The curious case of the commodity currencies (Part 1)

For decades, terms of trade signals were reliable predictors of the direction and scale of commodity currency movements. Not so in the 2020s ...

WHEN I BEGAN my career as Chief Economist of BHP, the world’s largest mining company, roughly nine years ago, iron ore was around $60 per tonne and AUD/USD was around 80¢. When I left BHP a few months ago, the iron ore price was around $100, and the AUD was below 70¢.

For those who have spent any time around commodity currencies and their dynamics, there is clearly something odd about that pair of observations. The price of Australia’s largest export increased by two-thirds and the AUD … lost value?

Not just AUD

Looking further afield, it turns out that this was not an Australian specific anomaly. It was a common change in dynamics across commodity currencies. In New Zealand, despite milk prices rising about 50% over that same period, the NZD depreciated around 8%. In Canada, a 92% rise in crude oil prices was met with a mere 2% appreciation of the CAD. In Chile an 81% uplift in the copper price was accompanied by a 20% depreciation of the CLP.

These simplistic paired observations encouraged me to delve deeper into the evolution of the dynamics of this currency group [1]. Has the breakdown of the traditional relationships been happening incrementally for some time, and the pandemic and its inflationary aftermath just delivered the coup de grace to the ancien regime? Or is the anomaly readily explainable by special factors that are likely to reverse – and if so, what sort of adjustment path would that entail in coming years? How did monetary policy contribute to the observed disconnect? Or is this observation in commodity currencies a symptom of something deeper within the global macro complex with fundamental implications for broader asset markets and capital flows? Or is the non-technical concept of “pandemic weirdness” to blame, and time will resolve all problems?

That is a lot to unpack - hence this morphed into a two-part piece. Part one will present the empirical evidence for the relationship breakdown below. Part two will canvas the various narrative explanations for this phenomenon and speculate upon future adjustment paths based on stylized beliefs regarding the durability of the changes.

To pre-empt potential disappointment, I will table at the outset that I have not reached a single high conviction conclusion on these questions, partly because it is simply too early to tell if full blown regime change has occurred. Instead, I will present conditional assessments and future watchpoints for three stylized positions: (a) secular regime change, (b) mean reversion to the old relationships as “pandemic weirdness” fades. (c) a hybrid world where assessing shape-shifting short-term dynamics becomes a more important skill in these markets than the ability to accurately predict future rate gaps and commodity prices.

A linear time series model

Empirically, my approach has been to create linear time series models of three bellwether commodity currencies representing the three main value chains of energy, metals and food. As it happens, they are also from the Anglo-sphere: AUD, CAD and NZD, the 5^th, 6^th and 10^th most activity traded currencies respectively, prior to the pandemic. This has the advantage of having each currency backed by broadly comparable institutions, open borders in terms of both trade and capital, full deliverability of FX, respectable levels of market turnover, and mature financial markets beyond the FX element. Further, there is nothing so idiosyncratic in their recent history to overtly pollute the full sample or sub-samples^[2]. These guardrails led to the exclusion of CLP, COP, ZAR, RUB, PEN and BRL, among other commodity exporters, due to a failure on some combination of the conditions specified above (e.g. Chilean social unrest and constitutional uncertainty, Russian sanctions, South African electricity and corruption crises etc etc).  

The models are specified as simply as possible for ease of comparison, with the USD nominal bilateral rate (expressed as USD per target currency – so CAD flips from the conventional quotation) regressed on an appropriate terms of trade proxy deflated by US CPI, US short rates^[3], a risk proxy and a GFC dummy. Over my 20+ years in the markets, models of this general type have been a workhorse technology in the FX strategy community, explaining three-quarters or more of the variation in commodity currencies from the mid-1990s forward^[4] and have been reliable tools for assessing valuation outliers. Periodic re-estimation tended to smooth out any wrinkles in the short run, with the basic framework consistently revalidated. Indeed, Rogoff formally stated in a milestone 2002 paper that his famous random walk critique (along with Richard Meese) of fundamental models of exchange rates needed explicit qualification for the same Anglo-sphere subset we are considering today.

Each model was estimated using monthly data over 5 overlapping sample periods: 1996 to 2023; 2004 to 2023; 1996 to 2019; 2004 to 2019; and 2016 to 2023 (the era of Brexit and Trump).

