No-one questions any longer the powerful feedbacks between financial conditions and the real economy. In recent years and in many countries, credit and housing booms have gone hand-in-hand with strong spending and production. Similarly, during downturns, deteriorating financial conditions have weighed heavily on growth. Such linkages have challenged the long-standing intellectual chasm between macroeconomics and finance. Efforts are underway to bridge it. To a large extent, however, these efforts have neglected the concept and measurement of potential output. This is surprising.
Potential output: From 'inflation-neutral' to 'finance-neutral' measures
From such a unified macro-finance perspective, prevailing views of potential output have two shortcomings.
The first has to do with the concept. A defining feature of potential output is sustainability. From at least Okun (1962) onwards, sustainability has generally been defined exclusively in terms of inflation.1 In Macroeconomics 101, it is the behaviour of inflation that provides a key signal of unsustainability. But identifying sustainable output with non-inflationary output is too restrictive. As the recent financial crisis has powerfully reminded us, output may be on an unsustainable path even if inflation remains low and stable: financial developments may be out of kilter.2
The second shortcoming has to do with measurement. There is little doubt that financial developments contain information about the cyclical component of output. Variations in financial conditions are associated with, and in many instances drive, fluctuations in economic activity. Ignoring them is bound to provide less accurate estimates of potential output whenever this is measured by the non-cyclical component of output fluctuations.
In a recent paper (Borio et al (2013)), we take a first step towards addressing these shortcomings. Our approach fully recognises the critical role that financial factors may play in the evolution of output, and in determining which of its paths are sustainable and which are not.3 More specifically, we rely on two variables that we have found in other work to be the best proxies for the financial cycle: credit and property prices (Drehmann et al (2012)). And we allow these variables to capture the cyclical component of output at traditional business-cycle frequencies.
By operating at these frequencies, we arrive at a measure of potential output, and a corresponding output gap, that are comparable to standard ones in economic analysis and policy. But since it is the behaviour of financial variables, and not inflation, that signals deviations of output from its potential, one can think of the corresponding estimates as 'finance-neutral', as opposed to 'inflation-neutral', measures of potential output.
Statistically, the novelty of the approach is that it incorporates economic information very flexibly. In particular, we do not force the output gap to explain economic variables, as is typically done in systems-based approaches and most prominently through the inclusion of a Phillips curve (eg, the fully fledged production-function approaches commonly used in policy institutions). Rather, we include the proxies for the financial cycle as potential explanatory variables for the transitory, or cyclical, fluctuations in output at the chosen frequencies. If the explanatory power of a given variable is small, it will make little difference to the estimate of potential output. In other words, the approach lets the data speak and, by so doing, avoids a common source of misspecification. Moreover, it is simple and very transparent.
The estimates: Much more precise and robust in real time
Of course, the proof of the pudding is in the eating. And this particular pudding happens to taste quite nice. We illustrate this in the case of the US, although in the paper we also apply the procedure to the UK and Spain, for which the findings are similar. The results are striking.
For one, credit and property prices have significant explanatory power: their inclusion greatly improves the statistical precision of the estimates. Figure 1 compares the 95% confidence bands for the output gap derived from our approach (right-hand panel) with those associated with the output gap resulting from a very common statistical technique based only on output itself, the so-called Hodrick-Prescott (HP) filter (left-hand panel).4 Our procedure roughly halves the size of the error bands. Moreover, these bands are likely to be considerably smaller than those for fully fledged production function approaches, which rely on many assumptions about economic relationships and on several HP trends for key variables.
Figure 1. US: Finance-neutral output gaps – statistical precision (as a percentage of potential output)
Source: Authors’ calculations.
More importantly, our procedure yields output gaps that are much more robust in real time. In other words, history does not get heavily rewritten as time unfolds: the output gap estimated based on information available today (the 'real-time' estimate) is not very different from that estimated a few years later, when more information is available ('ex post'). This is shown in Figure 2, which compares real-time estimates with those based on the full sample of data.
Figure 2. US output gaps: Full sample and real-time estimates (as a percentage of potential output)
Sources: IMF; OECD, Economic Outlook; authors’ calculations.
The difference between our approach and traditional ones is immediately apparent. The HP filter gap and the full-fledged approaches of the OECD and IMF – a representative sample of current approaches – did not detect that output was above sustainable levels during the boom that preceded the financial crisis. In fact, the corresponding real-time estimates indicated that the economy was running below, or at most close to, potential. Only after the crisis did they recognise, albeit to varying degrees, that output had been above its potential, sustainable level. By contrast, the finance-neutral measure sees this all along (bottom right-hand panel). And it hardly gets revised as time unfolds.
Why the difference? It reflects how the approaches deal with what is technically known as the 'end-point' problem. Statistical filters typically have a hard time measuring trends accurately because they are very sensitive to the latest observations. As long as the financial variables soak up the cyclical fluctuations, and do so reliably, they can obviate this problem. This is precisely what is happening here.
