When making cross-border capital allocation decisions, exchange rate movements can trip up good investments. Is it possible to predict currency movements? We give a cautious yes when the time horizon is 3-10 years, i.e. a strategic outlook (as opposed to currency traders’ horizon of seconds to at most months).
The graph to the left above shows the over- and under-valuation of all the world’s currencies from 1982 till 2017. Is it possible to make any sense of such a spaghetti?
We believe so. We used the Canback FX Tool to estimate over- and under-valuation. It is not complicated and builds on inflation-adjusted exchange rates. A twist is that we separate out USD volatility, so that all currencies, including USD, are evaluated against a “world currency.”
The interactive graph below allows you to see each currency individually or against each other. We do not pass judgment on the exact over- or under-value of individual currencies. The vertical scale is therefore missing. The scale is the same regardless of country chosen.
Each year’s exchange rate is the annual average except for 2017 where it is the rate on the day of analysis (April 40.
Statistically, the future change in currency value is always negative as a function of over/under value, as should be expected. This is based on a pooled regression (XT = currency/year).
For a 1-year horizon R² is low, for a 3-year horizon R² is moderate, and for 5-10 year horizons R² is good (within-R²; between-R² is always high). t-stat ranges from significant to highly significant.
Finally, we wonder why there is a global minimum over/under value around 2003-2005 (select all countries and it is visible). Perhaps the end of Bretton Woods led to progressively better monetary policies that gradually reduced differences up till the early 2000s. Then the gradual onset of the financial crisis increased differences and since then, other policy concerns have overridden exchange rate policies. Also, those are the years when the fewest number of countries had a recession.
Caveats: We do not make any recommendations based on the above analysis. Past behavior may not hold in the future. Individual currencies may differ from the global findings above. We unknowingly may have made analytical mistakes.
The analysis was updated April 4, 2017.
The over-values (x) have been transformed by x/(1+x). This gives a pleasing symmetry so that the maximum over-value of 100% and a maximum under-value of -100%. This in turn leads to good statistical properties. This simple transformation is seldom used, yet it applies to many processes where the upside is infinite.