The return of currency volatility
With the global financial crisis, currency volatility has leapt about fourfold after a long period of false calm. For sterling, over the duration of this equity bear market, this has coincided with almost unprecedented declines of between 30% and 40% against the currencies of the equity markets we invest in. Since the UK was one of the economies with the worst financial imbalances, both between domestic sectors and with the foreign sector, this drop is a significant element of the required economic adjustment process. In this Insight, we look at the prospects for key exchange rates from two different perspectives: the underlying process, which we believe to be a function of Purchasing Power Parity theory, and a geopolitical view of possible exceptional developments, ranging from the loss of dollar hegemony to partial unwinding of the euro zone.
How currency movements affect investors focused on real wealth outcomes.
Our interest in currencies is as an integral part of the process by which any equity market anywhere in the world acts as a system for generating real returns for any investor whatever their home currency.
Our local-currency analysis of single equity markets demonstrates clearly the return-generating characteristics that are broadly common to all equity markets, regardless of differences in their culture, economic performance or stage of development. These characteristics emerge only when equity returns are measured: in real terms, deflated by each market’s own inflation in total-return terms, including both capital movements and the reinvestment of dividends.
In a real, total-return model of the equity process, foreign- or local-currency real returns are translated into home-currency real returns via the exchange rate process. If this process is itself a function of experienced differences in inflation between two countries, then there will be a tendency for real returns in different currencies to equilibrate. The theory that posits that exchange rate changes over long periods are explained by inflation differentials is known as Purchasing Power Parity (PPP). The assumed underlying economic driver is an equilibrium process, where movements in the real exchange rate (ie movements not explained by inflation differences) lead to swings in the foreign trade account that, to ensure continued competitiveness, correct via subsequent exchange rate movements. This equilibrium explanation is consistent with ‘mean reversion’ in real exchange rates.
The validity of PPP is most obvious where the inflation differentials are very large, such as between one country with low inflation and another with high inflation. The theory has less explanatory power when the inflation differentials are quite small, such that even if trade equilibrium remains the best explanation of the underlying process other factors may then be driving the trade balance.
It is also clear that PPP may explain very long-period shifts in exchange rates but does not explain short-period volatility. The dispersion of returns around a measured real exchange rate is quite large and the deviations also tend to be serially-correlated, trending successively higher or lower than PPP would suggest. For short-term investors or traders, momentum or trend-following is therefore a more popular approach to managing currency positions. However, as was demonstrated by the collapse of the yen ‘carry trade’ (borrowing low-yielding yen to invest in high-yielding currencies like the NZ$), years of gains from following a persistent short-term trend in deviations from PPP can be wiped out in just a few weeks if PPP reasserts itself.
What are real exchange rates telling us about future currency movements?
We show below plots of the real exchange rates, adjusted for observed differences in inflation, for each of the three cross rates relevant to our equity exposures: the US dollar, Japanese yen and euro.
The graphs are intuitive. When the plot of the currency paired with sterling is rising, its real exchange rate against sterling is rising, ie the actual exchange rate is increasing by more than the inflation differential for the same two counties, and vice versa when the real exchange rate is falling. Each series starts at 100 and there is no attempt to recalibrate them so that a base of 100 is actually equivalent to parity. You can explore this point further when we examine an appropriate currency forecasting model as a technical appendix to this article.
Both the dollar and the euro show the characteristics sought with PPP, with relatively little overall trend on the whole 40-year period. The yen/sterling real exchange rate, however, trended fairly persistently higher, reaching a peak index level of 2.5 times the starting level in 1995. However, the subsequent correction on the yen/sterling real exchange rate, up to July 2007, has restored consistency with PPP.
Even without recalibrating each series to estimate parity, we can make some general observations about current real exchange rates:
Sterling general weakness was ‘predicted’ by extremely low real exchange rates against each of the dollar, yen and euro in the winter of 2007
The subsequent fall of sterling against the euro has left the euro nearly as overvalued as at earlier turning points relative to the deutschemark, in 1973, 1976 and 1995
Against the dollar sterling is fairly close to the mid-point of its trading range over the whole period but after its recent devaluation looks to be at an extreme in terms of the narrower range that has held since 1987
The recent yen strength has offset almost half the long correction of the overvaluation of the yen between 1995 and 2007 and the high yen real exchange rate has only been exceeded during the period of exceptional strength in the early 1990s that coincided with the onset of Japan’s economic crisis.
The implications for investment returns are that, unless sterling weakness shows continued momentum (which will quickly take real exchange rates to significant extremes), investors seem likely to give back some of the benefits of currency gains that have cushioned the bear market so far. This is allowed for in our equity return projections, even for quite short periods, because of the addition of currency risk to our measures of equity risk.
