Macro-prudential policy making: where’s the evidence? - Deputy Governor Sharon Donnery

Remarks by Deputy Governor Sharon Donnery at the joint Banco de Portugal - European Central Bank conference on macro-prudential policymaking

It is a great pleasure to speak to you all today on the topic of evidence-based macro-prudential policymaking.¹

One of the great philosophers of science of the 20th century, Karl Popper, once said “the role of evidence is, in the main, to correct our mistakes, our prejudices, our tentative theories—that is, to play a part in the critical discussion, in the elimination of error.”²  But, “by correcting our mistakes, we raise new problems. And in order to solve these problems, we invent conjectures”…and so on.³

 Analytical approaches to macro-prudential policy are in their infancy.⁴ I therefore believe we must be humble about what it can do, and clear about what it cannot. We must learn from our experience, use evidence to refine our hypotheses and theoretical models, correct our errors and develop our thinking, and indeed our instruments.

In this context, my aims for my intervention today are threefold.

I will begin with a justification for the implementation of macro-prudential policies and the importance of an evidence-based approach.

Secondly, I will discuss the range of approaches that can be placed under the umbrella of a macro-prudential “evidence base”.

Finally, I will outline what I believe to be the state of play currently in this sphere and where additional and deeper thinking is required of all of us as a policy-making and research community.

***

The imposition of system-wide policies, which could help to alleviate systemic risks and dampen credit and asset price cycles, predates the recent financial crisis.

However, the intensity with which macro-prudential policies are being implemented, as well as the breadth of research and discussion that surrounds them, have both expanded greatly in recent years.

Advanced economies also now have much more formal frameworks for the design and introduction of macro-prudential policies including national macro-prudential and national designated authorities in EU Member States.

This increased interest in macro-prudential policy relates, in part, to the legacy of the financial crisis and the realisation that monetary and fiscal policies and micro-prudential supervision are not by themselves sufficient to ensure financial stability.  It is also in my view closely related to the fundamental paradigm shift that has faced the central banking community globally in recent years: that of accommodative monetary policy, with unprecedented expansions in central bank balance sheets when policy rates faced the zero lower bound.

We currently have a growing evidence base that tells us that these unconventional policies have had a range of beneficial effects on outcomes such as borrowing costs, inflation and economic activity in the real economy.⁵ Moreover, the recovery in the broader macroeconomic environment contributes in a positive way to financial stability.

However, observers have rightly pointed out that with this unprecedented monetary accommodation, which of course has as its primary objective the restoration of price stability in the euro area, come a range of financial stability risks in asset markets.

More particularly, accommodative monetary policy conditions and low interest rates can induce investors to seek riskier, higher-yielding assets and result in increasing leverage and higher asset prices.

In my own view, the optimal policy mix in this environment is for central banks to continue to focus monetary policy on the restoration of price stability, while allowing macro-prudential policy to act heterogeneously across asset classes, sectors and countries in response to developing risks, some of which may be related to this monetary accommodation.

This underlines the importance of an evidence-based framework for macro-prudential policy, both in terms of the design, introduction and evaluation of policies but also our communications in relation to policies.⁶

 Such a framework is essential for many reasons:

• The instruments are new to many countries, and will be unfamiliar to the public.
• There is potential for side-effects in unrelated sectors of the economy.
• There may be distributional consequences, which impact on particular cohorts of the economy and spark political interest.
• That we face a “horizons problem”, in that the costs are often borne immediately, whereas the benefits may be longer term, nebulous and difficult to quantify.
• And, the effectiveness of a policy may also be determined in part by the “expectations channel”, whereby credible signalling by a macro-prudential authority alters how the public assesses risks and allows them to infer likely future policy actions.⁷

For all these reasons and more, clear communication with the public, built on a bedrock of solid analytical evidence, is essential to the successful implementation of these policies.

In contrast to monetary policy, macro-prudential policy, and the evidence base and analytical approaches that underpin it, remain in their infancy.

For most countries, the empirical evidence may simply not exist to directly answer the questions that need answering in as direct and robust a way as exists for monetary policy.

