted over time at the discretion of policymakers, and include fiscal policy tools.

  • Banks were required to raise their long-term foreign currency borrowing from 80 percent to 90 percent of their long-term lending, and hold at least 2 percent of their foreign assets in liquid investments rated A or higher (November 2009);

  • The value of banks’ foreign exchange forward transactions was limited to 125 percent of exporters’ future export revenues (November 2009); the limit was subsequently lowered to 100 percent (June 2010);

  • Foreign exchange derivative positions were limited to 50 percent of capital for domestic banks and 250 percent for foreign bank branches (June 2010); the limits were subsequently lowered to 40 percent and 200 percent, respectively (June 2011);

  • A withholding tax was reinstated on foreign purchases of domestic bonds, bringing it back in line with purchases by residents (January 2011); and

  • A macro-prudential levy is planned on banks’ non-deposit foreign currency liabilities (August 2011).

Outcome

The measures appear to have limited growth in banks’ external liabilities, with banks’ short-term external borrowing remaining some 30 percent below its pre-crisis levels as of 2010 (Figure II.17). However, the measures have not stemmed portfolio inflows into both debt and equity markets. The impact of the withholding tax has also been limited by existing double-taxation agreements.

Figure II.17
Figure II.17

Banks’ Short-Term External Borrowing (in US$ billion)

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IFS.

United States

Background

There were no prior formal interagency capital standards prior to the 1980s, and supervision in this area was governed by state laws or federal policies. During the 1980s, a surge in the number of bank failures, combined with a steady increase in the leverage position of some large banks, prompted the U.S. authorities to re-examine their regulatory and supervisory standards in order to ensure the safety and soundness of banks.

As part of this effort, and under the Federal Deposit Insurance Corporation Improvement Act of 1991, the authorities introduced a leverage ratio that would be applied on a consolidated basis (at the level of the bank holding company) as well as at the level of individual banks. The leverage ratio is a simple capital-to-assets ratio used to monitor a bank’s overall risk. It is intended to be used as a supplement to the risk based capital ratio. Its principal objective is to place a constraint on the maximum degree to which a bank can leverage its equity.

Although the U.S. authorities did not have macroprudential objectives in mind when the leverage ratio was introduced, it served the purpose of a macroprudential tool by containing the risk of excessive leverage building up in the financial system.

Action
How the instrument was used

The leverage ratio is expressed as a minimum ratio of Tier 1 capital to total average adjusted assets, where the latter is defined as the quarterly average total assets less deductions that include goodwill, investments deducted from Tier 1 capital, and deferred taxes.

The leverage ratio is set at 3 percent for banks rated “strong” (those that present no supervisory, operational, and managerial weaknesses and are therefore rated highly under the supervisory rating system) and at 4 percent for all other banks.42 Banks’ actual leverage ratios are typically higher than the minimum. A higher ratio may be required for any institution if warranted by its risk profile or circumstances.43

The main advantages of the leverage ratio are its simplicity and ease of application. It can be adopted quickly and monitored effectively without leading to high administrative costs. It also serves as a “back-up” against the possible failure of model-dependent, risk-based capital ratios by ensuring a minimum amount of capital. The disadvantage is that, as a balance-sheet measure, it does not take into account off-balance-sheet exposures.

The new standards were implemented by each of the Federal banking agencies44 according to their supervisory responsibilities. The leverage ratio was introduced with a broad-based application, and was not adjusted over time or accompanied by any other policy.

In 2004, a change in SEC regulation allowed investment banks to raise their leverage from 15:1 (6.7 percent) to 40:1 (2.5 percent).45

Outcome

Leverage of U.S. investment banks rose significantly after 2004 while leverage at U.S. commercial banks remained relatively low. The divergence reflected in large part the different regulatory provisions on leverage for commercial banks and investment banks. While the leverage ratio was intended to limit risk at individual banks, it appears to have helped prevent the buildup of excessive leverage in the commercial banking sector, even though the degree of effectiveness was limited. This observation is corroborated by the evidence seen in other countries, where leverage caps constrained excessive risk-taking in financial institutions.46

This said, the leverage ratio can be circumvented by financial institutions assuming leverage through off-balance sheet exposures. Its coverage should be sufficiently comprehensive, and the ratio adjusted counter-cyclically to adequately reflect rising systemic risk.

Figure II.18.
Figure II.18.

Leverage of Large International Banks and Hedge Funds*

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

1/ Source: Fitch-lbca. Leverage is defined as total assets over capital. The data points linked by a continuous line represent the weighted average of the leverage of the sample of banks in turn considered; the top of the vertical segments corresponds to the maximum value of leverage of the sample of banks considered. Continental Europe: 10 major commercial banks; Switzerland: three major banks; U.K.: five major commercial banks; US: five major commercial banks and five major investment banks. Data are as of June 2008./2 Source: Merrill Lynch. Global Fund Managers Survey. Based on a monthly survey of about 30-40 hedge funds.* Figure from Panetta. F. and P. Angelini—coordinators-and U. Albertazzi, F. Columba, W. Cornacchia, A. Di Cesare, A. Pilati, C. Salleo and G. Santini [2009) “Financial Sector Pro-cyclicality: Lessons from the Crisis.” Bank of Italy Occasional Papers. No.44.

