Speech Future Directions in Financial Stability Analysis: Learning from Others, Learning from the Past

My thanks, once again, go to the Paul Woolley Centre for the opportunity to participate in this conference. I have had the privilege of speaking at all four of the Centre's annual conferences. My previous three speeches were very much from the perspective of someone from a policymaking institution – on how to detect risks to financial stability, how to respond, and what mix of skills you might need to do so. With this being the last conference of this type at the Centre, I thought it would make sense to talk about future directions for research. This is, after all, an academic conference.

Before I get to that, I'd like to make a couple of comments about recent developments. The Reserve Bank has been closely monitoring emerging risks in property markets for some time now. At least some of the growth in housing prices since the middle of last year was a bit of a catch-up from a period of weakness. Some increase in prices was to be expected given the current level of interest rates. And housing prices are still within the same range, relative to incomes, that we have seen over the past decade, albeit towards the top of that range. But if you look beyond the national aggregates and averages, risks do seem to have been building in some parts of the housing market, specifically the investor segments of the Sydney and Melbourne markets.

Currently close to half of all net new housing finance is extended to self-declared investors rather than owner-occupiers. That share is noticeably higher than rental housing's share of the housing stock, even allowing for a possible faster rate of churn in investor loans. Obviously that can't continue forever. We raised the issue over a year ago in the September 2013 Financial Stability Review; it is a reason why, in the latest Financial Stability Review, we described the composition of the housing and mortgage markets as becoming unbalanced.

Part of the anticipated effect of monetary policy is to induce more construction activity. Higher prices are the incentive to get that expansion, which is indeed happening. But it is worth noting that the vast bulk of that new borrowing is to purchase existing properties (Graph 1).

We talked about these issues at some length in the Financial Stability Review last month, and last week before the Senate. I do not have anything new to add today to what we have already said in the past few weeks. We obviously need to keep a close eye on market developments, including lending standards. Amongst other things, banks and other lenders need to consider the risks they are taking on, not just from individual loans, but from the collective effects of lending decisions on the system as a whole.

Graph 1
Graph 1: Lending to Investors for New Construction

APRA is the regulator and supervisor of banks and other deposit-taking businesses in Australia. It has already been turning up the dial in its supervisory activities in this area, though this might not have been obvious to observers outside those regulated institutions. As has already been described publicly, APRA is consulting with the Reserve Bank and with other members of the Council of Financial Regulators, about additional steps that could be taken to reinforce sound lending practices. In doing so, it will balance the advantages and disadvantages in the context of financial system stability, safety, and efficiency, and it will consider how those measures can best be targeted.

Everything I have mentioned so far today is consistent with the existing framework for financial stability policy in Australia: APRA has the mandate to use its powers to promote financial stability; it can take a holistic approach to risk; it can and does coordinate with other regulators, under the auspices of the Council of Financial Regulators. So what I am describing is very much in keeping with the ‘state of mind’ of macroprudential policy that I described in my speech here in 2012.

Authorities in many countries have looked to prudential and supervisory measures as part of the policy response to financial risk. Different countries have designed their responses differently, according to the institutional arrangements they have and the specific risks they face. In doing so, they have found relatively little existing theory to draw upon in the academic literature. That's where you come in, and broadly speaking it's the topic of the remainder of my talk today.

A Crisis for Theory?

It has been said that the global financial crisis was also a crisis for macroeconomic theory. New models and approaches are required, so it's claimed. I don't think I can add much new to that discussion. Economists far more knowledgeable and eminent than me have had quite a bit to say on this recently.[1]

Some of the participants in that debate have argued that the problem with mainstream theory was that it assumes that people have rational expectations. I agree that we need more research into the biases that pervade human behaviour. But I'm not so sure that ‘rational expectations’ is the first assumption we should throw out of the standard models. The standard models that were being taught in graduate schools – those coming under the umbrella label of ‘DSGE’ models – didn't just assume that people were rational and forward-looking.[2] Most of them also assumed that there was a representative agent, who was infinitely lived and who had full information about the structure of the economy.