However, given the aforementioned possibility of emerging financial stability risks globally, combined with the memory of the crisis experienced in the last decade, I believe it is important that policymakers recognise that there are potential costs to inaction resulting from a lack of perfect data or perfect models.⁸

 Given that our frameworks are still emerging and thinking is evolving constantly, I would encourage all of us in the central banking community to take a broad and flexible approach to the issue of the macro-prudential evidence base.

There may well not exist one unified quantitative model that can incorporate all costs, benefits and side-effects. Rather, it may well be necessary to take sector-specific, partial-equilibrium, empirical work on the one hand, and combine it with more aggregate, general equilibrium approaches on the other, and collate lessons from the combined evidence base, using our judgement and discretion, to arrive at a decision.

***

Turning to my second point, the range of approaches that can be placed under the umbrella of a macro-prudential “evidence base”, let me outline four analytical approaches that can be used in combination by central banks to help motivate macro-prudential policy implementation and evaluation.

First, cross-country empirical research allows for the examination of responses at a macro financial and a bank level to previous use of macro-prudential policies across both advanced and developing economies.

While not always conclusive, this literature has generally shown that the use of macro-prudential policies can affect financial sector resilience and the pro-cyclicality of credit.

This literature also highlights the role of a country’s structural characteristics, such as the openness of its economy and the extent of its financial development, and indeed its stage in the credit cycle in determining responses to macro-prudential policies.⁹

Second, country-specific empirical evidence using within-country micro data is helpful for the effective calibration of macro-prudential instruments. At the heart of this approach lies the issue of granular, timely and high-quality data.

At the Central Bank of Ireland, one of the key lessons for us as an institution when moving from the crisis to the recovery phase was the importance of such data. Starting from a detailed stress-testing exercise in 2011, our staff have gathered data on the majority of loans outstanding on Irish banks’ mortgage, SME, Corporate and Commercial Real Estate portfolios every six months since.

I cannot overstate the importance of this exercise to you. It has been essential to building credibility, to the provision of relevant information into the Irish public discourse, as well as to our ability to have the most relevant and timely information available to help motivate policy decisions.

Without as clear a picture as possible of the state of play within one’s own financial system, it is difficult to build credibility and trust which are essential ingredients to the introduction of a macro-prudential instrument.

In our own case, a prominent direct example came with the introduction of borrower-based measures (BBMs) in the mortgage market, namely the introduction of loan to value (LTV) and loan to income (LTI) limits on new lending in 2015.

Using loan-level data, our staff were able to clearly communicate to the public that looser originating credit conditions were strongly associated with subsequent mortgage default and interactions between LTV and LTI ratios on the one hand and mortgage default and loss given default on the other.¹⁰

Following on from this initial communication with the public, which established that originating LTV and LTI were plausible sources of financial instability, our staff established that the relationship between these credit conditions and loan default appeared to be different for First Time Buyers versus those purchasing a second or subsequent home with a mortgage.¹¹ This evidence base helped us to motivate a differentiated treatment for these borrower groups when introducing the policy in February 2015.

The use of granular data does not however end at the implementation phase.

As part of compliance with the new borrower-based measures, banks must submit to the Bank a six-monthly return profiling all new mortgages issued. Using these returns, our staff carry out comprehensive regular timely analysis of current new lending developments. The results of these exercises are communicated to the public and form part of an annual review process of the impact and effectiveness of the measures, which last year led to changes to some of the parameters in the restrictions.¹²

We believe this annual review process is important and is in line with the quarterly and annual reviews of counter-cyclical and O-SII buffers, respectively.  With the review processes, we exhibit clearly to the public our commitment to remain as informed as possible about the range of effects, both direct and indirect, that our policies may be having.

However, a characteristic of this approach is that results may reflect country specific characteristics (demographics, homeownership, banking market structure, etc.) and as such may not be generalisable.