A word of caution, though. The finance-neutral estimates shown in the graphs are illustrative. For simplicity, they refer to the linear version of our approach. This version does not allow for the possibility that, as other evidence suggests, the distorting impact of the financial booms increases with their size (eg, Borio and Drehmann 2009). Allowing for this possibility, which we think is more realistic, yields output gaps that are considerably larger during the boom and smaller during the bust.5
Policy application: the case of cyclically adjusted budget balances
For policymakers tasked with keeping the economy on an even keel, large revisions as time unfolds are daunting. Based on traditional methods, policymakers would have completely missed that output was on an unsustainable path ahead of the financial crisis. Had they relied on finance-neutral potential output measures, they would have been in a better position to assess potential vulnerabilities and to take remedial action.
We illustrate this for fiscal policy – although the point applies also to monetary policy. Consider cyclically adjusted government budget balances. Because tax revenues tend to be low and expenditures high during booms, and vice versa, taking into account the state of the economy gives a better indication of the underlying stance of fiscal policy and health of the public accounts. Given that financial booms can flatter the fiscal accounts, neglecting the financial cycle can make assessments unreliable. The recent experiences of Spain and Ireland are quite telling (eg, Benetrix and Lane 2011). The fiscal accounts looked strong during the financial boom: debt-to-GDP ratios were low and falling and fiscal surpluses prevailed. And yet, following the bust and the banking crises, sovereign crises broke out.
The finance-neutral output gap measures help correct for the flattering effect of financial booms.6 Consider, again, the US. Figure 3 shows the actual fiscal balances (red line, right-hand scale) together with the real-time cyclical adjustments based on the HP filter, production function and the finance-neutral potential output measures (bars, left-hand scale). In this context, a difference of more than half a percentage point of GDP is generally regarded as economically significant. During the financial boom that preceded the financial crisis, cyclical adjustments based on the HP filter and production function approaches were small and sometimes even positive. By contrast, those based on the finance-neutral measure were persistently negative, generally above 0.5 percentage points and often in the order of one percentage point, if not larger. Clearly, underlying fiscal positions were substantially weaker than the headline figures suggested.
Figure 3. US: Budget balances and cyclical adjustments (as a percentage of output)
Notes: 1As a percentage of GDP. 2Cyclical correction of the unadjusted budget balance implied by the different output gap estimates. In percentage points.
Sources: OECD, Economic Outlook; national data; authors’ calculations.
Financial developments are an integral part of cyclical output swings. Our findings reflect this simple fact. Assessments of sustainable output that ignore financial factors are fundamentally incomplete.
Authors’ note: The views expressed are those of the authors and do not necessarily represent those of the Bank for International Settlements or the Bank of Thailand.
Benetrix, A and P Lane (2011): "Financial cycles and fiscal cycles," mimeo, Trinity College Dublin.
Borio, C and P Disyatat (2011): “Global imbalances and the financial crisis: Link or no link?”, BIS Working Papers No. 346, May.
Borio, C, P Disyatat and M Juselius (2013): “Rethinking potential output: Embedding information about the financial cycle,” BIS Working Papers No. 404, February.
Borio, C and M Drehmann (2009): “Assessing the risk of banking crises – revisited”, BIS Quarterly Review, March, pp 29–46.
Congdon, T (2008): “Two concepts of the output gap”, World Economics vol 9, no 1, pp 147–75.
Drehmann, M, C Borio and K Tsatsaronis (2012): “Characterising the financial cycle: don’t lose sight of the medium term!”, BIS Working Papers, no 380, June.
Mishkin, F (2007): “Estimating potential output,” Speech delivered at the Conference on Price Measurement for Monetary Policy, Federal Reserve Bank of Dallas, May 24, 2007.
Okun, A (1962): “Potential GNP, its measurement and significance”, Cowles Foundation, Yale University.
1 See, for instance, Congdon (2008) and Mishkin (2007) for useful discussions of the literature.
2 On reflection, there are several reasons for this. Unusually strong financial booms are likely to coincide with positive supply side shocks. Economic expansions may themselves temporarily weaken supply constraints (eg, inducing increases in labour supply and the capital stock). Financial booms often go hand-in-hand with a tendency for the currency to appreciate, as domestic assets become more attractive and capital flows surge; this, in turn, puts downward pressure on prices. And unsustainability may have to do more with sectoral misallocation of resources than with overall capacity constraints (eg, unsustainable expansion of the construction sector).
3 These factors can give rise to what elsewhere we have termed the 'excess elasticity' of the financial system (Borio and Disyatat (2011)). Just like a piece of rubber that stretches too far and eventually snaps, the self-reinforcing interaction between credit creation, asset prices and the real economy can lead to a build-up of financial imbalances that eventually derails economic activity.
4 Technically, the graph refers to a dynamic HP filter, which allows for autocorrelation in the output gap.
5 Moreover, as we argue in the paper, we suspect that our current procedure does not adequately capture these non-linearities. Doing so would likely reduce the estimate of the output gap in the bust further.
6 That said, they do so only partially. They capture the impact of output being above potential, but they ignore compositional effects (the fact that financial booms are revenue-rich) and the build-up of contingent liabilities to address the subsequent bust.