Home » Research Blog » Merchandise import as predictor of duration returns

Merchandise import as predictor of duration returns

Jupyter Notebook

Local-currency import growth is a widely underestimated and important indicator of trends in fixed-income markets. Its predictive power reflects its alignment with economic trends that matter for monetary policy: domestic demand, inflation, and effective currency dynamics. Empirical evidence confirms that import growth has significantly predicted outright duration returns, curve position returns, and cross-currency relative duration returns over the past 22 years. A composite import score would have added considerable economic value to a duration portfolio through timing directional exposure, positioning along the curve, and cross-country allocations.

The below post is based on proprietary research of Macrosynergy. This post ties in with this site’s summary of the importance of macro trends.

A Jupyter notebook for audit and replication of the research results can be downloaded here. The notebook operation requires access to J.P. Morgan DataQuery to download data from JPMaQS, a premium service of quantamental indicators. J.P. Morgan offers free trials for institutional clients.

Also, there is an academic research support program that sponsors data sets for relevant projects.

The theory

We define nominal merchandise imports as the local-currency value of all goods imported into a currency area from the rest of the world over a certain period. These flows are reported in customs trade statistics and balance of payments and are released shortly after the observed period with publication lags of a few days up to 6-7 weeks. Nominal import growth is not widely or intensely watched by market participants, judging from published research and financial news. That is because it is not formally part of the monetary policy reaction function and does not directly affect the value of many liquidly traded assets.

However, local-currency import values happen to be related to three key macro trends that do have an impact on local monetary policy:

  • Import value growth depends positively on aggregate domestic demand growth.
  • Import value growth depends positively on local inflation and relevant international tradable goods inflation.
  • Import value growth depends negatively on changes in the effective international value of the local currency.

For example, rapid nominal import growth typically indicates some combination of strong domestic spending and inflationary pressure. In accordance with the theory of rational inattention, we posit that nominal import growth has predictive power on subsequent returns in the fixed-income markets in the following ways:

  • Reported import growth over and above a neutral rate is negatively related to subsequent duration returns across all maturities. Here we measure duration returns as the returns on interest rate swap fixed receiver positions.
  • Reported import growth over and above a neutral rate positively predicts the returns on curve flattening positions in the 2-year to 5-year space. This reflects that excess import growth mainly affects the prediction of monetary policy rather than the long-term structural state of the economy and hence has a greater effect on shorter maturities.
  • Relative import growth over and above a neutral rate of one currency area versus a basket of others negatively predicts relative duration returns of that area versus the others. This hypothesis implies that import growth excesses are broadly comparable across the chosen set of markets.

The data

For a meaningful analysis of the impact of local-currency import trends on market returns, we use the related indicators of the J.P. Morgan Macrosynergy Quantamental System (“JPMaQS”). Quantamental indicators are real-time information states of the market and the public concerning an economic concept and, hence, are suitable for testing relations with subsequent returns and backtesting related trading strategies. In particular, we choose three standard quantamental import growth metrics based on seasonally- and calendar-adjusted series:

  • % of latest 3 months over previous 3 months at an annualized rate
  • % of latest 6 months over previous 6 months at an annualized rate
  • % over a year ago, 3-month moving average.

The underlying data vintages have been taken from external merchandise trade statistics of either the balances of payments or customs trade statistics. Seasonal adjustment factors are sequentially re-estimated as new data are released. Every re-estimation gives rise to a new data vintage, the basis of concurrent trend estimation. For complete documentation of the quantamental import trend series see the downloadable indicator notebook here.

For subsequent empirical analysis, we transform the import growth rates in two steps:

  • We subtract the 2-year or 5-year annualized nominal IRS swap rates from annualized import growth, depending on what maturity of duration exposure we use as a target. The yields serve as a metric for variations in the “neutral” rate of nominal import growth across countries. In equilibrium, nominal yields are related to nominal activity growth in an economy, and this serves as a benchmark for whether nominal import growth is considered high or low.
  • We normalize the differences between nominal import growth and yields by using the median and standard deviation of the entire data panel (time series across 25 currency areas) up to the day for which the normalization is performed. Thus, we create panel-based z-scores based on expanding windows, capping absolute values at three standard deviations. This normalization is helpful because it provides statistical information on the indicator’s “neutral” value: import growth historically has exceeded nominal GDP growth or nominal yields, plausibly a reflection of globalization. Using z-scores also makes it easier to calculate a composite import growth score as an average of the individual growth metrics.

