The Macroprudential Framework
Policy Responsiveness and Institutional Arrangements

This paper gauges if, and how, institutional arrangements are correlated with the use of macroprudential policy instruments. Using data from 39 countries, the paper evaluates policy response time in various types of institutional arrangements for macroprudential policy and finds that the macroprudential framework that gives the central bank an important role is associated with more timely use of macroprudential policy instruments. Policymakers may also tend to use macroprudential instruments more quickly if the ability to conduct monetary policy is somehow constrained. This finding points to the importance of coordination between macroprudential and monetary policy.

Abstract

This paper gauges if, and how, institutional arrangements are correlated with the use of macroprudential policy instruments. Using data from 39 countries, the paper evaluates policy response time in various types of institutional arrangements for macroprudential policy and finds that the macroprudential framework that gives the central bank an important role is associated with more timely use of macroprudential policy instruments. Policymakers may also tend to use macroprudential instruments more quickly if the ability to conduct monetary policy is somehow constrained. This finding points to the importance of coordination between macroprudential and monetary policy.

I. Introduction

The global financial crisis has underscored the need for a new macroprudential policy framework to prevent the buildup of systemic financial risks. In many countries, efforts are under way to establish such a policy framework by making new, or improving on existing, institutional arrangements. In the quest for a new policy framework, a natural question to ask is: what type of institutional arrangement is most effective? While no one size fits all, previous research by the IMF indicates that it is desirable for an institutional arrangement to provide for the timely use of macroprudential policy tools, which are shown to be effective in countering the cyclicality in the financial system.2

This paper attempts to gauge if institutional arrangements can affect the timely use of macroprudential policy instruments. This is achieved by evaluating policy response time under different institutional arrangements in a cross-country study of 39 countries, using newly updated data based on the 2010 IMF survey on Financial Stability and Macroprudential Policy (see Appendix). Policy response in this paper refers to the use of macroprudential instruments to address risks that contribute to the buildup of systemic vulnerabilities over time, or the time dimension of systemic risk.

A key assumption of the paper is that all observed policy actions are warranted by emerging risks. No attempt is made to judge when or whether policy action should be taken, which is beyond the scope of this paper. The assumption seems plausible for policy instruments used to address the time dimension of systemic risk, i.e., tightened in the upswing of the credit cycle and loosened in the downswing. To ensure that this assumption holds, all instruments in the sample are carefully examined to exclude those that seem to be unrelated to the credit cycle and have a microprudential focus. As it turns out, these instances are extremely rare in the sample.

The paper finds a negative correlation between policy response time and the central bank’s involvement in the macroprudential policy framework. This finding supports the Fund’s position that “the central bank needs to play an important role” in the macroprudential policy framework.3 While not purporting to show that a short response time is, in and of itself, effective or even desirable, the finding is consistent with previous Fund research showing that an institutional arrangement can enhance the timeliness of policy responses if it (i) facilitates systemic risk monitoring and identification, and (ii) fosters cross-agency policy coordination. The central bank is in a unique position to do both.

II. Measuring Institutional Arrangements

Real-life institutional arrangements have certain distinguishing dimensions. The Fund’s previous work has identified five such dimensions from institutional arrangements in place or being developed around the world:4 (i) the degree of institutional integration between the central bank and financial regulatory/supervisory functions; (ii) the ownership of the macroprudential mandate; (iii) the role of the government (treasury) in macroprudential policy; (iv) the degree of organizational separation of decision-making and control over instruments; and (v) the existence of a coordinating body for macroprudential policy.

These dimensions may be quantified to indicate the respective roles of the central bank and the government in various institutional arrangements. To that end, this paper constructs three indices: a macroprudential index (MaPP) indicating the role of the central bank in the macroprudential policy framework; a microprudential index (MiPP) indicating the degree of involvement of the central bank in prudential regulation and supervision;5 and a government index (MoF) indicating the degree of involvement of the government in macroprudential policy.6 The indices measure de facto arrangement and are not mutually exclusive.7 The indices assign a score of 1 to 4, with a higher value indicating a more important role:

  • The MaPP index:

    • 1 – The financial stability/macroprudential policy mandate is shared by multiple agencies including the central bank, but there is no coordination body,

    • 2 – The mandate is shared by multiple agencies including the central bank, and the central bank is a member of a coordination body,

    • 3 – The mandate is shared by multiple agencies including the central bank, and the central bank chairs the coordination body, and

    • 4 – The central bank, or a committee of the central bank, is the sole owner of the mandate.