We have also calculated some other key cross rates, not including sterling, below. In these graphs the plots measure the real exchange rate of the first of the pair. These suggest that the US dollar real exchange rates against the euro and yen are not particularly extreme and that the euro/yen rate has also now moved away from an extreme. In this situation, momentum may explain most of the next exchange rate moves, momentum possibly being with the dollar rather than the yen or euro.
Geopolitical developments Here, we move way outside our comfort zone as economic commentators and are simply reflecting the occasional speculation in the broad sheets and regular speculation in economic blogs.
In much the same way that the break-up of Bretton Woods, floating exchange rates and the dismantlement of post-war exchange controls mark a hiatus between exchange rate behaviour pre and post about 1970, there is always a possibility that geopolitical influences will arise that are large enough to overwhelm trade competitiveness or momentum as an explanation of the currency return process.
A key development that cannot be dismissed is the US losing its dominant role as a reserve currency, which in conjunction with the post-Bretton Woods changes has meant that the US government has been free to create dollars at will and foist them on foreigners. Often referred to as ‘dollar hegemony’, this would be put at risk if other nations decided they were unwilling to accept the dollar as the main currency for their official foreign exchange reserves.
Since the economic problems we confront now originated in a particular example of dollar hegemony, namely excessive US consumption and indebtedness financed by the nations providing the resulting imports, it is not unreasonable to think that foreigners may choose to replace the dollar in favour of a multi-currency reserve approach. The Obama Administration is presumably acutely sensitive to the risk that China will want to take this route, even though China cannot diversify its currency holdings quickly without also increasing currency losses on its existing holdings of financial assets.
Reserve currency status should go hand in hand with deep and liquid financial markets. China and other large creditor nations may feel that these will be hard to replicate outside the US. This issue has led to consideration of using IMF Special Drawing Rights, already defined by a mix of reserve currencies, as an alternative to public financial markets.
This dollar speculation is necessarily linked to another possible geopolitical event: a break-up of the euro zone. However, the two may be mutually exclusive. Fragmentation of the euro zone may not be consistent with the loss of dollar hegemony since the euro zone and its separate financial markets represent one of the few realistic rivals to the dollar and US financial markets.
The threat to the cohesion of any monetary union arises from pressures on individual members that other members are either unable or unwilling to tolerate. In the euro zone, the pressures originate in structural financial imbalances that grew in significance during the growth phase for the economy and are now being aggravated, instead of alleviated, by the necessary phase of adjustment. These structural imbalances have for a long time appeared greatest in Spain, Greece and Italy but the credit crisis has since highlighted particular challenges for Ireland.
We are not competent to comment on whether the fragmentation risk is only a function of euro zone member states or whether strains emerging within EU member states (or intended new members) that are outside the euro zone can trigger fragmentation of the zone members themselves. This is obviously relevant because so many of the former Eastern bloc countries are heavily indebted, particularly to euro zone lenders, and widespread defaults would add a new dimension to the banking crisis as it already affects EU banksm
History suggests monetary unions have a short shelf life yet most serious commentators are dismissive of the chance of the euro zone fragmenting. How to resolve this contradiction is part of what puts this topic outside our comfort zone.
Conclusion The scale, but also the unpredictability, of such cataclysmic events strongly suggests that investors need to find some accommodation with currency risk that does not require taking a view on the chances of these events.
Clearly, geographical and currency diversification of global equity exposure is an easy and practical form of accommodation. On the other hand, trying to avoid foreign equity exposure and focusing instead on the UK, though directly reducing exposure to these event risks, also introduces increasing dependence of long holding-period outcomes on the economic performance of the UK. Caught in the cross fire of much larger exchange rate developments, UK economic risk will itself tend to increase as a direct consequence of the factor investors are trying to avoid.
This is part of the appeal of simply following a PPP approach to currency strategy in an investment portfolio. It means an investor focused primarily on long-horizon real wealth outcomes can diversify his or her exposure to a series of self-similar real-return generating systems.
But it is also possible that the real exchange rate process we have described directly impacts on the chances of geopolitical events themselves, such as because the principle source of the pressure for cataclysmic change is itself eased by a move from currency undervaluation to fair value or vice versa. This convenient relationship would be hinted at if, for instance, there has been a reduction in speculation about the euro zone break up since its key cross rates with the dollar and yen started to correct from earlier extremes.