The targets of the present analysis are 2-year and 5-year interest rate swap receiver returns for 25 countries with reasonably liquid markets: AUD, CAD, CHF, CLP, CNY, COP, CZK, EUR, GBP, HUF, IDR, ILS, INR, JPY, KRW, MXN, NOK, NZD, PLN, SEK, SGD, THB, TWD, USD, and ZAR. For an explanation of the currency symbols see the annex below.

In particular, we look at vol-targeted returns, i.e., returns on receiver positions scaled at the end of each month to produce a 10% annualized return on the deployed risk capital based on historical standard deviations for an exponential moving average with a half-life of 11 days. For more details see the downloadable indicator notebook here.

Predicting IRS receiver returns

The first hypothesis is that excess import growth predicts duration returns negatively. Plausibly, this should be relevant for the whole curve. However, concurrent import trends inform more on the monetary policy outlook than long-term growth and inflation and, hence, should be a better predictor for two years than five years.

Panel analysis for 25 countries from 2002 (or from the earliest observation available) to May 2023 confirms the significance of the predictive power of the composite import growth score for subsequent monthly or quarterly IRS receiver returns for either the 2-year or 5-year maturity. Higher import growth has heralded lower or negative duration returns.

We assess the significance of this relationship based on the Macrosynergy panel test. It checks the significance through a panel regression model with period-specific random effects. The model adjusts targets and features of the predictive regression for common (global) influences. The stronger these global effects, the greater the weight of deviations from the period-mean in the regression. In the presence of dominant global effects, the significance test relies mainly upon its ability to explain cross-sectional target differences. Thus, the method automatically accounts for the similarity of experiences across markets (a type of “pseudo-replication”) when assessing the significance and, hence, can be applied to a wide variety of target returns and features.

The panel tests show a near 100% probability of the historical relation between reported import growth and subsequent returns not being accidental for both 2-year and 5-year IRS receiver returns. The monthly balanced accuracy of monthly 2-year IRS return predictions (the average ratio of correct positive and negative return predictions) has been 53.9%. Indeed, positive precision (57.2%) and negative precision (50.5%) have been above par, meaning that positive and negative return predictions have been correct more than half the time.

The predictive correlation of the import growth score with subsequent returns has been lower for the 5-year maturity than the 2-year contract. This is consistent with import growth being mainly an indicator of short- and medium-term economic fluctuation rather than structural changes.

We calculate naïve PnLs based on standard rules consistently used in Macrosynergy posts. This objective is an analysis of the concept, not optimization. Positions are taken based on import scores in units of vol-targeted positions. The z-scores are winsorized at three standard deviations to reduce the impact of data outliers. Positions are rebalanced monthly with a one-day slippage added for trading. Transaction costs are disregarded because they depend heavily on the value of assets under strategy management. The long-term volatility of the PnL for positions across all currency areas has been set to 10% annualized for ease of presentation.

The 22-year Sharpe ratio of a strategy based on the composite import score has been very high at 1.4, with no correlation to the S&P500 returns. However, the performance has also been very seasonal, with the most value generated in the 2020s, benefiting from the economic fluctuations related to the pandemic. Looking across different import growth signals, the least volatile annual growth rates produced the highest prediction accuracy and PnL value.

An essential feature of the imports-based PnL has been its negative relation with the long-duration portfolio during sell-offs. This pattern manifested in the period of Fed tightening in the second half of the 2000s, the 2017-18 period, and the post-pandemic inflation-related sell-off. This means that imports-based strategies would have functioned well as an overlay of a long-duration portfolio, not just adding performance but offsetting all larger drawdowns.