  • The MiPP index:

    • 1 – The central bank has no regulatory/supervisory functions,

    • 2 – The central bank supervises the banking sector,

    • 3 – The central bank supervises the banking sector and part of the nonbank financial sector, and

    • 4 – The central bank supervises the entire financial sector.

  • The MoF index:

    • 1 – There is no macroprudential policy coordination body or the government is not a member,

    • 2 – The government is a member/observer of the policy coordination body,

    • 3 – The government co-chairs the policy coordination body with other agencies, and

    • 4 – The government chairs the policy coordination body.

The indices thus constructed show some common features of institutional arrangements across countries. For instance, the central bank shares the financial stability/macroprudential policy mandate with other agencies as a member of a policy coordination body (2 in MaPP) in a majority of countries in the sample (Table 1). Similarly, the central bank has prudential regulation functions in a majority of the sample countries, with 41 percent having responsibility for banking supervision (2 in MiPP), 18 percent for banking and some nonbank supervision (3 in MiPP), and 8 percent for all financial regulation and supervision (4 in MiPP). The government also tends to share the financial stability/macroprudential policy mandate with other agencies and plays a leading role in only a minority of the sample countries (4 in MoF). A full tabulation of the results is presented in Table 4.

Table 1.

Institutional Arrangement Indices

(% of countries)

article image
Source: Fund staff calculations.
Table. 3.

Response Time and Dummy Variables

article image
Source: Fund staff calculations.Note: Standard deviations are in parentheses, and ***, **,* denote significance levels of the coefficients at 1, 5 and 10 percent, respectively.
Table 4.

Institutional Arrangement Indices

article image
•The MaPP index:1–The financial stability/macroprudential policy mandate is shared by multiple agencies including the central bank, but there is no coordination body,2–The mandate is shared by multiple agencies including the central bank, and the central bank is a member of a coordination body,3–The mandate is shared by multiple agencies including the central bank, and the central bank chairs the coordination body, and4–The central bank, or a committee of the central bank, is the sole owner of the mandate.•The MiPP index:1–The central bank has no regulatory/supervisory functions,2–The central bank supervises the banking sector,3–The central bank supervises the banking sector and part of the nonbank financial sector, and4–The central bank supervises the entire financial sector.•The MoF index:1–There is no macroprudential policy coordination body or the government is not a member,2–The government is a member/observer of the policy coordination body,3–The government co-chairs the policy coordination body with other agencies, and4–The government chairs the policy coordination body.Note: The indices reflect de facto institutional arrangements, on which information is obtained from the 2010 IMF survey on Financial Stability and Macroprudential Policy, IMF country economists, the authorities’ websites, and FSAP reports.

III. Measuring Response Time

Response time measures elapsed time between the emergence of a risk and the subsequent use of a policy instrument. While the use of a policy instrument is usually well documented, it is not always clear when a risk has emerged. The judgment of risks depends on the policymaker’s risk tolerance, and there does not seem to be a universally accepted level of risk tolerance across countries. Given the difficulty in identifying the emergence of risks, this paper uses significant and distinctive changes in the behavior of a risk variable as the start point of response time. These changes are identified using two different approaches, i.e., a break-in-trend analysis and a distance from peak/trough analysis, both providing a benchmark for measuring response time without the need to assess the risk tolerance. A third approach, a threshold analysis, is also tried, but it involves judgment on the emergence of risks and is included only as an experiment.8

  • Approach 1: break-in-trend analysis. Under this approach, a structural break in the level and slope of a trend function represents a change in the behavior of a risk variable and is used as the start point of response time. To identify such structural breaks, the methodology of Carrion-i-Silvestre, Kim and Perron (2009) is used, which detects the break date by minimizing the sum of squared residuals across all possible break points. From an initially assumed number of breaks, OLS is used to remove insignificant levels and slopes to identify the final set of break dates (Figure 1).9

Figure 1.
Figure 1.