Or in other words, you are the only person in the world; you are going to live forever; and you know everything. It might be a reasonable way to model an economy comprised entirely of teenagers, but I think it might have some shortcomings when analysing the economy we actually have. More seriously though, without some diversity of actors in the model, some diversity of beliefs, there are no financial markets, no borrowing and lending transactions.

It is true that many mainstream macro models did not contain financial sectors, or admit the possibility of financial crises. They were built to answer different questions – and that's ok! No single model can do everything.

This does not mean that policymakers were unaware of the importance of financial factors. Some of the most senior policymakers had been authorities on financial frictions and crises in their academic lives.[3] And there were plenty of recent crises to remind us that crises matter. In fact, in the lead-up to the crisis, two of the most frequent policy debates were, first, what to do about so-called ‘global imbalances’ and second, whether and how monetary policy should respond to asset prices.

These were not just conversations among policymakers, either. Plenty of serious academic work was published on these topics during the pre-crisis period, often using more or less mainstream models (e.g. Bordo and Jeanne (2002) and Kiyotaki and Moore (1997)). Many of these papers were written by distinguished academics and published in top journals. These papers had welcoming audiences. If you wanted to work on these topics, nobody was stopping you. I've read some reports of sniggering in seminar rooms, when people presented papers outside the mainstream models. Let me assure you that good papers on financial stability topics had interest and respectful audiences at central banks and international agencies long before the crisis.

So if the core mainstream models couldn't be used to investigate financial crises, it wasn't because they were inherently limited. It was because (like Jessica Rabbit) they were drawn that way. That was a choice of the modellers. It comes down to the questions they were trying to answer. So perhaps the real issue is not so much whether our models and theory needed to change, but why mainstream macro theory was not more interested in financial crises and financial stability issues generally.

Nowadays, of course, there is more interest in the subject. Whatever the methods to be used, this is an incredibly exciting time to be researching topics related to financial stability. Indeed, never before have there been so many interesting questions to pursue, yet so little time for us policymakers to pursue them ourselves.

Even more exciting is that it's a game that many people can play, not just economics and finance types. For example, I think there is a lot that sociologists and organisational psychologists could have to say about why the major global banks take the risks that they do. And because we know that so much of the financial stability policy framework centres on communicating and warning about risk, I'd love to hear from academics in the field of communications, about whether we could do it better.

Learning from the Past

So how should academics interested in financial stability topics approach the subject? Some observers have suggested that what we need is better models of asset price bubbles (Allen and Carletti 2013). Honestly, I'm not so sure that this is the first priority. I just don't see asset markets as having an on/off state – bubble or no bubble. There will always be some speculators in any asset market, more at some times than others. And even in the hottest market, there will always be some people buying for completely non-speculative reasons based on expected dividends or the suitability of the home, or whatever other ‘fundamentals’ are important to them in the context of that particular asset. This kind of diversity is particularly relevant in property markets, where every property is different in some way. Even if prices seem overinflated in general, there can always be a well-priced property somewhere.

On a more practical level, as I've explained elsewhere previously, asset prices can go through boom-bust cycles that seem completely in line with fundamentals like rents, if those fundamentals are also cyclical. Busts of this type can be just as painful as the bubbly sort. Another limitation of a bubbles-focused approach is that your estimate of the bubble is only as good as your model of fundamentals. Unfortunately, many models of fundamentals are quite limited. In particular, many of them implicitly assume the same single-agent structure that is so problematic for macroeconomic models.

Perhaps a more productive approach might be to think of asset markets as having many different type of trader operating in them, some more speculative in their intentions than others. Models with different types of trader have a longstanding place in the literature.[4] I think there is more to be done in this vein.

Rather than focusing on bubbles specifically, I'd like to step back and suggest a couple of guidelines that I think are likely to be important for staying on the right track when investigating financial stability issues.

