“The rate of return on everything”

This is the focus of a paper titled “The Rate of Return on Everything, 1870-2015 that seeks to address some fundamental questions that underpin, not only economic theory, but also investment strategy.

To quote the abstract:

This paper answers fundamental questions that have preoccupied modern economic thought since the 18th century. What is the aggregate real rate of return in the economy? Is it higher than the growth rate of the economy and, if so, by how much? Is there a tendency for returns to fall in the long-run? Which particular assets have the highest long-run returns? We answer these questions on the basis of a new and comprehensive dataset for all major asset classes, including—for the first time—total returns to the largest, but oft ignored, component of household wealth, housing. The annual data on total returns for equity, housing, bonds, and bills cover 16 advanced economies from1870 to 2015, and our new evidence reveals many new insights and puzzles.ets.

“The Rate of Return on Everything” Òscar Jordà, Katharina Knoll, Dmitry Kuvshinov, Moritz Schularick, Alan M. Taylor – December 2017

The paper is roughly 50 pages long (excluding appendices) but the 5 page introduction summarises the four main findings which I have further summarised below:

  1. Risky Returns: The study finds that “… residential real estate and equities have shown very similar and high real total gains, on average about 7% per year. Housing outperformed equity before WW2. Since WW2, equities have outperformed housing on average, but only at the cost of much higher volatility and higher synchronicity with the business cycle”.
  2. Safe Returns: The study finds that “Safe returns have been low on average, falling in the 1%–3% range for most countries and peacetime periods“. However, “the real safe asset return has been very volatile over the long-run, more so than one might expect, and oftentimes even more volatile than real risky returns.” This offers a long-run perspective on the current low level of the safe returns with the authors observing that “… it may be fair to characterize the real safe rate as normally fluctuating around the levels that we see today, so that today’s level is not so unusual. Which begs the question “… why was the safe rate so high in the mid-1980s rather than why has it declined ever since.”
  3. The Risk Premium: The study finds that the risk premium has been very volatile over the long run. The risk premium has tended to revert to about 4%-5% but there have been periods in which it has been higher. The study finds that the increases in the risk premium “… were mostly a phenomenon of collapsing safe rates rather than dramatic spikes in risky rates. In fact, the risky rate has often been smoother and more stable than safe rates, averaging about 6%–8% across all eras” . This for me was one of the more interesting pieces of data to emerge from the study and has implications for the question of what should be happening to target return on equity in a low interest rate environment such as we are currently experiencing. In the Authors’ words “Whether due to shifts in risk aversion or other phenomena, the fact that safe rates seem to absorb almost all of these adjustments seems like a puzzle in need of further exploration and explanation
  4. On Returns Minus Growth: This is the question that Thomas Piketty explored in his book “Capital in the Twenty-First Century”. Piketty argued that, if the return to capital exceeded the rate of economic growth, rentiers would accumulate wealth at a faster rate and thus worsen wealth inequality. The study finds that, “for more countries and more years, the rate of return on risk assets does in fact materially exceed the rate of growth in GDP… In fact, the only exceptions to that rule happen in very special periods: the years in or right around wartime. In peacetime, r has always been much greater than g. In the pre-WW2 period, this gap was on average 5% per annum (excluding WW1). As of today, this gap is still quite large, in the range of 3%–4%, and it narrowed to 2% during the 1970s oil crises before widening in the years leading up to the Global Financial Crisis.

So why does this matter?

There is a lot to think about in this paper depending on your particular areas of interest.

The finding that the long run return on residential housing is on par with equity but lower volatility is intriguing though I must confess that I want to have a closer look at the data and methodology before I take the conclusion as a fact. In particular, I think it is worth paying close attention to the way that the study deals with taxation. Fortunately, the paper offers a great deal of detail on the way that residential property is taxed (Appendix M in the December 2017 version of the paper) in different countries which is useful in its own right. I have been looking for a source that collates this information for some time and this is the best I have seen so far.

For me at least, the data on how the Equity Risk Premium (ERP) expands and contracts to offset changes in the return unsafe assets was especially interesting. This observation about the relationship is not new of itself but it was useful to find more data in support of it. I have been thinking a lot about Cost of Equity in a low interest rate environment and this seems to support the thesis that the target Return on Equity (ROE) should not necessarily be based on simply adding a fixed measure of the ERP (say 4%-5%) to whatever the current long run risk free rate is. It is at least worth having the question in mind when considering the question of whether Australian bank ROE is excessive at this point of the cycle.