Break-in-Trend Analysis

Citation: IMF Working Papers 2013, 166; 10.5089/9781484377819.001.A001

Source: Fund staff calculations.
  • Approach 2: distance-from-peak/trough analysis. Under this approach, a turning point (peak/trough) in the path of a risk variable represents a change in its behavior and is used as the start point of response time. The turning points are estimated from a six-month moving average of seasonally adjusted data. Like Approach 1, this approach is capable of ranking policy responsiveness under different institutional arrangements by providing a common yardstick that is independent of any judgment of risk tolerance or when the risks emerged (Figure 2).

Figure 2.
Figure 2.

Distance-from Peak/Trough Analysis

Citation: IMF Working Papers 2013, 166; 10.5089/9781484377819.001.A001

Source: Fund staff calculations.
  • Approach 3: threshold analysis. Under this approach, a risk is considered to have emerged only after the risk variable crosses a certain threshold, and the threshold is used as the start point of response time. Following previous work by the Fund, a growth rate (yoy) that is 1.5 standard deviations from the historical mean during the sample period is chosen as the threshold.10 The date corresponding to this threshold is then selected as the start point for calculating response time.

Figure 3.
Figure 3.

Threshold Analysis

Citation: IMF Working Papers 2013, 166; 10.5089/9781484377819.001.A001

Source: Fund staff calculations.

In the calculation of response time, credit and credit growth are used as the risk variable. Credit is widely considered an important variable to focus on for systemic risk oversight.11 Most of the countries responding to the IMF survey have used macroprudential policy to target credit or credit growth.12 Using credit and credit growth as the risk variable to calculate response time also has the advantage of covering the largest number of instances of macroprudential policy instruments being used. On the other hand, response time calculated for other risk variables, such as leverage, liquidity and house prices, based on the use of instruments to address credit-related risks may not be accurate.13 The calculation of response time for these other risk variables based on instruments used to target them turns out not to be feasible as very few macroprudential policy instruments were used to specifically target these risks. This paper uses eight macroprudential policy instruments in calculating response time: caps on the LTV, caps on the DTI, limits on foreign currency lending, ceilings on credit/credit growth, reserve requirements, capital requirements, provisioning requirements, and restrictions on profit distribution (see Appendix).

The estimated response time reflects policy action taken in both the upswing and downswing of the credit cycle. The sample includes monthly data from 39 countries that used macroprudential instruments during the period of 2008–2011, in which both tightening and loosening of macroprudential instruments occurred. Thus, response time is calculated in a two-step process: the response times for the upswing and downswing are measured separately and then averaged to arrive at the final number. If an instrument is used multiple times in either the upswing or the downswing, the first time of its implementation is used in the calculation of response time.

The calculated average response times seem close for Approaches 1 and 2 but not for Approach 3. The results under Approach 3 excluded more than half of the countries in the sample, for which response time could not be calculated. These are countries where policy action is taken well before the threshold is reached, and excluding them makes the average response time under Approach 3 the longest among the three approaches (Figure 4). The fact that many countries take action before a threshold is reached seems to indicate that the assumption of a common risk threshold across countries, such as one represented by a growth rate 1.5 standard deviations from the historical mean, is arbitrary and unrealistic.14 A threshold estimated from cross-country panel data may be, on average, a good indicator of the likelihood of an impending financial crisis, but it may not be a probable trigger for preventive policy action in any given country. Policymakers may be more likely to look at the behavior of a risk variable over time, and in combination with other variables, than focus on a single threshold in deciding to use macroprudential policy instruments.

Figure 4.
Figure 4.

Calculated Response Time

Citation: IMF Working Papers 2013, 166; 10.5089/9781484377819.001.A001

Source: Fund staff calculations.

IV. Response Time and Institutional Indices

There seems to be some correlation between response time and the MaPP index. In particular, the average response time seems negatively correlated with the MaPP index under both Approach 1 and Approach 2 (Figure 5).15 On the other hand, the correlation is not obvious under Approach 3, but this approach is probably not representative of the sample. Approach 3 includes fewer than half of the countries in the sample and has a much longer average response time because countries that take action before a threshold is breached are excluded. The negative correlation seems to suggest that it is desirable for the central bank to play an important role.16

Figure 5.
Figure 5.

Response Time for Credit in Different Institutional Arrangements|

(months, vertical axis)

Citation: IMF Working Papers 2013, 166; 10.5089/9781484377819.001.A001

Source: Fund staff calculations.

