Currency valuation methodologies
- Valuation is a key pillar of active currency management, despite the long time horizon required.
- Approaches to measuring valuation vary, but purchasing power parity is most widely used by currency managers and is supported by academic research.
- Our research has demonstrated the importance of ensuring the comparability of price level data, a task that is more challenging than many appreciate.
- We have found little benefit to other theoretically motivated enhancements to valuation modelling. We prefer to address the impact of cyclical developments on currencies separately.
Approaches to valuation
Valuation is a key pillar of our approach to active currency management and one of the few methods of currency analysis widely accepted by the academic community. Currency valuation models provide alpha that is very diversifying compared to other asset classes and factors, but requires a long time horizon and a tolerance for drawdowns.
Purchasing power parity (PPP) is the basis for the majority of currency valuation models, though in practice there are a wide range of approaches that vary from the relatively simple concept that exchange rates should keep prices stable in common currency terms, to the more complex behavioural equilibrium exchange rate (BEER) models. The other broad category of academic approaches known as fundamental equilibrium exchange rate (FEER) or macro balance (MB) models suffer in practice from being highly sensitive to the calibration chosen by the user and are not widely used by practitioners due to poor performance.
Generally, attempts to adjust for the Balassa Samuelson effect1 - where higher productivity should theoretically result in a stronger exchange rate - suffer from challenges with productivity data quality, timeliness and vulnerability to historical data revisions. However, we have found using GDP per capita as an adjustment to valuations provides a significant enhancement for emerging market currencies, while the effect for the major developed market currencies is modest but additive. We have found a single, simple measure, such as GDP per capita, to be more effective than more complex approaches.
We also prefer to address the cyclical drivers of exchange rates separately from valuations, consistent with recent European Central Bank (ECB) research2 that shows no benefit to including these factors within valuation models.
Not all purchasing power parity models are the same
While the underlying concept of PPP is simple enough, there is a significant level of complexity in measuring price levels and in comparing data across countries where national statistical agencies may use differing methodologies.
Some examples of data issues include difference in coverage or methodology for imputed costs, such as housing; different hedonic pricing in areas of rapid technological progress, such as communications; and basic differences in index methodology, such as the historical use of the discredited Carli method3 or, more recently, the different approaches to missing price data during the height of the pandemic.
As a specific example, imputed housing rental costs have one of the largest consumer price index (CPI) weights in many countries, yet are not included in others. The International Labour Organization (ILO) guide to CPI methodology4 allows three approaches that have materially different outcomes in terms of the level of overall CPIs through time. In general, the “use approach” is most favoured, resulting in a higher CPI than the “payment approach” but lower than the “acquisition approach”. The bias introduced by this methodology is a contribution to the persistent difference in inflation rates between the US and Europe (Exhibit 1), though there is ongoing work by Eurostat to address this issue5.
Exhibit 1: A Comparison of CPI and HICP For the US Demonstrates the Impact of Housing Inflation Methodology (Index)
Extensive efforts are made to correctly adjust prices for the impact of technological change, yet the smartphone has presented a particular challenge for these methods. Smartphones cross the boundaries between many different products, and pricing is often bundled in with telephone services.
The resulting differences in price trends in sectors with rapid technological progress, such as communications, may relate more to the statistical methodology than divergence in underling economic developments. In particular, we are sceptical about the greater deflationary impact of communications prices in Sweden since 2004 (Exhibit 2).
Exhibit 2: Communication Price Divergence Amplified by Hedonic Adjustment (Price Level Index, Log Scale)
The impact of the pandemic on missing prices and changing consumption patterns will also have exacerbated the impact of methodological decisions (Exhibit 3). Examples include items unavailable during lockdown, such as beer prices in British pubs, where any price increase on reopening would be missed as no base period prices were available6. There are reports of promotions that typically reduce the cost of groceries becoming less prevalent7. The materiality of such effects remains to be seen.
Exhibit 3: Share of UK CPI Basket where Prices were Unavailable (%)
Generally, these issues are still present in data from international sources, since they tend to have quite broad criteria for data standards and simply aggregate information provided by national statistical agencies.
There is one source that produces comparable global data:, the International Price Comparison Project (IPCP), run by the World Bank8. This research project brings together national statistical agencies and aims to produce comparable pricing data across countries, addressing a wide range of theoretical challenges through both methodology and the collection of new primary data. The trade-off from the perspective of a currency manager is the timeliness of the data, which lags significantly behind traditional sources.
Our research has demonstrated a noticeable improvement to our quantitative valuation modelling by systematically excluding housing inflation from national price series data. We have also found the IPCP data to provide a complement to the more timely national data, with a blend of approaches achieving additional diversification benefits within our investment process.
There are, of course, other country-specific distortions to pricing data, but none have as much impact on valuations as those of the housing sector. While it is theoretically possible to attempt to adjust for all distortions, such an approach would be unwieldy and potentially lead to an over-fitted deflator construction. Instead, we seek to take advantage of our understanding of these more specific issues through a qualitative approach, where we cross check our valuations with balance of payments analysis.
In practice, valuation models have been a key tool for currency management at J.P. Morgan Asset Management for more than 30 years. Even such a simple concept requires ongoing research and evolution to maintain confidence in the accuracy of valuation estimates throughout an economic cycle.