The search for the elusive twin goals of monetary and financial stability

Abstract

ing from some cross-country differences, the graphs illustrate that since the 1970s two major cycles have taken place and a third is under way, in sympathy with real economic activity. They correspond to the early to mid-1970s, the mid-1980s to the early or mid-1990s, and the second half of the 1990s to the present. Japan did not take part in the latest upswing following the bust in asset prices at the turn of the 1990s and the subsequent “lost decade”. The data indicate that, if anything, the size and amplitude of the cycles may be growing. These cycles have typically coexisted with similar fluctuations in credit. Since the 1980s, the ratio of credit to GDP has risen markedly in most countries. In addition, the evolution of this ratio has tended to exhibit a generally positive correlation with medium-term swings in asset prices, as further confirmed by more detailed econometric evidence (Borio et al (1994)). Importantly, the same formal empirical work indicates that this correlation appears to have become tighter following financial liberalisations. More recent studies suggest that the correlation is especially close with real estate prices, as might be expected (Hofmann (2001), Davis and Zhu (2004)). Financial imbalances herald financial instability A look at the empirical record indicates that financial imbalances, in the form of unusually strong and sustained credit and asset price expansion, have preceded most of the episodes of serious financial instability and strong financial “headwinds” that have become more common since the 1980s, in industrial and emerging market economies alike. This retraces a pattern that was quite familiar under the Gold Standard (Kindleberger (2000), Goodhart and Delargy (1999)). This observation has been confirmed by more formal statistical work. With Phil Lowe, for instance, we have shown that it is possible to “predict” fairly well episodes of major banking distress based on very simple proxies of “financial imbalances” (Borio and Lowe (2002a), (2002b), (2004)). The proxies are based on two key variables. The first is a measure of misalignment in some key asset price, which can be taken as an indicator of the likelihood and size of a reversal. The second is some measure of the private sector leverage (here private sector credit in relation to GDP), which can be taken as an indicator of the likely damage caused to the economy by the reversal in asset prices. Both of these elements measure deviations from the “normal” range of historical experience, or “gaps”, defined in terms of deviations that exceed certain critical thresholds. Given the limited time available, let me just highlight a few points about this empirical evidence, summarised in Tables II.2 to II.4. First, the gap variables are constructed based exclusively on information available at the time when the predictions are made; technically, the trends are calculated recursively based on ex ante information only. This is important to make sure that they can be useful for policy. Second, they are calibrated based on developments during the boom only. Thus, they can be thought of as helping to distinguish sustainable from unsustainable economic expansions. This is reinforced by the fact that the proxy for financial imbalances contains information well beyond that contained in traditional output gaps (same tables). Third, it is the requirement that the thresholds be exceeded simultaneously by the indicators of price misalignment and leverage that helps to improve the accuracy of the prediction. It does so by eliminating a lot of “noise”, namely by not predicting too many crises. This underscores the point that what matters are financial imbalances, not asset price misalignments per se. Fourth, the financial imbalance proxies contain information that extends beyond short horizons, with the accuracy of the predictions improving as the horizon is lengthened. On annual data, we have shown that this is the case if one extends the horizon from one to three years ahead (Tables II.2.and II.3)). More recently, based on quarterly data for industrial countries, we have shown that this is also the case if the forecasting horizon is three to five years ahead. This highlights the significant time that the build up and unwinding of imbalances can take and that the information content goes beyond the typical horizons used for monetary policy.