Macroprudential Policy Implementation in Europe. Session 8: First lessons learned: calibration, transmission, impact

Transcription

1 Macroprudential Policy Implementation in Europe Session 8: First lessons learned: calibration, transmission, impact Katarzyna Budnik and Marco Lo Duca October 2018

2 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 2

3 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 3

4 Key messages from the literature Macroprudential policy relatively new in advanced economies: Emerging markets actively used macroprudential policies (e.g. borrower based measures, capital flow management measures, etc) Literature on effectiveness of macroprudential policy Cross countries studies are predominant Literature learns also from other fields (e.g. impact of capital measures on banks behavior) Key datasets of macroprudential measures extends to financial stability policies used in the past with macroprudential goals Findings of the literature Overall macroprudential policy effective Heterogeneity effectiveness across outcome indicators, instruments and countries 4

5 Cross countries studies Key research question Do outcome indicators differ across countries depending on the activation of macroprudential instruments? Datasets of enacted macroprudential policies IMF, ECB, BIS Outcome variables Macro: credit growth, house prices, household/corporate debt, etc Micro: bank balance sheet variables Shortcomings Measurement of macroprudential policy actions and identification Binary variables: instrument or more instruments activated Not clear whether policy binding / which policy binding Endogeneity issue (e.g. activation in high credit growth countries) 5

6 Cross countries studies: selected findings Article Macropru dataset Empirical strategy results Cerutti et al 2017 IMF Survey; quarterly, coverage 113 countries, 18 measures, between 2000 and 2013 Unit: country; quarter Outcome indicators: credit growth Identification: dummy 0, 1 when instrument is active Econometrics: panel regression LTV and DTI: Overall credit and house price growth no impact; DTI: reduces household credit growth Claessens et al IMF Survey; annual, coverage 48countries, 2000 and 2010 Unit: bank; year Outcome indicators: leverage; assets; non-core liabilities; Identification: dummy 0, 1 when instrument active Econometrics: panel regression LTV and DTI reduce the growth rate of outcome variables. They also reduce credit growth; Less conclusive on house prices Vandenbussche al Own data; quarterly,16 European countries, 1990s to 2010; 29 measures Unit: country, year or quarter Outcome indicators: house price growth; Identification: index of stringency Econometrics: panel regression DTI slightly reduce house price growth; LTV no impact. Kuttner and shin (2016) Own data; quarterly,57 countries, 1980s to 2012 measures Unit: country; quarter Outcome indicators: credit and house price growth; Identification: for activation dates Econometrics: panel regression DSTI rules more consistently associated with declines in credit growth than LTVs. No impact of house prices. Note: capital measures covered more extensively in subsequent slides 6

7 Evidence from literature: impact of capital measures 7

8 Evidence from literature: impact of capital measures 8

9 Evidence from literature: impact of capital measures 9

10 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 10

11 Calibration of measures and concept of policy sufficiency Source: ESRB Flagship Report 11

12 Calibration of measures and policy sufficiency Policy sufficiency: Policies are sufficient when they are calibrated in a way: 1. that meets policy objectives (and has the desired impact on target variables) 2. That ensures that benefits exceed costs Key issues Identification of target variables in relation to policy objectives Cost benefit analysis (what are the costs?) Ex-ante calibration and ex-post assessment Ex-ante calibration: the expected impact of policy measures meets goals Ex-post assessment: evaluation of whether policy measures meet the goals 12

13 Calibration of measures: quantitative and qualitative approaches Qualitative approaches: International best practices / practices in peer countries / realized losses in past situations of financial distress (either in the domestic economy of abroad) Easy to communicate but might not lead to accurate calibration Quantitative approaches: Calibration informed by models Importance of using a range of tools: different models emphasize different aspects of the transmission of macroprudential policies (e.g. different costs and benefits) and therefore they complement each other Complex to communicate but might lead to more accurate calibration (if methods are robust) 13

