All other things equal, an improvement in a country’s terms of trade, the ratio of export to import prices, translates into increased demand for its currency and a boost for its growth outlook. However, terms of trade are a rather subtle and sporadic influence. Therefore, many market participants are rationally inattentive to smaller changes and unwilling to trade on large changes in times of turmoil. This points to investor value in the systematic consideration of monthly or annual terms-of-trade dynamics, which can be approximated by commodity-based export and import price indices. Empirically, standard terms-of-trade dynamics have indeed predicted FX returns positively since 2000, across developed and emerging market countries. However, while this relation has been fairly stable in the developed world since 2000, for emerging markets the trading value of terms-of-trade indicators has only become evident since the great financial crisis.
The below post is based on proprietary research of Macrosynergy Ltd. It uses data from the JPMorgan Macrosynergy Quantamental System.
This post ties in with this site’s summary of trading strategies based on macro trends.
What are commodity-based terms of trade?
In macroeconomics terms of trade refer to the ratio of an area’s export prices and import prices. Put differently, it measures the volume of imports that an area can get for a unit of its exports. All else equal, an increase in terms of trade implies an improvement in competitiveness and economic performance.
Official terms-of-trade statistics are typically released with a few months lag to their observation period. Also, they are completely backward-looking, measuring only prices that have actually been paid for past shipments. For timely information, financial markets have come to pay more attention to commodity-based terms of trade. Put simply, these indices measure the effects of primary commodity prices on an area’s terms of trade. While primary commodities make up only about 50% of international trade their prices tend to fluctuate more than those of final goods and services and tend to dominate short-term dynamics. Moreover, these price changes can be tracked or even predicted based on real-time quotes from international commodity markets.
Commodity terms-of-trade data are being published by various international organizations. However, the most popular terms-of-trade data set in financial markets have been the Citibank commodity terms-of-trade indices. They have been published daily with historical information going back to the 1990s, always incorporating the influence of the latest price changes in commodity markets.
For this post, we use the J.P. Morgan Macrosynergy Quantamental System (“JPMaQS “) commodity terms-of-trade, a newer data set, which has several advantages over the Citibank indices for the purpose of developing trading strategies:
- JPMaQS series are point-in-time, which means that for each date in history, they use the latest concurrently available commodity trade weights and producer prices.
- While the Citibank terms-of-trade proxy implicitly assumes stable non-commodity prices (effectively biasing terms-of-trade proxy in favor of commodity exporters), the JPMaQS proxy assumes that non-commodity export and import prices have grown in line with the U.S. core consumer price index.
- JPMaQS uses locally relevant commodity prices, wherever possible, not just one single global benchmark per commodity. For example, U.S. crude export values are proxied by WTI, while for most other countries crude prices are approximated by Brent.
JPMaQS terms-of-trade dynamics are based on ratios of approximate export and import price dynamics, whereby the latter are calculated daily using concurrent vintages of commodity prices (in USD), commodity trade shares, and (as a non-commodity tradable USD goods’ price proxy) the U.S. core producer price index. Export and import price dynamics are collected as ratios of the latest price index divided by the index at the base period. The underlying vintages of all categories are based on commodity trade shares and related prices of liquidly traded commodity contracts. Since the range of available traded contracts has broadened over time, so has the range of commodities that are used for the index. For example, in 1995 the number of usable contracts was 36. By 2005 it had grown to 44 and by 2015 to 62.
For more details please check out the technical annex in the JPMaQS terms-of-trade documentation here (which requires access to J.P. Morgan markets for now)
Why should commodity-based terms of trade have trading value for FX?
All other things equal, an increase in the terms of trade of a currency area implies a rise in demand for the local currency and an improvement in its growth outlook. That is because residents receive more foreign currency for their output and/or pay less for their consumption and investment inputs. For example, a country that mainly exports base metals and imports agricultural commodities would obtain relatively more revenues from external trade if the price of the former rose relative to the latter, boosting currency inflows, corporate profits, and real disposable income.
While FX markets should plausibly respond to exogenous larger changes in terms of trade, it is unlikely that they fully and consistently account for such information, for two reasons:
- Most of the time, terms-of-trade changes are modest, and, hence, their direct influence on the exchange rate outlook is not easily noticeable at a weekly or monthly frequency. Neglect of subtle factors is consistent with “rational inattention”. Rational inattention acknowledges that agents cannot process all available information and choose which exact pieces of information to attend to by considering its value [view paper here].
- Large changes in commodity prices often occur in the context of broad price changes in financial markets. For example, a global risk shock triggered by fears of financial stability may push down a wide array of prices, including those of commodities and currencies that benefit from that commodity price decline. There is typically little appetite in markets for taking tactical positions in risky currencies in times of turmoil merely because their terms of trade have improved. The power of global market factors explains why historically an improvement in terms of trade has not for all countries been positively correlated with currency appreciation. Indeed, regression of real effective appreciation of terms-of-trade suffers from what econometricians call an “omitted variable bias”.
Simple empirical checks of trading value
The dataset
We look at the predictive power of standard commodity-based terms of trade changes for subsequent monthly FX forward returns.