Third, moving on from micro approaches, there is also room in the macro-prudential evidence base for empirical macroeconomic work. These will often take the form of Vector Autoregression (VAR) and have the great advantage that they allow for feedback loops between aggregate credit market variables and other economic aggregates such as economic growth, construction activity, house prices and unemployment to be modelled. Such feedback loops are an important part of our understanding of the overall effects of our policies in the wider economy, and are a crucial complement to the detailed understanding of within-sector effects which can be gleaned from micro data analysis.

Fourth, and finally, model-based simulations based on approaches such as the DSGE must also be taken into account. Like the aforementioned empirical macro approaches, they have the distinct advantage that they explicitly take account of feedback loops between credit markets and other sectors of the economy, giving us a general equilibrium understanding of how agents across the economy will react to our policies. Given their micro-founded nature, they also alleviate concerns related to the Lucas Critique, which are harder to brush off when running policy simulations using parameters estimated from any of the aforementioned approaches.¹³

Let me summarise up to this point by saying that I do not believe there to be one “first best” analytical approach underpinning macro-prudential policy implementation or evaluation.

Referring back to the Greek poet Archilochus, it is incumbent on us to behave more like the fox, who must know many things and take inspiration from a disparate range of sources, than the hedgehog, who knows one big truth.

We must also be cognisant of Popper’s warning that “the way of science is paved with discarded theories which were once declared self-evident.”¹⁴

 ***

I will close my speech with an overview of the areas in which progress is likely to be concentrated in the coming years.

Our achievements as macro-prudential policymakers up to this point have been wide-ranging and numerous. Thought leadership has come from within the Central Banking community, from academia, from the wider research community, as well as from private industry. There remains progress to be made in a number of areas. I would encourage all of you attending this conference to bear these areas in mind as you develop your own research agendas.

A more systematic approach to the definition and communication of the costs and benefits of macro-prudential policies may be necessary. This boils down to the definition of a type of “loss function”, which may not have to take on a formal quantitative form, but which must be informing our decision making process. This may help answer questions such as:

• What are the benefits we envisage arising from policy implementation?
• To what extent is the policy likely to hamper economic activity?
• Which sectors of economy or financial system are likely to experience side-effects?
• How high is the likelihood that the policy will “leak”, with credit being extended in unregulated parts of the financial system?
• And, how do we place weights on each of these components as policy makers?

I believe we will be doing a stronger job in serving the public when the answers to these questions form a more systematic part of our communications with the public in this sphere.

We must continue to provide the highest-quality ongoing monitoring and evaluation of our policies, and commit to allow our “opinions to change as the facts change”. Such an ongoing rigorous approach will help to build public trust, and allow for policy tweaking where necessary.

We are also likely to benefit from the “wisdom of crowds” by expanding the range of researchers working on this topic. I would encourage us all to ensure that we are doing what we can to allow for collaboration between our own central banks and each of our domestic research communities. The wider the range of voices being heard, the more likely we are to arrive at better outcomes. In this light, the initiative that we are celebrating today, the provision of a publicly available database on macro-prudential instruments, is particularly welcome.

------------------------------------------------------- 

¹ I would like to thank Mark Cassidy, Fergal McCann, Mícheál O’Keefe and Ellen Ryan for their contribution to my remarks.

² Popper, Karl R. (1994), The Myth of the Framework: In defence of science and rationality, Routledge: London.

³ Ibid.

⁴ Although the term "macro-prudential" only surfaced in public documents in the mid 1980’s, throughout much of the twentieth century, and particularly during the post-war period, there were numerous examples of central bank interventions in the financial system that can, in retrospect, be considered "macro-prudential”. See ‘Lessons from the past, safeguarding stability for the future’, address by Sharon Donnery, at the Centre for Economic Policy Research (CEPR) Economic History Symposium, Dublin, 9 June 2016. Available here.