14 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 14

15 Ex- ante calibration of capital measures Primary Goal: to increase bank resilience to the materialization of risk Calibration should look at whether bank capital is commensurate to risks Calibration should look at measures of bank risk Example: the capital measure covers losses emerging in adverse scenario where the identified macroprudential risks materialize Secondary goal: to affect bank incentives (impact on pricing, risk features, and quantity of new lending) Calibration should quantify the impact on lending, portfolio rebalancing; spreads, etc; Potential costs: lower economic activity; lower lending, higher margins; intermediation moves to unregulated entities; Assessment of costs depends on financial stability goals 15

16 Ex- ante calibration of capital measures: findings from a DSGE model 3D DSGE model (Mendicino et al. 2015) Developed to quantitatively assess the impact of macroprudential policy instruments on financial intermediaries and the economy Compares short- and long-run costs and benefits, taking into account general equilibrium effects and optimizing behavior of agents 3D stands for 3 layers of default : borrowing households (HH), nonfinancial corporations (NFC) and banks can all default on their liabilities Financial frictions costly state verification implies that lending premium depends on agents default probability 16

17 Ex- ante calibration of capital measures: findings from a DSGE model 17

18 Ex- ante calibration of capital measures: findings from a DSGE model Capital requirements: increase equity to be held in relation to loans Prevailing effect depends on initial conditions and specific features of the national banking system 18

19 Ex- ante calibration of capital measures: findings from a DSGE model Short run costs: lower GDP and total credit Gradual adjustment to equilibrium: short run costs dissipate and benefits from lower default probability and deposit premium materialize 19

20 Ex- ante calibration of capital measures: findings from a DSGE model 20

21 Ex- ante calibration of capital measures: findings from a DSGE model 21

22 Ex- ante calibration of capital measures: findings from a DSGE model 22 Note: propagation of a shock to bank risk. Upper panel: baseline capitalization. Lower panel: 1 p.p. higher capitalization.

23 Ex- ante calibration of capital measures: benefits of capital measures 23

24 Ex- ante calibration of capital measures: macroprudential use of stress tests 24 Source: ESRB Final report on the use of structural macroprudential instruments in the EU. Adaptation of Benanni et. Al (2017)

25 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 25

26 Borrower Based measures Authorities specify rules that banks should apply on credit standards: Loan to Value (LTV) ratio It requires borrowers to use equity ( skin in the game ) It lowers incentives to default (lowers PD) and it lowers the loss given default (LGD) Debt/Loan to Income (D/LTI) ratio It links the borrowing capacity to the income of the borrower It lower credit risk (it lowers PD) Debt/Loan service to Income (D/LSTI) ratio It links the borrowing capacity to servicing costs in relation to income of the borrower It ensures the existence of buffers to meet regular payments (it lowers PD) Maturity limits and amortisation requirements Also complementary role (eg DSTI) 26

27 Ex- ante calibration of borrower based measures Primary Goal: improve quality and reduce quantity of new lending ( leaning against the wind ) Calibration should look at the impact on the quality and quantity of lending, and on other cyclical indicators of risk (e.g. house prices) E.g. the expected effect of the measures brings lending standards and lending growth in the comfort area of the macroprudential authority Secondary goal: bank and borrower resilience (medium term) Calibration should look at the impact on PDs and LGDs of borrowers and banks in the medium term (via lower credit risk of the loan portfolio) E.g. the expected effect of the measure decreases bank capital needs in adverse scenarios ceteris paribus Potential costs: access to finance for categories of borrowers (political issue); profit opportunities for banking industry; macro implications (lower housing investment); Assessment of costs depends on financial stability goals 27

28 Ex ante calibration of borrower based measures: link between credit standards PDs and LGDs Strong link: LTI PD; LTV LGD Implications for bank PD and capital 28

29 Ex- ante calibration of BB measures: key choices Guidance on prudent lending standards vs legally binding actions Example: PT and AT (comply or explain policies) Example: SK, LT, LV, EE, FI, NL (legally binding limits) Activation in the cycle of hard limits freeze lending standards early in the cycle with limited immediate impact on lending Impose tighter limits than actual lending standards to cool down lending Use of speed limits (or exemption) Speed limit = a fraction of loans is exempted from the limit Other choices Differentiation across categories of borrowers Definition of key parameters (e.g. value, income, etc) 29