- In JPMaQS various terms-of-trade dynamics are calculated based on concurrent data vintages. Since this post mainly looks at predictive power one month or one week ahead we focus on the terms-of-trade change of the latest week over the previous 4 weeks and the latest month over the previous 12 months, as well as an average of these two. The average will be our main point of reference. FX returns are calculated based on vol-targeted positions to equalize the impact of different currencies for the analysis.
- FX forward returns are measured against their dominant benchmark, which is typically USD, except for European currencies, which trade either mainly against EUR or against both (GBP, RUB, and TRY).
Terms of trade changes and FX returns are considered for 26 developed and emerging currencies. The sample period is a bit more than 23 years, i.e. 2000 to 2023 to date. For a full list of currencies and symbols used see Annex 1 at the bottom of the post. Not all markets were always tradable or flexible during the sample periods and we exclude periods of inconvertibility, illiquidity, and tight exchange rate pegs.
The relation of terms-of-trade and subsequent returns
As hypothesized, there has been a positive correlation between the average terms-of-trade changes and subsequent FX forward returns of a currency. The relation has been subtle and not visually striking in scatterplots but has been significant with 98% probability at a monthly frequency and 99% at a weekly frequency, based on the Macrosynergy panel test. For details on the test methodology, one can view a post here. Similarly, the Kendall non-parametric correlation coefficient has been positive and significant with a probability of 99.5% at monthly frequency.
The balanced accuracy of directional return predictions, i.e. the average of the ratios of correctly detected positive returns and correctly detected negative returns, has been 51.4% across the panel of currency areas and months. The balanced accuracy has been above 50% in two-thirds of all years since 2000 (for the full set of currencies) and in above 70% of all currency areas (for the full set of years), suggesting that accuracy was pervasive rather than concentrated.
PnL generation of terms-of-trade signals
We calculate naïve PnLs based on the following assumptions. First, positions are taken based on z-scores of the terms-of-trade changes. The z-scores are winsorized at 2 standard deviations to mitigate data outliers and to avoid excessive risk-taking in any single market or period. Second, positions are rebalanced monthly with a one-day slippage for trading. And third, the long-term volatility of the PnL for positions across all currency areas has been set to 10%. These are standard procedures that we have been using in previously published proof-of-concept analyses. Note that this PnL is called “naïve” because it does not consider transaction costs and realistic risk management rules.
Naïve PnLs based on standard tactical terms-of-trade signals have produced long-term Sharpe ratios of 0.45 and 0.5, without significant correlation to either equity baskets or the dollar. However, backtested value generation has been highly seasonal and all positive contribution has occurred after the great financial crisis.
The phenomenon of “long seasons” in terms of trade-based FX trading strategies apparently reflects the subtle nature of terms-of-trade in general and the dominance of capital flows into emerging markets in the 2000s in particular. Note that 18 of the 26 tradable currencies in the set are classified as emerging markets. The 2000s prior to the great financial crisis saw the transition of many EM currencies to convertibility and flexibility and the development of local fixed-income markets. This coincided with the rising interest of global investors in local EM currency markets. In periods of large capital flows into a set of countries global funding conditions are the dominant driver and shocks to financial conditions can easily push terms-of-trade and currency value in opposite directions (see scatter plot panel above). As emerging currency markets matured, subtle forces such as terms of trade gained importance.
The EM capital flows hypothesis is supported by two facts.
- First, the PnLs of long-only positions in small and EM currencies and terms-of-trade strategies have been highly complementary. As long as “long-only” strategies performed well, terms-of-trade strategies performed poorly. And when “long-only” exposure failed to produce value, terms-of-trade strategies generated significant positive PnLs. A combined managed long-biased strategy based on terms of trade would have produced strong and consistent value.
- Second, the PnL generation of the terms-of-trade signal for developed market currencies has not displayed such extreme seasonality, even though it only makes use of a subset of 8 currencies. The average and medium-term terms-of-trade signal would have produced a positive PnL in the 2000s, the 2010s, and the 2020s (until March 2023). Long-term Sharpe ratios would have been between 0.29 and 0.32, which is respectable for subtle factors and a small set of positions.
Accuracy and balanced accuracy of directional return prediction would have been above 50% for all developed countries since 2000. This means in over 23 years of history not a single DM currency would have failed to pick up an above-par accuracy of terms-of-trade-based predictions.
Annex 1: Currencies and tickers used in the post
The currency names are in alphabetical order: AUD (Australian dollar), BRL (Brazilian real), CAD (Canadian dollar), CHF (Swiss franc), CLP (Chilean peso), CNY (Chinese yuan renminbi), COP (Colombian peso), CZK (Czech Republic koruna), GBP (British pound), HKD (Hong Kong dollar), HUF (Hungarian forint), ILS (Israeli shekel), JPY (Japanese yen), KRW (Korean won), MXN (Mexican peso), MYR (Malaysian ringgit), NOK (Norwegian krone), NZD (New Zealand dollar), PEN (Peruvian sol), PHP (Philippine peso), PLN (Polish zloty), RON (Romanian leu), RUB (Russian ruble), SEK (Swedish krona), SGD (Singaporean dollar), THB (Thai baht), TRY (Turkish lira), TWD (Taiwanese dollar), ZAR (South African rand).