⁵ See for example:

• Antonio Garcia Pascual, Tomasz Wieladek (2016), “The European Central Bank's QE: A new hope”. Centre for Economic Policy Reform DP11309.
• Luca Gambetti, Alberto Musso (2017), “The macroeconomic impact of the ECB's expanded asset purchase programme (APP)”. European Central Bank Working Paper No. 2075.
• Matteo Ciccarelli, Juan Angel Garcia, Carlos Montes-Galdón (2017), “Unconventional monetary policy and the anchoring of inflation expectations”, European Central Bank Working Paper No. 1995.
• Christiane Baumeister and Luca Benati (2017), “Unconventional Monetary Policy and the Great Recession: Estimating the Macroeconomic Effects of a Spread Compression at the Zero Lower Bound”, International Journal of Central Banking, vol. 9(2), pages 165-212

⁶ See ‘Communication, Calibration and Coordination: Challenges Implementing Macro-prudential Policy in the Euro Area’, speech by Sharon Donnery at the second annual ECB macro-prudential policy and research conference, 12 May 2017. Available here.

⁷ For further discussion see Committee on the Global Financial System (2016), “Objective-setting and communication of macro-prudential policies”, CGFS Papers No. 57. CFGS (2016) (PDF 359.44KB).

⁸ See ‘Macro-prudential policy: action in the face of uncertainty’, address by Sharon Donnery, at the Dublin Economic Workshop Annual Economic Policy Conference 24 Sep 2016. Available here.

⁹ For example see:

• International Monetary Fund, 2011. "Macro-prudential Policy; What Instruments and How to Use them? Lessons From Country Experiences," IMF Working Papers 11/238, International Monetary Fund
• Cerutti, Eugenio, Claessens, Stijn and Laeven, Luc, (2015), The Use and Effectiveness of Macro-prudential Policies; New Evidence, No 15/61, IMF Working Papers, International Monetary Fund
• Claessens, Stijn & Ghosh, Swati R. & Mihet, Roxana, 2013. "Macro-prudential policies to mitigate financial system vulnerabilities," Journal of International Money and Finance, Elsevier, vol. 39(C), pages 153-185.
• Codruta Boar, Leonardo Gambacorta, Giovanni Lombardo and Luiz Awazu Pereira da Silva (2017), “What are the effects of macro-prudential policies on macroeconomic performance?” BIS Quarterly Review, September 2017.

Available here: Lim et al. (2011); Cerutti, Claessens and Laeven (2015); Claessens, Ghosh and Mihet (2014); and Boar et al (2017).

¹⁰ Hallissey, Niamh et al. (2014). Economic Letter, Central Bank of Ireland. Macro-prudential Tools and Credit Risk of Property Lending at Irish banks. Available here (PDF 455.84KB).

¹¹ Kelly, Robert et al. (2014) Economic Letter, Central Bank of Ireland. Do first time buyers default less? Implications for macro-prudential policy. Also Kelly, Robert et al. (2015) Research Technical Paper, Central Bank of Ireland. Designing Macro-prudential Policy in Mortgage Lending: Do First Time Buyers Default Less? Available here (PDF 505.47KB) and here (PDF 1.53MB).

¹² Kinghan, Christina et al. (2016, 17). Economic Letter, Central Bank of Ireland. Macro-prudential Measures and Irish Mortgage Lending: Insights from H1 2016. Macro-prudential Measures and Irish Mortgage Lending: Insights from H1 2017. Available here (PDF 1.76MB) and here (PDF 15.12MB). 

¹³ For further detail see:

• Jaromir Benes & Douglas Laxton & Joannes Mongardini, 2016 "Mitigating the Deadly Embrace in Financial Cycles; Countercyclical Buffers and Loan-to-Value Limits," IMF Working Papers 16/87, International Monetary Fund.
• Lozej, Matija & Rannenberg, Ansgar, 2017. "The macro-economic effects of the regulatory LTV and LTI ratios in the Central Bank of Ireland's DSGE model," Economic Letter, Central Bank of Ireland. Available here (PDF 469.63KB).
• Lozej, Matija & Onorante, Luca & Rannenberg, Ansgar, 2017. "Countercyclical Capital Regulation in a Small Open Economy DSGE Model," Research Technical Papers, Central Bank of Ireland. Available here (PDF 1.4MB).

¹⁴ See Popper, Karl. R.. (2013, p232), The Open Society and Its Enemies, Princeton University Press.