One of the drawbacks of the imports-based directional trading signal is that international trade is correlated across all larger countries and hence the diversification benefit of trading a range of countries is limited. Also, the strategy needs economic turbulence to get large signals from import growth. This fosters seasonality.

An imports-based duration strategy for 5-year maturity contracts would have only produced half the value and did so predominantly in the 2020s. Longer-date maturity receivers carry higher risk premia and displayed a greater ratio of positive return months (54.6%), making it harder for a short-biased strategy to create value independently. However, like the 2-year maturity, the imports-based signal would have highly complemented a long-only strategy.

Predicting curve flattening returns

The hypothesis is that strong import growth predicts swap curve-flattening through a more hawkish monetary policy stance in the near term. This hypothesis relies on the transitory nature of import growth information and the expectations that central banks defend their inflation targets in the long run. Here we define a curve flattening return as the difference between the returns of 5-year and 2-year IRS vol-targeted receiver positions.

Empirical panel analysis confirms a positive relationship between the composite import growth score and subsequent monthly or quarterly curve flattening returns. The forward correlation coefficient is comparable in size to outright duration return predictions. However, the Macrosynergy panel test only assigns 83% significance to the quarterly relation, reflecting the high cross-country correlation of curve-based returns.

Predictive accuracy for curve flattening returns has been similar in value to directional 2-year returns, with balanced accuracy at 53.9%.

The naïve standard PnL for a curve position portfolio across all 25 markets displays material economic value, with a long-term Sharpe ratio of 0.84 before transaction costs and an S&P500 correlation of nearly zero. Like the directional strategy, the imports-based curve strategy has performed well in economic turbulences.

Predicting relative IRS returns across currency areas

The hypothesis is that countries with stronger imports will subsequently experience lower duration returns than those with weaker import growth. Differentials in local-currency import growth are seen as indicative of differences in growth and inflation and, hence, for resultant differences in changes in monetary policy. The monetary policy links suggest that predictive power should be stronger for relative 2-year IRS receiver returns than for relative 5-year IRS receiver returns.  The term “relative” here refers to the value of the local currency area minus the average value for all tradable currency areas at the given time period.

Again the hypothesized relationship is born out by the empirical evidence for the 25-country panel for 2002-2023 (May). Relatively strong local import growth predicts relatively weak local duration returns for 2-year or 5-year IRS contracts. The Macrosynergy panel tests suggest that the relationship is highly significant with a probability of nearly 100% for both maturities at either monthly or quarterly frequency.

The balanced accuracy of the prediction of monthly return directions has been lower than in the case of directional signals. For the 2-year relative IRS receiver returns it has been at 51.4%. Lower precision is a common feature of relative economic signals because data are not fully comparable across countries and a part of the signal reflects irrelevant differences in data conventions and economic structure.

Notwithstanding lower accuracy, the economic value of the relative import growth score trading signal has been considerable with a long-term Sharpe ratio of 1 and no equity market correlation. The strategy has been less seasonal than those based on directional signals. The consistent and robust performance of relative import growth as a trading signal reflects the diversification benefits of relative versus directional positions. The flip side is that this strategy requires higher leverage than a directional portfolio for the same return target, translating into higher transaction costs.

A relative duration exposure strategy based on the 5-year maturities would have produced a little less value, with Sharpe 0.9, and with greater seasonality.

Annex: Currency symbols

In alphabetical order, the currency symbols and their meanings are AUD (Australian dollar), BRL (Brazilian real), CAD (Canadian dollar), CHF (Swiss franc), CLP (Chilean peso), CNY (Chinese yuan), CZK (Czech Republic koruna), GBP (British pound), HUF (Hungarian forint), IDR (Indonesian rupiah), ILS (Israeli shekel), JPY (Japanese yen), KRW (Korean won), MXN (Mexican peso), MYR (Malaysian ringgit), NOK (Norwegian krone), NZD (New Zealand dollar), PHP (Phillipine peso), PLN (Polish zloty), SEK (Swedish krona), SGD (Singaporean dollar), THB (Thai baht), TWD (Taiwanese dollar), USD (U.S. dollar), ZAR (South African rand).



Related articles