Appendix III. The Simple Approach

Figure III.1.
Figure III.1.
Figure III.1.
Figure III.1.

Change in Risk Variables after the Implementation of Instruments47

Citation: IMF Working Papers 2011, 238; 10.5089/9781463922603.001.A001

Source: IMF Staff Estimates.

Appendix IV. GMM Methodology for Panel Regression

This appendix describes the methodology used to evaluate the effectiveness of macroprudential instruments and the full results of the panel regression.48

The goal of the exercise is to answer two simple questions:

  • What is the effect of an instrument in countries where it has been introduced?49

  • And, what would have been its effect in countries that have not actually used it?

To answer these questions, a dummy variable I is introduced with a value of 1 for countries and periods in which a particular instrument is used, and a value of zero otherwise. The dummy variable captures an average “treatment effect” of the instrument across countries, with countries and periods in which the instrument is not used as counterfactuals.

Specification

A fixed-effect dynamic panel specification is used here since a general theoretical framework for using macroprudential instruments is not yet available. The model is specified as follows:

ΔYi,t=ai,1+b1Ii,t+c1ΔYi,t-1+d1Xi,t+e1IXi,t+εi,tI(2)

For each country I, matrix I is the time-series of the value of a particular instrument (for example: the maximum LTV ratio) or a set of dummy variables that take a value of 1 during periods in which the instrument is used. Dummy variables representing a combination of instruments are also considered to capture the effect of multiple instruments used in the same period.

Matrix X represents macroeconomic variables used to control for GDP growth, the interest rate or other policy instruments already in place. IX is a matrix that captures the interaction between the macro-control variables and the instrument used. The coefficient of this matrix, e1, measures the change in the correlation between the risk variable and the control variable after an instrument is adopted. Countries that do not use the instrument are included as a counterfactual, where the value of the instrument is set to zero. Matrix Y represents the change in systemic risk after the introduction of an instrument.

For this specification, a total of 40 regressions are needed to show the interaction between four risk variables and 10 instruments. The coefficients of the interaction terms, e1, and the constant dummies, b1, are expected to be negative.

Most of the instruments are estimated for their effect on procyclicality, which is defined as the correlation between growth of GDP and growth of the risk variable on the left hand side. Some instruments may be used to reduce common exposure across institutions. These are estimated for their effect on the level of exposure to non-core funding (measured as credit to deposits of the banking sector) and foreign assets to foreign liabilities (a proxy for capital flow reversal risk).

Estimation challenges

Ordinary least squares (OLS) estimation of average treatment effects may be subject to biases. For instance, countries that adopt an instrument may need it the most (the so called endogeneity problem). Thus, if countries that introduce the instrument are those that would have had, for example, excessively high credit growth, the coefficient estimated with ordinary least squares are biased upwards. Instrumental variables are needed to address endogeneity.

The use of a dynamic panel—required to fully capture the time-series component of the effectiveness of the instruments—adds difficulties. The estimation of a dynamic panel by OLS with fixed-effects will be biased, since by construction there is a positive correlation between the lagged dependent variable and the unobserved individual-level effects.

The Generalized Method of Moments (GMM) addresses this problem, and is a standard choice for the estimation of panel data models with endogenous regressors. The GMM system estimator ensures orthogonality between the lagged endogenous variables, in both levels and differences, and the residual term. The lagged variables are used as instruments, appropriately weighted.

Data

The sample covers 49 countries for a period of ten years, from 2000 to 2010. The information required on the use of macroprudential instruments is obtained from a recent IMF survey on country authorities, as well as an internal IMF survey on country desk economists.50 Four risk variables, as identified by country authorities, are chosen: excessive credit growth risk, excessive leverage risk, liquidity risk and the risk of capital flows reversals.

Credit growth is measured as claims on the private sector from both banks and non-banking financial institutions (source: IFS). Leverage is measured as assets over equity, obtained from the IMF FSIs. Liquidity risk is proxied by non-core funding, measured as bank credit to deposits. Capital flow reversals risk is proxied by the ratio of foreign assets to foreign liabilities, for both bank and non-bank financial institutions. The source for these variables is IFS.

GDP growth and the prime lending rate are control variables for fiscal and monetary policies, respectively, both obtained from the IFS. The lending rate is used to capture the change in the price of credit, either due to changes in demand or supply in response to, for example, changes in monetary policy.