If you are interested in the issues covered above then it is also worth having a look at an RBA Research Discussion Paper titled “A History of Australian Equities” by Thomas Matthews that was published this month.

From The Outside

The financial cycle and macroeconomics: What have we learnt? BIS Working Paper

Claudio Borio at the BIS wrote an interesting paper exploring the “financial cycle”. This post seeks to summarise the key points of the paper and draw out some implications for bank stress testing (the original paper can be found here).  The paper was published in December 2012, so its discussion of the implications for macroeconomic modelling may be dated but I believe it continues to have some useful insights for the challenges banks face in dealing with adverse economic conditions and the boundary between risk and uncertainty.

Key observations Borio makes regarding the Financial Cycle

The concept of a “business cycle”, in the sense of there being a regular occurrence of peaks and troughs in business activity, is widely known but the concept of a “financial cycle” is a distinct variation on this theme that is possibly less well understood. Borio states that there is no consensus definition but he uses the term to

“denote self-reinforcing interactions between perceptions of value and risk, attitudes towards risk and financing constraints, which translate into booms followed by busts. These interactions can amplify economic fluctuations and possibly lead to serious financial distress and economic disruption”.

This definition is closely related to the concept of “procyclicality” in the financial system and should not be confused with a generic description of cycles in economic activity and asset prices. Borio does not use these words but I have seen the term “balance sheet recession” employed to describe much the same phenomenon as Borio’s financial cycle.

Borio identifies five features that describe the Financial Cycle

  1. It is best captured by the joint behaviour of credit and property prices – these variables tend to closely co-vary, especially at low frequencies, reflecting the importance of credit in the financing of construction and the purchase of property.
  2. It is much longer, and has a much larger amplitude, than the traditional business cycle – the business cycle involves frequencies from 1 to 8 years whereas the average length of the financial cycle is longer; Borio cites a cycle length of 16 years in a study of seven industrialised economies and I have seen other studies indicating a longer cycle (with more severe impacts).
  3. It is closely associated with systemic banking crises which tend to occur close to its peak.
  4. It permits the identification of the risks of future financial crises in real time and with a good lead – Borio states that the most promising leading indicators of financial crises are based on simultaneous positive deviations of the ratio of private sector credit-to-GDP and asset prices, especially property prices, from historical norms.
  5. And it is highly dependent of the financial, monetary and real-economy policy regimes in place (e.g. financial liberalisation under Basel II, monetary policy focussed primarily on inflation targeting and globalisation in the real economy).

Macro economic modelling

Borio also argues that the conventional models used to analyse the economy are deficient because they do not capture the dynamics of the financial cycle. These extracts capture the main points of his critique:

“The notion… of financial booms followed by busts, actually predates the much more common and influential one of the business cycle …. But for most of the postwar period it fell out of favour. It featured, more or less prominently, only in the accounts of economists outside the mainstream (eg, Minsky (1982) and Kindleberger (2000)). Indeed, financial factors in general progressively disappeared from macroeconomists’ radar screen. Finance came to be seen effectively as a veil – a factor that, as a first approximation, could be ignored when seeking to understand business fluctuations … And when included at all, it would at most enhance the persistence of the impact of economic shocks that buffet the economy, delaying slightly its natural return to the steady state …”

“Economists are now trying hard to incorporate financial factors into standard macroeconomic models. However, the prevailing, in fact almost exclusive, strategy is a conservative one. It is to graft additional so-called financial “frictions” on otherwise fully well behaved equilibrium macroeconomic models, built on real-business-cycle foundations and augmented with nominal rigidities. The approach is firmly anchored in the New Keynesian Dynamic Stochastic General Equilibrium (DSGE) paradigm.”

“The purpose of this essay is to summarise what we think we have learnt about the financial cycle over the last ten years or so in order to identify the most promising way forward…. The main thesis is that …it is simply not possible to understand business fluctuations and their policy challenges without understanding the financial cycle”

There is an interesting discussion of the public policy (i.e. prudential, fiscal, monetary) associated with recognising the role of the financial cycle but I will focus on what implications this may have for bank management in general and stress testing in particular.