The relationship between response time and the MiPP index is less clear. The average response time seems to be negatively correlated with the MiPP index under Approach 1, but a similar result does not hold for Approach 2 or Approach 3. It is therefore unclear whether such a negative correlation exists. While integrating prudential regulation in the central bank has the potential advantage of placing policy decision and tools in the same agency that can reduce response time, this advantage would not be achieved if the central bank were not given the decision-making role.

The relationship between response time and the MoF index is also unclear. The correlation seems nonlinear, with response time initially declining but reversing course later. This, however, does not indicate that the participation of the government in the macroprudential policy framework is not important. The fact that the average response time is the longest in the absence of a policy coordinating body (1 in MoF) under all three approaches suggests that the government’s involvement can improve the macroprudential policy framework. As noted by the Fund’s previous work, participation by government helps garner political support for policy actions, although a stronger role of the government relative to the central bank may increase the risk that short-term political considerations prevail over incentives to contain excessive exuberance in financial markets.17

V. Response Time in a Multivariate Analysis

There may be other factors, in addition to institutional arrangements, that may affect the timely use of macroprudential instruments. The previous section illustrates the possible correlation between institutional arrangements and response time. This section investigates if response time is correlated with other factors. Monetary policy, for instance, may be one such factor—the availability of the interest rate as a policy tool and the willingness of the policymaker to use it may influence how quickly a macroprudential policy instrument is used. The depth of the financial sector may also have a bearing on the use of macroprudential policy instruments as countries with small and unsophisticated financial sectors, ceteris paribus, may tend to use these instruments as a first resort more frequently. To verify the possible effect of such factors, the following cross-country equation is estimated:18

Responsetimei=α +β1MaPPi +β2MaPPi+β3MoFi+β4INTi+β5CGDPi+ɛi(a)

where i denotes country, MaPPi, MiPPi, and MoFi are the institutional indices representing the arrangement in place at the time of policy action, INTi is the standardized cumulative change in the policy rate during the period corresponding to the response time of macroprudential policy, CGDPi is the average credit-to-GDP ratio during 2008-2011 representing the depth of the financial sector.

Some caution is needed in interpreting the regression results. The sample size is so small that the regression results may be sensitive to small variations in the sample and to the influence of outliers.19 The estimated model is also too simple to identify any causal relationship, and some of the right-hand-side variables may be endogenous. While the institutional indices can be considered exogenous as they reflect the arrangement at the time of policy action, there is no sure way to ascertain that monetary policy is truly exogenous.

The caveats notwithstanding, the regression results seem to confirm the negative correlation between response time and the MaPP index. The estimated coefficients of the MaPP index under both Approach 1 and Approach 2 are negative and statistically significant (Table 2). They are also similar in size (-2.6 under Approach 1 and -3.1 under approach 2), suggesting that for each increase in central bank involvement, the response time would be reduced by about three months. The result doesn’t seem to be sensitive to the inclusion of other independent variables—the estimated coefficients of the MaPP index remain negative and statistically significant.20

Table. 2.

Response Time and Institutional Arrangements

article image
Source: Fund staff calculations.Note: Standard deviations are in parentheses, and ***, **,* denote significance levels of the coefficients at 1, 5 and 10 percent, respectively.

There seems to be no clear correlation between response time and the other two institutional indices. The estimated coefficients of the MiPP index are statistically insignificant under both approaches, although they have the same sign. This result seems consistent with Figure 4, indicating that response time may have little to do with whether the supervisory function is integrated in the central bank. Likewise, the estimated coefficients of the MoF index, though both negative, are statistically insignificant, providing inconclusive evidence on the role of the government.

The results are mixed for the control variables. The estimated coefficient of the policy rate change is positive and statistically significant under Approach 1 but not under Approach 2. A positive correlation perhaps indicates that adjusting interest rates reduces the need to use macroprudential instruments early, or country authorities tend to use macroprudential instruments quickly if, for some reason, their ability to conduct monetary policy is constrained.21 This result provides some evidence, though inconclusive, that coordination between macroprudential and monetary policies is important in addressing risks associated with credit growth. The estimated coefficient of credit-to-GDP is also statistically significant under Approach 1 but not under Approach 2. One possible explanation for the significant coefficient under Approach 1 is that, as credit-to-GDP is used as a proxy for the depth of the financial sector, countries tend to use macroprudential instruments more quickly if the financial markets are less developed. However, when per capita GDP and a few of the Worldwide Governance Indicators are tried in the regression in place of credit-to-GDP, none of the estimated coefficients is statistically significant.22 Another explanation may be simply that the risk, as represented by credit-to-GDP, becomes greater the longer it takes for policymakers to react.