30 Ex- ante calibration of BB measures: key choices, Slovakia example Key objectives and evolution of borrower based measures in Slovakia Original setting ( ) (non binding) 1 st revision ( ) (binding) 2 nd revision (as of July 2018) (binding) Main objective of measures Sound and sustainable credit growth (preventing excessive easing of lending practices) 1. Sound and sustainable credit growth 2. Addressing risks related to vulnerabilities emerging on the RRE market 1. Sound and sustainable credit growth 2. Addressing risks related to rising household indebtedness Original setting ( ) (non binding) LTV limit Max. share of LTV 90+: 10% 1 st revision ( ) (binding) Max. share of LTV 90+: 10% Max. share of LTV 80+: 40% (phase in) 2 nd revision (as of July 2018) (binding) Max. LTV 90% Max. share of LTV 80+: 20% (phase in) DSTI limit 100 % 80 % (phase in) 80 % Interest rate sensitivity test Maturity limit Amortization rule Applies to new loans only RRE secured loans: 30Y (excep. 10 %) Unsecured loans: 8Y (phase in applied) Applies to all customer s loans with variable interest rates No change No change No change Mandatory amortization with annuity No change No change DTI Not set Not set Source: NBS DTI limit: 8 (after tax income) Max. share of DTI 8+: 10% 30

31 Ex ante calibration of BB measures: considerations for the calibration Modelling the transmission of BB measures: At the individual level they are effective if the credit standard is binding for the borrower At the aggregate level, effectiveness depends on how many borrowers are constrained Micro data are crucial to understand tightness of measures E.g. financial wealth of individuals affects effectiveness of LTV ratios Use of combined micro and macro approaches to assess the impact of BB measures 31

32 Ex ante calibration of BB measures: combined micro and macro approach A combined micro and macro approach for the calibration of BB relies at least on the following steps 1. Uses micro data to identify the impact of the policy on the population of borrowers 2. Quantifies the impact on lending on the basis of step 1 (+ assumptions) 3. Quantifies the impact of lending (step 2) on other key variables 32

33 Ex ante calibration of BB measures: combined micro and macro approach 1. Use micro data to identify the impact of the policy on the population of borrowers 33

34 Ex ante calibration of BB measures: combined micro and macro approach Step 2: Quantify the impact on lending on the basis of step 1 (+ assumptions). Options: 1. Strict: loan applications above caps are denied (Gross Poblacion 2017); 2. Limited: loan applications above caps are accepted up to the cap; 3. Conditional: part of the gap (between demand and caps) is supplied, conditional on borrowers LTV, DSTI, LTI metrics (Kelly et al 2015). 34

35 Ex ante calibration of BB measures: combined micro and macro approach Step 3: Quantify the impact of lending (step 2) on other key variables Transform new mortgage lending shocks into shocks to mortgage credit outstanding feed into Macro model and assess impact on key macro variables. Simpler alternatively: use elasticities at fixed time horizon (i.e. of house prices to credit) 35

36 Ex ante calibration of BB measures: combined micro and macro approach 36

37 Ex ante calibration of BB measures: extended micro and macro approach 37

38 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 38

39 Ex-post assessment: policy evaluation Source: ESRB Flagship Report 39

40 Ex post assessment of the impact of policies Is the observed impact of the enacted policy in line with the exante expectation? Assess whether policy meets the initial goals Is the impact on risk indicators visible? are benefits observable and in line with expectations? are costs observable and in line with expectations? What is the information content of deviations from the expected path? Policy not effective? Role of other drivers in observed developments developments / counterfactual analyses Is there evidence of leakages or circumvention? 40