Insights and questions we can derive from the paper

The observation that financial crises are based on simultaneous positive deviations of the ratio of private sector credit-to-GDP and asset prices, especially property prices, from historical norms covers much the same ground as the Basel Committee’s Countercyclical Capital Buffer (CCyB) and is something banks would already monitor as part of the ICAAP. The interesting question the paper poses for me is the extent to which stress testing (and ICAAP) should focus on a “financial cycle” style disruption as opposed to a business cycle event. Even more interesting is the question of whether the higher severity of the financial cycle is simply an exogenous random variable or an endogenous factor that can be attributed to excessive credit growth. 

I think this matters because it has implications for how banks calibrate their overall risk appetite. The severity of the downturns employed in stress testing has in my experience gradually increased over successive iterations. My recollection is that this has partly been a response to prudential stress tests which were more severe in some respects than might have been determined internally. In the absence of any objective absolute measure of what was severe, it probably made sense to turn up the dial on severity in places to align as far as possible the internal benchmark scenarios with prudential benchmarks such as the “Common Scenario” APRA employs.

At the risk of a gross over simplification, I think that banks started the stress testing process looking at both moderate downturns (e.g. 7-10 year frequency and relatively short duration) and severe recessions (say a 25 year cycle though still relatively short duration downturn). Bank supervisors  in contrast have tended to focus more on severe recession and financial cycle style severity scenarios with more extended durations. Banks’s have progressively shifted their attention to scenarios that are more closely aligned to the severe recession assumed by supervisors in part because moderate recessions tend to be fairly manageable from a capital management perspective.

Why does the distinction between the business cycle and the financial cycle matter?

Business cycle fluctuations (in stress testing terms a “moderate recession”) are arguably an inherent feature of the economy that occur largely independently of the business strategy and risk appetite choices that banks make. However, Borio’s analysis suggests that the decisions that banks make (in particular the rate of growth in credit relative to growth in GDP and the extent to which the extension of bank credit contributes to inflated asset values) do contribute to the risk (i.e. probability, severity and duration) of a severe financial cycle style recession. 

Borio’s analysis also offers a way of thinking about the nature of the recovery from a recession. A moderate business cycle style recession is typically assumed to be short with a relatively quick recovery whereas financial cycle style recessions typically persist for some time. The more drawn out recovery from a financial cycle style recession can be explained by the need for borrowers to deleverage and repair their balance sheets as part of the process of addressing the structural imbalances that caused the downturn.

If the observations above are true, then they suggest a few things to consider:

  • should banks explore a more dynamic approach to risk appetite limits that incorporated the metrics identified by Borio (and also used in the calibration of the CCyB) so that the level of risk they are willing to take adjusts for where they believe they are in the state of the cycle (and which kind of cycle we are in)
  • how should banks think about these more severe financial cycle losses? Their measure of Expected Loss should clearly incorporate the losses expected from business cycle style moderate recessions occurring once every 7-10 years but it is less clear that the kinds of more severe and drawn out losses expected under a Severe Recession or Financial Cycle downturn should be part of Expected Loss.

A more dynamic approach to risk appetite get us into some interesting game theory  puzzles because a decision by one bank to pull back on risk appetite potentially allows competitors to benefit by writing more business and potentially doubly benefiting to the extent that the decision to pull back makes it safer for competitors to write the business without fear of a severe recession (in technical economist speak we have a “collective action” problem). This was similar to the problem APRA faced when it decided to impose “speed limits” on certain types of lending in 2017. The Royal Commission was not especially sympathetic to the strategic bind banks face but I suspect that APRA understand the problem.

How do shareholders think about these business and financial cycle losses? Some investors will adopt a “risk on-risk off” approach in which they attempt to predict the downturn and trade in and out based on that view, other “buy and hold” investors (especially retail) may be unable or unwilling to adopt a trading approach.

The dependence of the financial cycle on the fiscal and monetary policy regimes in place and changes in the real-economy also has potential implications for how banks think about the risk of adverse scenarios playing out. Many of the factors that Borio argues have contributed to the financial cycle (i.e. financial liberalisation under Basel II, monetary policy focussed primarily on inflation targeting and globalisation in the real economy) are reversing (regulation of banks is much more restrictive, monetary policy appears to have recognised the limitations of a narrow inflation target focus and the pace of globalisation appears to be slowing in response to a growing concern that its benefits are not shared equitably). I am not sure exactly what these changes mean other than to recognise that they should in principle have some impact. At a minimum it seems that the pace of credit expansion might be slower in the coming decades than it has in the past 30 years.

All in all, I find myself regularly revisiting this paper, referring to it or employing the distinction between the business and financial cycle. I would recommend it to anyone interested in bank capital management.