First, get your stocks and flows straight. I believe that no model will capture financial stability dynamics unless it incorporates stocks as well as the flows into and out of them. The obvious case is a balance sheet. If you are interested in a topic related to financial stability, tracking flows of income and expenditure isn't enough. You must also track the underlying balance sheets – the debt, the liquidity position, leverage and so on. That is not a new insight. The role of balance sheets, and especially debt, was a central theme of work by Irving Fisher, Wynne Godley and many others.

I'd go further than that and say that balance sheets are not the only stock that matters. Physical stocks – of equipment, of real estate, or of particular types of capital goods such as ships – can also be very important. Long-lived physical assets can experience large price cycles because their supply – a stock – is inherently sluggish. Speculative bubble dynamics are actually not required for a painful boom-bust cycle in asset prices.

An example might help illustrate this point. The recent financial crisis was sparked by a meltdown in the US mortgage market. Mortgage defaults started rising rapidly, long before unemployment did. Many countries had booms in housing prices before the crisis. But only a few had large subsequent increases in mortgage defaults. And only in the United States did defaults rise dramatically before unemployment rose significantly. Why was that? As I've discussed elsewhere, a lot of factors contributed, not least the utter breakdown in lending standards in the US mortgage market (Ellis 2010). But also crucial was the overbuilding, the build-up of a supply overhang.

A good indicator of this is the US Census Bureau's series on the share of non-rental housing that is both vacant and for sale – the owner-occupier vacancy rate (Graph 2). You can think of it as a kind of unemployment rate for houses. The vertical line shows the December 2006 release, which was published in early 2007, before the financial part of the crisis had started.

Over more than half a century, this vacancy rate had never before risen so high or so quickly. It should be clear from this that the US housing market was in a different position to Australia's, or the United Kingdom's, or many other countries that had seen strong growth in housing prices up to that point. Because it had a supply overhang, the United States was set for a larger downturn than those other countries.[5] Looking at the flow of building work wasn't enough to see the problem. You also had to look at the resulting stocks.[6] Unfortunately a lot of people missed this point. That might have been partly because they weren't looking across different countries' experiences.

Graph 2
Graph 2: US Homeowner Vacancy Rate

Second, allow for incomplete information, uncertainty and thus default. Incomplete information is especially important when some people have information that others do not. From these asymmetries spring agency problems of the type that seem endemic in finance. There is already a substantial body of literature on these issues.[7] I expect that it will remain a fruitful area for research, particularly when one considers issues of organisational culture, fraud and so on.

The specific possibility of default on a loan is a fundamental building block in any investigation of financial stability issues: people who borrow cannot know for sure that they will be able to repay the debt. They might expect and plan to do so, but some unexpected bad event might prevent it. Neither borrower nor lender knows for sure who will default. While lenders might estimate a probability of default, they cannot observe it.[8]

Uncertainty about default induces lenders to impose constraints on how much they will lend to a particular borrower. Again, this is all in much older and relatively (though in some cases not entirely) mainstream literature. Some of these papers focused on companies as customers of the banking system (Bernanke and Gertler 1989); others applied similar ideas to the banks themselves, allowing for the possibility of bank runs (Diamond and Dybvig 1983).

So there are credit constraints. They take different forms but often boil down to working out how large a repayment the borrower can be expected to service. Credit constraints of this type can have macro consequences, and in particular they affect how we should interpret movements in certain macro-level quantities. If average interest rates fall, the loan amount that corresponds to that repayment is larger. Australia went from being a high-inflation country to a low-inflation country in the early 1990s. One result of that was that average nominal interest rates fell, and so the sustainable amount of debt rose relative to income – permanently. People should therefore not expect ratios of housing prices or debt to income to revert to their long-run, multi-decade averages. We have discussed this conclusion numerous times over the past 20 years.[9] Obviously if we are wrong about this, we would love to know about it. So far, nobody has come up with a counterargument to this idea, let alone a compelling one.