As a robustness check, an alternative specification with dummy variables is estimated for response time. Equation (a) assumes a particular subjective ranking of the institutional arrangements. To see if the negative correlation still holds even if this ranking does not, the institutional indices are replaced with dummy variables representing the most distinguishing feature of the indices, i.e., which agency is in charge, and the following equation is estimated:23

Responsetimei=α+β1D1i +β2D2i+β3D3i+β4INTi+β5CGDPi+ɛi(b)

where

  • D1 substitutes MaPP, and has a value of 1 if the central bank chairs the policy coordination body or is the sole agency with the financial stability/macroprudential mandate and 0 otherwise;

  • D2 substitutes MiPP, and has a value of 1 if banking supervision is inside the central bank and 0 otherwise;

  • D3 substitutes MoF, and has a value of 1 if the ministry of finance chairs the policy coordination body and 0 otherwise;

Results of the alternative regressions corroborate the negative correlation between response time and the MaPP index. The estimated coefficients of the dummy variable for the role of the central bank, D1, are negative and statistically significant under both approaches (Table 3). This confirms the result of the regression on the indices themselves, pointing to the importance of the role of the central bank in the macroprudential policy framework. On the other hand, the estimated coefficients of the dummy variables D2, representing the integration of prudential supervision in the central bank, are statistically insignificant under both approaches. Similarly, the estimated coefficients of D3 are statistically insignificant under both approaches, consistent with the results of the regression on the indices themselves and providing inconclusive evidence on the role of the government.

VI. Conclusion

This paper finds a negative correlation between policy response time and the involvement of the central bank in the macroprudential framework. This seems to indicate that giving the central bank an important role is conducive to reducing policy response time. This finding is consistent with the Fund’s position that “the central bank needs to play an important role,” the benefit of which has been well documented:24 the central bank is in a unique position to monitor macrofinancial linkages in its capacity as the monetary policymaker and the supervisor of payments systems; it has the expertise in systemic risk identification and monitoring; and it has the experience in communicating risks to markets and the general public through the publication of a financial stability report.

Monetary policy may affect the timing of the use of macroprudential instruments, although the evidence is inconclusive. The paper’s finding of a possible positive correlation between macroprudential policy response time and changes in the policy interest rate suggests that coordination is important between macroprudential and monetary policies. Indeed, the positive correlation seems to suggest that smaller changes in the policy rate may necessitate a quicker response of macroprudential policy to mitigate risks generated by credit growth. While monetary policy should be used as the first line of defense, it is often constrained by a number of factors, including the exchange rate regime, the prevalence of foreign currency lending and an inefficient policy transmission mechanism. In those cases, macroprudential policy would be a useful complement.

A common risk “threshold” that would trigger policy actions probably does not exist. While a threshold estimated from panel data may be useful in raising a “red flag” and analyzing the likelihood of a future financial crisis, this paper finds that policy actions are often taken long before an arbitrary threshold is reached. Rather than taking action when credit growth crosses a certain threshold, policymakers are likely to monitor a range of indicators, including changes in risk variables and market intelligence, in considering policy options. Policymakers also rely on judgment that cannot be easily captured by the risk variables, and often take action only after a confluence of evidence suggests that action is needed.

Policy response time is only one aspect of an effective macroprudential policy framework. This paper has focused on the time dimension of systemic risk, i.e., how quickly policy has responded to changes in the credit cycle under various institutional arrangements. An effective institutional arrangement, however, should be equally capable of addressing the cross-sectional dimension of systemic risk, i.e., common exposures, linkages, and interdependencies that may be sources of contagion and spillover risks to the whole financial system. In addition, the results of a cross-country study will only hold “on average,” and country-specific factors may be important in determining the responsiveness of institutional arrangements. These factors include the quality of supervision, policy coordination mechanisms, and the approach to economic management, which cannot be easily captured in a cross-country study but should be given adequate attention in establishing macroprudential frameworks in individual countries.

The Macroprudential Framework: Policy Responsiveness and Institutional Arrangements
Author: Cheng Hoon Lim, Mr. Ivo Krznar, Mr. Fabian Lipinsky, Mr. Akira Otani, and Mr. Xiaoyong Wu