Fernando Eguren-Martin and Andrej Sokol
Asset prices tend to co-move internationally, in what is often described as the ‘global financial cycle’. However, one such asset class, exchange rates, cannot by definition all move in the same direction. In this post we show how the ‘global financial cycle’ is associated with markedly different dynamics across currencies. We enrich traditional labels such as ‘safe haven’ and ‘risky’ currencies with an explicit quantification of exchange rate tail risks. We also find that several popular ‘risk factors’, such as current account balances and interest rate differentials, can be linked to these differences.
Exchange rates and the global financial cycle
The past five to ten years have witnessed a boom in research articles highlighting the co-movement of asset prices across countries. Proponents of this idea claim that such co-movement goes beyond simply reflecting synchronicity in macro-economic developments and has a finance-specific side to it, dubbed the ‘global financial cycle’ (see here). But in a world of globally co-moving asset prices, there is an asset class that stands out from the rest: exchange rates. Because currency values are measured against each other, the scope for co-movement at the global level is limited by definition. That is, we cannot have a situation in which all exchange rates are depreciating or appreciating in tandem.
This feature of exchange rates opens the door for many interesting questions, including one about the relationship between different exchange rates and global financial conditions, which we address in a recent Staff Working Paper. In particular, we study how global financial conditions can help understand the behaviour of currency returns, with a focus on the tails of their returns, that is, on the likelihood of sharp appreciations or depreciations. We also highlight how some country characteristics, such as the level of the current account or of international reserves, can help explain differences in currency behaviour.
Our analysis adds new insights and nuances to widely-held views about particular currencies, such as whether they’re deemed to be ‘safe havens’ or rather ‘risky’, and rests on three building blocks. We first construct a Global Financial Conditions Index (Figure 1), a summary measure of the ‘global financial cycle’ that captures common variation in eight market-based variables across a large panel of countries. We have already used this measure in a previous post to study its relationship with GDP at risk, where the details of its construction can be found. We then rely on quantile regression and related techniques to study how the distributions of different currencies’ returns change shape with the unfolding of the global financial cycle, and with it the chances of large appreciations or depreciations. Finally, we rely on portfolio sorting, an approach common in empirical finance, to identify which country characteristics are associated with currencies’ tail behaviour.
Figure 1: A quantification of the ‘global financial cycle’
Attention to the tail(s)
Much of the literature on exchange rates to date has focused on the mean. So, for example, a number of papers that share our interest in the relationship between exchange rates and global measures of financial conditions (see here, here and here) would typically be interested in the average response of currency returns; a ‘safe haven’ currency would be one that appreciates on average when global financial conditions tighten, etc — the typical answers provided by standard regression analysis and its many refinements. Our goal is to go beyond the mean. In parallel fashion to the growing interest in ‘GDP at risk‘ and other analyses that aim to better capture the full set of risks implied by a particular economic relationship, we want to understand what global financial conditions can tells us about the risk of sharp appreciation or depreciation episodes for a large set of exchange rates. Because of the nature of our question, we model the response of the entire distribution of exchange rate returns to global financial conditions, to then zoom in on the tails. To this end, we make use of the increasingly popular quantile regression methodology, which allows us to model in a tractable way the whole distribution of a variable (in our case currency returns) given a set of explanatory variables (in our case, global financial conditions).
Figures 2 and 3 show this exercise applied to the returns of the Japanese yen and the Australian dollar, two polar cases that are often taken as examples of a ‘safe haven’ and ‘risky’ currency, respectively. In blue we plot the distribution of these currencies when global financial conditions are at their average, while the orange distributions give us information about the likelihood of different returns in the event of a one standard deviation tightening in global financial conditions. We can see that the shifts in the distributions are in line with prevailing market narratives: in the face of tighter global financial conditions, the likelihood of observing a sharp appreciation (right tail) in the yen goes up significantly, while in the case of the Australian dollar it is the likelihood of a sharp depreciation (left tail) that goes up markedly.
Figure 2: Distribution of Japanese yen returns
Figure 3: Distribution of Australian dollar returns
While it is reassuring to see that our results match ‘market talk’, it is important to note that the insight is much richer: we can quantify the shifts in the tails of the distributions for different currencies (by making use of relative entropy measures) and compare them. This way, we can rank currencies according to how much ‘fatter’ the left (downside entropy) and/or right (upside entropy) tails of their respective return distributions become in the event of a tightening in global financial conditions, ie how much more likely it becomes that they experience a large depreciation or appreciation. This is what we do in Figure 4. We can see that the resulting ranking of a wider set of currencies also goes in line with prevailing narratives: the Australian dollar remains top of ranking in terms of an increased likelihood of a sharp depreciation (large downside entropy), while at the opposite end of the spectrum there are a series of currencies with large upside entropies that join the yen as ‘safe havens’, notably the Swiss franc and the US dollar.
Figure 4: A quantification of the sharp appreciation and depreciation risks
A natural question is why certain currencies behave the way they do. This is not a new question, but one that has a tradition in the literature about ‘safe haven’ status. However, while the focus to date has been on average behaviour, we’re explicitly interested in the risks: for example, we want to know why, in the face of a tightening in global financial conditions, the likelihood of a sharp yen appreciation goes up, while that of an Australian dollar appreciation does not, and in fact it is the likelihood of a sharp depreciation that increases.
Addressing such questions comes with new challenges: while the distributions we have shown are estimated over the whole sample, it is reasonable to assume that the underlying currency (or country) characteristics associated with such behaviour, ‘risk factors’ in technical terms, change over time. Suppose we thought that currencies of countries with large current account deficits become more exposed to the risk of experiencing a sharp depreciation in the face of tight global financial conditions; it would be unreasonable to try to relate such shifts in distributions with average current account values over the 20 years in our sample. Thus, we need to incorporate a degree of time variation in our analysis. In order to do so, we rely on portfolio sorting techniques, widely used in equity and FX pricing literatures (see here and here), and consider portfolios constructed on the basis of countries’ current account balances, interest rate differentials with respect to the rest of the world, level of international reserves and fiscal balances.
The idea behind portfolio sorting is to group currencies into portfolios according to a given common characteristic and design an investment strategy whose return is determined by the relative performance of portfolios that score differently on these characteristics. This relative performance is informative about market perceptions of the underlying sorting characteristics as risk factors (see paper for more details). By running the same quantile regression machinery on these relative returns series, we can study whether the risk of a sharp negative return goes up in the face of tighter global financial conditions. If this is the case, then the framework is telling us that a particular characteristic, say current account balances, is indeed a risk factor; that is, a currency characteristic plausibly taken into account by market participants when choosing their currency positions.
Our results point to current account deficits and positive interest rate differentials as the most robust indicators of increased chances of seeing a sharp depreciation in the event of a tightening in global financial conditions. A low level of international reserves, another indicator previously considered by the literature, seems to play some role too, while fiscal deficits do not seem to contain much useful information in this respect. Further potential risk factors, as well as a better understanding of their relative importance, are both being investigated in ongoing work.
While obviously stylised, our framework is intuitive and simple to implement, and could be of use to policymakers monitoring macro-financial risks at the country level, as well as to market participants modelling the FX risk of their investment strategies.
If you want to get in touch, please email us at firstname.lastname@example.org or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.