Instead we see some people confidently opining that property prices are however many per cent overvalued. What is the model behind these statements? As far as I can tell, there is never a real model, only an assumption that if the national ratio of prices to incomes or rents is however many percentage points away from some multi-decade long-run average, it must revert to that average. Yet there is no theory that says things should work that way. Credit constraints matter, and if those constraints should change, say because of regulatory change or disinflation, then the sustainable level of these kinds of ratios will also shift.[10]

Learning from Others

I feel a little as though I have spent most of today's talk citing all the well-known seminal papers in this field. And they are the well-known seminal papers. The point I am making is that none of this is completely new. The profession as a whole was not ignoring financial stability issues, even if the whole profession wasn't focused on the topic.

Certainly some of the authors I've cited in the footnotes to this talk might have felt themselves to be outside of the mainstream of macroeconomics as it is taught in the top graduate schools overseas. I have deliberately sought to show that these topics were being researched, were being published in top journals, and were getting the necessary attention from many policymakers.

That said, it is a collection of work that did not make it into those canonical macro models. More needs to be done on this front. Some of the most relevant ideas were coming from top academics such as the ones I've cited today, who could hardly be said to be outside the core of the profession. Certainly there are other important insights coming from those who aren't big names. And there are some quite big names whose influences are somewhat outside the graduate school canon.[11] There is much to be learned from all those voices.

Beyond the diverse voices in the fields of economics and finance, I think there is a lot to be learned from a different kind of ‘other’: researchers in other fields entirely. Many of the issues relevant for financial stability analysis are quite difficult to tackle using standard economics models. It is hard enough to model a system with diverse kinds of atoms; it's even harder to model a system with diverse kinds of humans, who think ahead and plan and mould their behaviour in response to what they think the other guy is going to do.

There are so many interesting ideas in other fields. For example, I already mentioned the problems of assuming a single type of agent in the system. Recently, researchers in central banks and elsewhere have been using network models to obtain some useful insights on the structure of particular financial markets and relationships. Inherently, this means recognising that you can't treat each market participant as being interchangeable with the others. Each of them will have a different set of relationships, and therefore pose a different set of risks to the others and to the system. Networks have a long history in the mathematical literature, but they are relatively new tools in economics and finance.

Another interesting angle comes from so-called agent-based modelling, which really aggregates up from the individual. The overall behaviour develops organically, rather than assuming that the micro represents the macro.

When I titled this talk ‘learning from others’, though, I really did mean ‘learning’. Economics as a field is particularly known for co-opting techniques from other disciplines. There is econophysics, ideas from evolutionary biology, ideas from geography. There are people looking at networks, searching for power laws, and many other examples. But economics and finance will remain the fields with the most to say about financial stability. As we adopt these models and techniques, we have to be aware of their strengths and weaknesses, and be mindful of using them in ways that are not misleading. We have to engage with the expertise in those other fields, and learn from them.

One of the great things about universities is that those multidisciplinary conversations can happen right here on campus. Unlike at central banks and international financial institutions, where almost everyone doing research comes from some variety of economics or finance training, at universities, there are experts in those other fields right down the hall or in the next building. Not all of their approaches and techniques will be appropriate for our purposes. We are much more likely to be able to tell the difference between the useful transplants and the less useful ones if we engage with the originating fields. Be wary of just reading a paper or two and thinking that you understand a whole field or technique.

So if I am to articulate a vision of what research in financial stability topics could look like, it would be one of both continuity and change. I anticipate that many of the most useful ideas will be developed from existing ideas, especially those on asymmetric information and uncertainty. I also anticipate that some techniques adapted from other fields will prove fruitful, though there might be others that end up being dead ends for our purposes, no matter how useful they are in their original fields. And I anticipate that, try as I might, there will still be more interesting papers than policymakers like me have time to read. For that, I apologise. I will try, though, because I'm looking forward to seeing what you come up with.

Thank you for your time.