41 Ex post evaluation of policies Belgium Ferrari, Pirovano, Kaltwasser (2017) Policy: 5 ppt add-on to RW of domestic RRE mortgage loan portfolio of IRB banks Evaluation of the impact of the policy on credit pricing (link to paper) Methodology: difference-in-difference estimation using bank-level panel data. allowing the average lending spread across banks to vary according to banks balance sheet characteristics allowing IRB banks response to the macroprudential measure in terms of mortgage loan pricing to differ according to IRB banks balance sheet characteristics Sample of monthly observations for 13 banks 8 IRB banks -> affected by the policy 5 STA banks -> not affected (control group) 41

42 Ex post evaluation of policies Belgium Ferrari, Pirovano, Kaltwasser (2017) 42

43 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 43

44 Structural Factor Augmented Vector Autoregression (FAVAR) [Please select] [Please select] Captures the joint dynamics of macroeconomic aggregates and bank-level variables GDP, residential house prices and commercial real estate prices Bank-level credit to the non-financial private sector, credit risk, banks capitalization, funding costs Distinguishing features Capturing both time dimension (the speed of pass-through) and distributional effects; Tailored for cases where we have to deviate from the representative bank assumption; Allows the identification of structural shocks but lets data speak; Parsimonious (easily extendable). 44

45 Augmented reduced-form VAR: A FAVAR model 45 t t t t t F Y L F Y 1 1 ) ( t Y - a vector of observable variables (GDP, residential house prices) - a (small) vector of latent factors F t Factor identification: t t F t Y t u F Y X t X - a (large) vector of bank level variables (credit, credit risk, banks capitalisation) - loading matrices

46 Benefits and costs of capital regulation Benefits from higher capitalisation = an unrealised reduction in bank credit under adverse circumstances (banks resilience/smoothening of credit supply) Costs of higher capitalisation = a reduction in bank credit along the transition path Applicable to the assessment of system-wide and banklevel capital buffers 46

47 Three questions = three specifications of SFAVAR One model: countrylevel structural FAVAR Three specifications => Three analytical questions => Banks capitalistation => banks resilience? => Three policy messages: Long-run benefits of capital regulation; Short-run costs of systemwide capital regulation; Short-run costs of selective (e.g. O-SII) capital regulation. Structural FAVAR Short-run effects of system-wide capital buffers on credit and economic activity? Short-run effects of selective capital buffers? Type of banks adjustment vs. the outcomes. 47

48 Workings: macroeconomic outcomes Median IRFs of standardised macroeconomic variables to a one standard deviation positive aggregate demand, residential house price and commercial house price shocks Legend: Figures are constructed on the basis of median responses of macrovariables from the country models. All IRFs correspond with the behaviour of standardised variables on a cumulative basis. Solid line marks the median response, while dark blue fan represents 50% and light blue 68% of the distribution of median responses of macrovariables on a country level. IRFs of commercial property prices and representing the response to structural shocks to commercial property prices are based on a narrower sample of countries: Finland, France, Ireland and the Netherlands. Source: Budnik et al. (2018), OMR TF Macro-level results comparable to those from any other macroeconometric (DSGE) model 48

49 Workings: bank-level IRFs Standardised IRFs of bank-level non-performing loans to a one standard deviation positive aggregate demand shock Legend: Blue lines correspond with IRFs of O-SII (or G-SII) banks. Red lines with IRFs of remaining (smaller) banks. Black dashed line represents the IRFs of aggregate (system-wide) credit to non-financial private sector. Budnik and Bochmann (2017) Comprehensive bank-level information comparable with micro models 49

50 Bank-level IRFs Standardised IRFs of bank-level non-performing loans to a one standard deviation positive aggregate demand shock Legend: Large banks include mostly G-SII and O- SII banks. Figures are constructed on the basis of median responses of banklevel variables from the country models. All IRFs correspond with the behaviour of standardised variables on a cumulative basis. Red (blue) broken line marks the median response, and dark red (blue) fan represent 50% and light red (blue) 68% of the distribution of median responses of bank-level variables of G-SII and O-SII (other) banks. Solid black line marks the median response for the full sample of banks. Source: Budnik et al. (2018), OMR TF 50