Other variables such as the policy rate or fiscal imbalances are also tested as control variables but not used. The policy rate has the disadvantage of being identical for all euro-area countries, reducing the variability across countries. In addition, the pass-through mechanism in some emerging economies is rather weak, making the interpretation difficult. Similar results are obtained when the policy rate is used instead of the lending rate and are not reported. The high correlation of fiscal imbalances with GDP (and interest rates) would result in biased estimates. In addition, the theory on fiscal policy and financial frictions suggests that fiscal policy shocks are transmitted through GDP (in the form of higher demand) and the lending rates (through risk premia).51 Since the right hand variables already capture these effects, fiscal imbalances are not used.

All variables are tested and found to be covariance stationary. Interaction terms are also tested for significance, with no further significant results. Variables in the form of dummies are used to control for the exchange rate regime, the degree of financial development and the use of other macroprudential policies.52 Possible interactions between these dummies and the instruments are also tested, without significant results.

Results

The regression results are summarized as follows. The regressions passed the Arellano-Bond for autocorrelation. However, due to the rather small number of countries in the sample, the large number of instruments used by Arellano-Bover causes the Sargan test to be weak. To further check the robustness of the results under GMM, the equation is estimated under restrictions on the lags used as instruments, as well as under ordinary least squares with fixed effects. The results are consistent, with significant coefficients of very similar magnitudes to the ones shown in the main text and the appendix.

The most significant coefficients are found on the interaction term between GDP growth and five instruments: caps on LTV, caps on DTI, ceilings on credit growth, reserve requirements and dynamic provisioning. As expected, the effect of an instrument differs in different phases of the cycle. Indeed, Table IV.1 provides the results obtained when no differentiation of the cycle is made, and the instrument is included on the right hand side as a dummy affecting the constant term and hence the level of the risk variable. Most coefficients have non-significant results.

Table IV.1

Effectiveness of Macroprudential Instruments in Reducing Credit and Leverage Growth

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variable is credit growth (top) or leverage growth (bottom), the log change in the real level of credit or leverage. Credit is measured as claims on private sector from both bank and non-bank financial institutions (source: IFS) and leverage is measured as assets over capital (source: IMF FSIs). The interest rate is the nominal long-term interest rate on prime lending, from the IMF’s International Financial Statistics. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

Non-Significant Results when Interest Rate included.

The coefficient corresponds to a dummy for the respective macroprudential instrument.

Source: IMF’s staff estimates.

However, when the instruments are analyzed during economic expansions alone (Table IV.2), the coefficients on the dummies become negative and even significant in some cases. This confirms the need to take account of different phases of the economic cycle, and in turn the rationale for focusing on procyclicality. Tables IV.3 and IV.4 provide the results under this framework.

Table IV.2

Effectiveness of Macroprudential Instruments in Reducing Credit and Leverage Growth during Booms

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variable is credit growth (top) or leverage growth (bottom), the log change in the real level of credit or leverage. Credit is measured as claims on private sector from both bank and non-bank financial institutions (source: IFS) and leverage is measured as assets over capital (source: IMF FSIs). The interest rate is the nominal long-term interest rate on prime lending, from the IMF’s International Financial Statistics. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

Not enough observations during booms

The coefficient corresponds to a dummy for the respective macroprudential instrument.

Source: IMF staff estimates.
Table IV.3

Effectiveness of Macroprudential Instruments in Reducing Credit Growth (both Level and Pro-cyclicality)

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variable is credit growth (top) or leverage growth (bottom), the log change in the real level of credit or leverage. Credit is measured as claims on private sector from both bank and non-bank financial institutions (source: IFS) and leverage is measured as assets over capital (source: IMF FSIs). The interest rate is the nominal long-term interest rate on prime lending, from the IMF’s International Financial Statistics. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

The coefficient corresponds to the interaction term between GDP growth and a dummy for the respective macroprudential instrument.

The coefficient corresponds to a dummy for the respective macroprudential instrument.

Source: IMF staff estimates.
Table IV.4

Effectiveness of Macroprudential Instruments in Reducing Leverage Growth (both Level and Pro-cyclicality)

article image
***, **, * indicate statistical significance at 1%, 5%, and 10% (two-tail) test levels, respectively.

The dependent variable is leverage growth, the log change in the level of leverage. Leverage is measured as assets over capital (source: IMF FSIs). The interest rate is the nominal long-term interest rate on prime lending, from the IMF’s International Financial Statistics. The estimation period is 2000–2010. The sample is composed of 48 countries. The regression includes dummy variables to correct for different degrees of flexibility in the exchange rate regime, individual (country) effects, a time trend (year effect) and a dummy variable for the use of other MPP instruments. Instrumental variables for the policy instrument and the GMM Arellano-Bond estimator are used to address selection bias and endogeneity.

The coefficient corresponds to the interaction term between GDP growth and a dummy for the respective macroprudential instrument.

The coefficient corresponds to a dummy for the respective macroprudential instrument.

Source: IMF staff estimates.