Appendix: a technical explanation of PPP for portfolio investors As noted earlier, isolating variance in the real exchange rates is highly relevant as the key measure of incremental risk from currency exposures in a real-return model. Also as noted earlier, this is convenient because of the ‘big idea’ that equity exposure in any market anywhere is equally capable of generating real returns over long holding periods for any investor in any home currency. Hence our adoption of PPP is specific to our overall investment approach.
Technically, PPP is also very convenient for us because the data properties are remarkably similar for real exchange rates and real equity returns. Over the holding periods relevant for foreign or UK equity exposures, we appear to be able to assume that both real exchange rate and real equity returns are log-normally distributed rather than having fat tails. Both show the characteristic of mean reversion. The data supports a view that real exchange rate deviations revert to the mean quicker than real equity returns and that the deviations are themselves smaller than for equity real returns. This looks consistent with the assumed underlying theory since the responses to differences in competitiveness in tradeable goods are likely to be much faster to emerge than differences in investment returns in a global equilibrium model.
Applying a PPP approach offers us two opportunities to incorporate currency-related estimates into our return-generating model for any projection period:
a mean expected currency return based on the observed deviation from, and eventual reversion to, PPP
a measure of currency risk derived from the historical variance not explained by observed inflation differences
If we want to combine the mean reversion effect of currency adjustments under PPP with the mean reversion effect of equity markets being valued above their own trend of achieved real return, we can do so simply by multiplication (or addition if using log numbers). We observe that currency risk and equity market risk are statistically independent. Independence means we can combine the two sources of uncertainty, or forecasting error, for market and currency, as the square root of the sum of the squares. In other words, adding currency risk to market risk expands the combined risk but not to the same extent as if the two were causally associated with each other.
Though the Lambda model supports both currency forecasts and currency risk estimates, No Monkey Business has chosen to ignore the former and only adopt the latter into its overall return-generating process. The impact is that foreign equity markets will always tend to be riskier than the UK but without a calculated reward for bearing that extra risk. When we move from the return-generating stage of the model to the portfolio-optimisation stage, this unrewarded risk will tend to be offset by diversification benefits arising from our estimate of the correlation between different markets.
If we had the same confidence in the currency model as we do in the market model, it would not make sense to ignore the mean reversion effect under PPP. If we had less confidence but nonetheless believed extreme deviations from PPP tended to produce more reliable forecasts, we could introduce a currency return element only when those extremes appeared to be present. Because the currency model, since it was designed by Lambda in 1999, has performed well over the time horizons relevant to us, we are considering both options. Indeed, as already noted, the recent adjustments in sterling rates were ‘predicted’ by the model. To reflect this prediction, many clients were asked to consider abandoning a ‘home bias’ in favour of an unconstrained approach to geographical diversification. Given what subsequently happened to sterling, this has helped their returns in the bear market (as set out in an earlier Insight on our portfolio performance).
There are some issues with a currency-return generating model.
The scale differences between the yen currency pairs and all the other main pairs poses a possible problem for modelling real exchange rates as a single global process
To convert these plots into a measure of deviation from parity, in order to generate forecasts based on mean reversion, requires some attempt to recalibrate the series so 100 represents parity, or equilibrium.
Many economists and currency analysts use trade data to try to measure parity, or to confirm parity implied by the time series (such as if periods of implied parity coincide with broadly balanced trade accounts). A more agnostic approach is to assume that, over a long enough period, the deviations from parity will even out and that a fitted regression will provide a reasonable estimate of parity. When the Lambda currency model was first devised in 1999, we used judgement based on trade data. Our view in 2006 and 2007 that sterling was overvalued was, on the other hand, simply based on the low levels of the uncalibrated indices, particularly for the yen and dollar pairs rather than the euro/sterling pair.
Now that we have as many as 40 years of data under floating exchange rates, there is an argument for relying purely on regressions to measure parity. However, these still have an observed slope slightly different from zero whereas zero is what we would hope to see if the data was fully representative and the underlying process was entirely consistent with PPP. With a few more years data, however, we may find, as we did for the yen, that the data set is really representative of an underlying PPP model.
We are currently researching with Chris Drew the basis of currency return forecasts, specific to the same periods for which we already calculate mean expected equity returns, and for integrating the two components so as to calculate a mean expected sterling real return from each foreign market. As with equities, we can differentiate our confidence in the forecast as a function of the data length and quality (in the equity model, less data history implies less confidence in the trend itself but also in the ‘current’ conditions relative to trend, increasing the standard deviation of expected returns).
The method for calculating the incremental currency risk for foreign markets, which is already part of our real outcome risk measures, will not change.
Part of the focus of this research project is whether to integrate currency forecasts all the time or only when we observe extreme real exchange rates. Clients will be kept informed of the progress of this research project.