51 Convex relationship between banks reaction and capitalisation level Empirical relationship between the change in credit to non-financial private sector following a positive aggregate demand shocks and the initial level of banks capitalisation As a rule: better capitalised banks reduce credit less when economic conditions turn sour (higher resilience) Source: Budnik et al. (2018), OMR TF For very high levels of banks capitalisation the negative relationship breaks. Beefing up bank capital does not bring in additional gains 51

52 Large banks differ The marginal effect of an increase in Tier1 ratio from the level of 8.5% on the cumulated credit to non-financial private sector following a positive aggregate demand shock On average: a negative effect of banks capitalisation on credit provision is highly persistent (4-5 years) For large banks the effect is larger and even more persistent (extending beyong 5 year horizon) Legend: Estimates based on bank size specific regressions with OLS estimator and robust standard errors. 52

53 Has banks resilience increased following the phase-in of CRDIV? The effect of an increase in Tier1 capital buffers from the levels that prevailed in 2010 to the levels at the end of 2015 on the outstanding credit to non-financial private sector under the adverse scenario (at the end of a year) Legend: Estimates based on bank size specific regressions with OLS estimator and robust standard errors clustered by banks. The effect of Tier1 capital buffers at the end of 2015 above the level of 8.5% on the outstanding credit to non-financial private sector under the adverse scenario (at the end of a year) The effects of the existing capital buffers assessed against an adverse but plausible senario (the combination of historical structural shocks as in the EBA ST methodology) Legend: Estimates based on bank size specific regressions with OLS estimator and robust standard errors clustered by banks. Source: Budnik et al. (2018), OMR TF 53

54 Short-run propagation of bank capital shock Legend: Figures are constructed on the basis of median responses of macrovariables from the country models. All IRFs correspond with the behaviour of standardised variables on a cumulative basis. Broken line marks the median response, and dark blue fan represent 50% and light blue 68% of the distribution of median responses of macrovariables on a country level. Source: Budnik et al. (2018), OMR TF 54

55 Credit and GDP losses following an increase in average capital ratio Total credit Lending spreads Economic activity (GDP) Path of capital increase A gradual increase in Tier1 capital ratio translates into a contraction in credit between 1 and 3pp High impact (IT, LT) and low impact (BE, FR) countries Source: Budnik et al. (2018), OMR TF 55

56 Deleveraging or accumulating capital? Median standardised capital shocks related to deleveraging and capital accumulation in the sample Both types of adjustment are empirically relevant 56

57 Selective capital regulation: large banks only IRFs of bank-level capital ratios to non-financial private sector to an increase in a capital ratio of G-SII and O-SII banks (80-100%) that adjust via asset deleveraging Large banks increase capital ratios via deleveraging resulting in a prolonged reduction of credit and assets Other banks (partially) follow the suit => Upper estimate of costs of capital regulation 57

58 The way banks adjust matters IRFs of bank-level credit to non-financial private sector to an increase in a capital ratio accommodated by > 80% of banks in the system via beefingup capital volume If banks adjust by beefing-up capital there is a persistent increase in total assets and credit to non-financial private sector => Lower estimate of costs of capital regulation 58

59 Findings Multi-model approach is advantageous for deadling with model uncertainties Different models can emphasize different elements: Theory versus data based approaches Representative bank versus heterogeneous banks structures and can serve different purposes: Ex post evaluation Ex ante calibration A FAVAR (Factor Augmented Vector Autoregression model) provides an example of a relatively universal approach that can complement foremost structural DSGE models 59

60 Overview 1. Literature review on impact of measures 2. Concept of policy sufficiency 3. Ex-ante calibration of capital measures 4. Ex-ante calibration of borrower based measures 5. Ex-post assessment 6. Model application: FAVAR 7. Measuring the macroprudential stance 60

61 Policy stance Stance establishes a relationship between actions and a policy objective Policy actions = the calibration of macroprudential instruments The objective = financial stability (preventing or mitigating systemic risk) The policy stance condenses the information about the direction in policies (their current intensity and the character of changes) 61