Thanks to Fiona Price for her assistance in preparing this talk. [*]

Most recently Blanchard (2014) and Stiglitz (2014). [1]

DSGE stands for ‘dynamic stochastic general equilibrium’, though in practice it refers to a particular subset of models that involve time paths (dynamic) and random shocks (stochastic), and where things add up with no production or income entering or leaving the system from unexplained sources (general equilibrium). [2]

Most obviously Ben Bernanke – see Bernanke and Gertler (1989). And of course plenty of career policymakers were working on these topics as well (Borio and Lowe 2002). [3]

See, for example, Shliefer and Summers (1990) and De Long et al (1990). [4]

This reasoning found its way into the BIS Annual Report in 2008. [5]

This is the point made in Ellis, Kulish and Wallace (2012). [6]

Two quite separate strands of this literature are well represented by Allen and Gale (2000) and Greenwald and Stiglitz (1986). [7]

Technically this is a case of Knightian uncertainty rather than risk in the usual sense, even though default risk is usually treated as a probabilistic risk. Thank you to Penny Smith for reminding me of this point. [8]

Most recently, this was discussed in Chapter 2 of the Bank's initial submission to the Financial System Inquiry. See also Ellis (2013). [9]

This state of affairs is not helped by the data limitations in some other countries, which mean that commendable efforts by international agencies to publish comparable property price information are limited to deviations of indexes from their own long-run averages. It is not possible to compare ratios across countries using such a data set, but many commentators have misunderstood these data sets and done so. [10]

I'm thinking here of people like Axel Leijonhufvud and Roger Farmer. [11]


Allen F and D Gale (2000), ‘Bubbles and Crises’, Economic Journal, 110(460), pp 236–255.

Allen F and E Carletti (2013), ‘New Theories to Underpin Financial Reform’, Journal of Financial Stability, 9(2), pp 242–249.

Bernanke B and M Gertler (1989), ‘Agency Costs, Net Worth, and Business Fluctuations’, American Economic Review, 79(1), pp 14–31.

Blanchard O (2014), ‘Where Danger Lurks’, Finance & Development, 51(3), pp 28–31.

Bordo MD and O Jeanne (2002), ‘Monetary Policy And Asset Prices: Does ‘Benign Neglect’ Make Sense?’, International Finance, 5(2), pp 139–164.

Borio C and P Lowe (2002), ‘Asset prices, financial and monetary stability: exploring the nexus’, BIS Working Paper 114.

De Long B, A Shleifer, LH Summers and RJ Waldmann (1990), ‘Noise Trader Risk in Financial Markets’, Journal of Political Economy, 98(4), pp 703–738.

Diamond DW and PH Dybvig (1983), ‘Bank Runs, Deposit Insurance, and Liquidity’, Journal of Political Economy, 91(3), pp 401–419.

Ellis L, M Kulish and S Wallace (2012), Property Market Cycles as Paths to Financial Distress, Property Markets and Financial Stability, Proceedings of a Conference, Bank for International Settlements and Reserve Bank of Australia, Sydney.

Ellis L (2010), ‘The Housing Meltdown: Why did it Happen in the United States?’, International Real Estate Review, 13(3), pp 351–394.

Ellis L (2013), ‘Housing and Mortgage Markets: The Long Run, the Short Run and the Uncertainty in Between’, Citibank Property Conference, Sydney, 23 April. Available at <https://www.rba.gov.au/speeches/2013/sp-so-230413.html>.

Greenwald B and JE Stiglitz (1986), ‘Externalities in Economies with Imperfect Information and Incomplete Markets’, Quarterly Journal of Economics, 101(2), pp 229–264.

Kiyotaki N and J Moore (1997), ‘Credit Cycles’, Journal of Political Economy, 105(2), pp 211–248.

Shliefer A and LH Summers (1990), ‘The Noise Trader Approach to Finance’, Journal of Economic Perspectives, 4(2), pp 19–33.

Stiglitz JE (2014), ‘Reconstructing Macroeconomic Theory to Manage Economic Policy’, NBER Working Paper 20517.