62 Monetary policy stance Neutral policy stance: The implemented policies are sufficient to meet (and maintain) the policy objectives in the desired time horizon Monetary policy e.g.: the central bank interest rate is set so that 2% inflation target is met in the medium run Loose policy stance: The implemented policies are not sufficient to meet the policy objectives in the desired time horizon Monetary policy e.g.: the central bank interest rate is set so that the inflation rate is expected to remain above the target 2% level Tight policy stance: The implemented policies are likely to overshoot the target in the reference time horizon Monetary policy e.g.: the central bank interest rate is set so that the inflation rate is expected to continue falling below its target 2% level 62

63 Monetary policy stance: observations The assessment of stance depends on the level of interest rate (the higher interest rate, the tighter the policy stance), but also on the intensity of inflation risks (to achieve neutral policy stance the central bank interest rate has to increase with rising inflation risks) The description of stance requires policymakers to form preferences regarding: Targets (e.g. 2% inflation rate) Horizon over which they are met (e.g. at each instance, or only over the medium run) Neutral policy stance will be sufficient to describe a tight and loose policy stance (the two latter are relative concepts) The notion of stance is generally positive (though it can be used normative statements e.g. too loose policies) 63

64 Complexity of macroprudential policy stance Neutral macroprudential policy stance: the implemented macroprudential policies are sufficient to meet and maintain financial stability in the desired time horizon The difficulty of measuring the policy objective (financial stability) Multi-dimensional instrument set Knowledge about the pass-through of policy instruments is still scarce Interactions between risks and instruments Overlaps with other policies: microprudential, monetary (via the lending channel), fiscal (e.g. taxation of real estate investments), competition policy 64

65 Policy objectives: between risks and resilience the ultimate objective of macro-prudential policy is to contribute to the safeguarding of the stability of the financial system as a whole, including by strengthening the resilience of the financial system and decreasing the buildup of systemic risks, thereby ensuring a sustainable contribution of the financial sector to economic growth. (ESRB Handbook of Macroprudential Policy) Reducing systemic risks Increasing system resilience Reducing the probability of the financial crisis Reducing the depth (consequences) of the financial crisis 65

66 Measurement of systemic risks: example Systemic Risk Indicator (Jan Apr. 2013; probability in percentages) Source: ECB Interpretation: Market perception of the probability of an adverse systemic event i.e. probability of simultaneous default of two large EU banks within the next two years Construction: based on CDS spreads and equity returns.

67 Measurement of systemic risks A great share of indicators used to measure systemic risks have a partial nature e.g. credit-to-gdp gap measures foremost cyclical risks Growth-at-Risk is a promising alternative. It links financial conditions to the distribution of future growth outcomes. It can assess whether a tightening or an easing of financial conditions may put financial stability and future growth at risk. 67 Source: IMF

68 Growth-at-Risk: example Source: IMF 68

69 Measurement of resilience Relative scarcity of approaches targeted at measuring system resilience Stress testing is a way to identify vulnerabilities in the system and acquire the forward-looking view of system s shockabsorption capacity under stress Source: FSR May 2018

70 Example: Banking Sector Macro Stress Test Contagion Scenario generator Satellite models Balance sheet and Profit&Loss Solvency models Source: Henry and Kok (2013), ECB Occasional Paper No Dynamic adjustment model Identify key systemic risks at the current juncture Design severe but plausible macro-financial scenario(s) 70 C. Macro feedback models Translate scenario into institution-specific credit risk parameters (e.g. PDs/LGDs), market risk (e.g. MTM valuation) and bank profitability Assess banks resilience (e.g. capital buffers remaining after shocks to balance sheet components).

71 Instrument measurement: stylised example 1996Q1: introduction of an LTV limit on mortgage loans of 90% [level] for second-home buyers [scope] [activation] 1998Q2: an introduction of a stricter LTV limit of 80% for FX mortgage loans [currency] for first-and second-home buyers 1999Q1: tightening of the LTV limit on FX loans to 70% and extending the LTV limit on domestic currency loans to second-home buyers 2003Q1: loosening of the LTV limit on mortgage loans in domestic and FX currency 10% of loans in bank portfolio can be exempted from the limit [exemptions] 2008Q2: LTV limit on FX currency loans removed 2014Q4: LTV limit on mortgage loans in domestic currency removed [deactivation] 71

72 Instrument measurement: state-of-the-art 1996Q1: introduction of an LTV limit on mortgage loans of 90% [level] for second-home buyers [scope] [activation] 1998Q2: an introduction of a stricter LTV limit of 80% for FX mortgage loans [currency] for first-and second-home buyers Examples of use: Lim et al (2011), Cerutti et al (2015) 1999Q1: tightening of the LTV limit on FX loans to 70% and extending the LTV limit on domestic currency loans to second-home buyers 2003Q1: loosening of the LTV limit on mortgage loans in domestic and FX currency 10% of loans in bank portfolio can be exempted from the limit [exemptions] 2008Q2: LTV limit on FX currency loans removed 2014Q4: LTV limit on mortgage loans in domestic currency removed [deactivation] Examples of use: Akinci and Olmstead-Rumsey (2015) 72

73 Sectoral stance measurement: a Czech National Bank CCyB example Measurement of risk/resilience: Composite financial cycle (FCI) indicator which gives early warning signals (6-8 quarters ahead) by aggregating a wide range of countryspecific financial risk factors (e.g. credit growth, property prices, lending conditions, etc.). Instrument: CCyB rate Policy target: the FCI guidance. For example, the maximum value of the historical distribution of the indicator (observed before the outbreak of the financial crisis) is associated with a CCyB rate of 2.5%. Starting from this maximum value, intermediate CCyB rates are set in line with a number of ranges for values of the indicator. Disclaimer: the CNB applies the guidance only discretionary. 73

74 Sectoral stance measurement: a Czech National Bank CCyB example Neutral stance = CCyB at the FCI guidance (blue line) e.g. 2016Q1 and 2016Q2 Tight stance = CCyB above the FCI guidance e.g. 1Q2018 Loose stance = CCyB below the FCI guidance e.g. between 2014Q4 2015Q3 and between 2016Q2 and 2017Q2. 74

75 Sectoral stance measurement: stylised euro area O-SII example Measurement of risk/resilience: institutions are given a score from 0 to bps representing their systemic riskiness as in EBA Guidance (size, importance, complexity and cross-border activities, interconnectedness). Instrument: O-SII buffer rate Policy target: the mean (cross-country) calibration of O-SII buffers at a given value of scores. Disclaimer: Caution is needed when interpreting this example as the neutral stance also depends on policy preferences and risk tolerance that are known not to be the same across countries. 75

76 Sectoral stance measurement: stylised euro area O-SII example Average country-level scores based on the EBA Guidelines and average buffers for EA countries. Neutral stance = OSII buffers at the regression line Tight stance = OSII buffers above the regression line Loose stance = OSII buffers below the regression line Source: ECB. 76

77 Additional complexity: Interactions Interactions => The overall macroprudential stance is not a simple sum of individual stances. between instruments, e.g.: maturity restrictions increase the effectiveness of DSTI limits general capital buffers (imposed on all banks exposures) can limit the effectiveness of borrower-based measures targeting the real estate sector between objectives, e.g.: limits on large exposures should reduce the direct contagion, but may increase indirect contagion (via common exposures) 77

78 Additional complexity: pass-through and policy preferences The novelty of macroprudential policy means that relatively little is know about its pass-through into policy objectives and time required for policy to work It also means that macroprudential authorities hardly ever communicate their policy preferences toward e.g. the horizon over which the target is met the volatility of the instrument versus the intermediary targets the hierarchy of different (intermediary) targets the hierarchy of different instruments 78

79 Wrap-up on the measurement of macroprudential policy stance Measurement of macroprudential policy stance poses more chellanges than that of monetary policy (though comes close to fiscal policy) Some problems will become less acute with time and experience, and e.g. better knowledge of the pass-through of instruments standardisation of macroprudential instruments (easing their comparability) At many practical instances, the measurement of changes in policy stance may be easier that of its level 79