In line with the high-level observations at the outset of the piece, the models imply very strongly that something has clearly changed in the way FX markets are incorporating terms of trade signals into AUD, CAD and NZD in recent years versus prior decades. Each sub-sample tested that captured the period from 2020 onwards sees a considerable deterioration in model fit (see Table 1) and diagnostics from those sub-samples that precede 2020, and a reduction in the explanatory power of commodity pricing – and in the shortest sample (isolating the post Brexit/Trump era), the terms of trade proxy is rendered statistically insignificant for both CAD and NZD (Table 2), and it declines in significance for the AUD. The degradation in fit is considerable – shifting from consistently reliable to coin-flip territory (bold row in Table 1). The US interest rate signal though retains normal explanatory power in the 2016-2023 period for Canada and it increases its direct influence on the Antipodean pair.

(For table 2, note again that all rates are quoted as “US dollar per unit of commodity currency”, which is the norm for AUD and NZD, but the inverse of the conventional quotation of the CAD. A positively signed coefficient means that an increase in that variable appreciates the target currency, and an increase in a variable with a negative coefficient depreciates it.)

Table 1. Goodness-of-fit by sample period.

Table 2. Terms of trade (ToT) proxy and US short rates: estimated coefficients and statistical significance by sample period.

This increasing influence for US interest rates though is not enough to offset the declining explanatory power of commodity prices. It is as if a previously invisible factor (or factors) has suddenly loomed into view that is now strong enough to disrupt the traditional relationships (blowing out the overall model residuals and compromising the statistical significance of other key variables). Or was the “invisible” factor known all along, it was just ignored or assigned an economic value insignificantly different from zero? Could this be another instance of a dark matter analogy fitting the economic debate?  Let’s see.

Chart 1. Model residuals (1996-2023 full sample coefficients).

There was a period in the early 2000s when residuals were sounding a similar alarm in terms of the magnitude of AUD and NZD undervaluation/USD overvaluation, although the CAD has never seen anything like the present episode (Chart 1). However, the commodity price backdrop in the early 2000s was very different, with energy and metals prices very weak at the time (Chart 2), supporting depressed relative valuations for the commodity complex, so one can make the case this was a simple case of overshooting to the downside.

Chart 2. Four bellwether commodity markets (US CPI deflated)

US equity market exceptionalism and associated US dollar bullishness characterized this historical period (the emerging market crises of the late 1990s intersecting and overlapping with the dotcom bubble) but this was unravelling well in advance of the maximum negative residuals attained for AUD and NZD (the Nasdaq peaked in March/April 2000, but the peak in the USD real effective index came a year later). With US equity market (and growth) exceptionalism again a feature in the current phase, this is potentially an important theme in the secular regime change versus transitory debate.

The role of interest rates in the two periods is another curiosity. Weak commodity prices and commodity currencies in the early 2000s coincided with an enduring negative rate spread (foreign minus USD). We have a similar flat to negative spread in rates today – but once again, a vastly different commodity backdrop.

Chart 3. Short interest rate differentials to the US

Conclusion of part one

The range of evidence described above points to a clear change in commodity currency dynamics in recent years versus prior decades. Part two of this piece will introduce and debate some of the narrative explanations that have emerged to account for this phenomenon. It will then draw out implications for future price formation dynamics and longer-run adjustment processes based on various stylized beliefs of how durable the observed changes may prove to be.

---

^[1] The BIS’ Daniel Rees investigated the positive correlation between a rising USD and rising commodity prices in the early 2020s, concluding that this was a response to the rising US terms of trade under its enhanced energy balances post the shale boom that occurred in the first half of the 2010s. A broader study by the same institution looked at the stagflationary impact of this correlation on oil importing regions. This piece looks at the issue from the perspective of commodity exporters that have a much greater leverage to the resources trade than the United States. 

^[2] The case could be made that the RBNZ’s experiment using a monetary conditions index as the operational target for monetary policy from 1997 to 1999 is such a pollutant. The subsequent empirical work detailed in the rest of the blog implies that if this is working as a distortion, it has a minor impact on the problem at hand.

^[3] Whether to use a spread, the local and US rates separately, or to use different points on the curve (e.g. 2-year swap spreads) to capture the interest rate signal, is something that consumes modellers’ time, but does little to drive additional fundamental insight. It has been noted within the modelling community that a standalone US rate signal can provide greater explanatory power when the Fed is aggressive, but that impact should logically be diluted somewhat over long samples. 

^[4] There was a time when the external balance sheet and/or government finances were an influential driver of these currencies (AUD “banana republic” crisis in the 1980s, Canadian public finance correction in the 1990s), but these variables have fallen out of favour over time.

Huw McKay is a Visiting Fellow at the Crawford School of Public Policy, ANU. His research merges economics, macro-finance, geostrategy, climate, energy and resources, technology and societal trends into a coherent whole. His recent book The Strategic Logic of China’s Economy has been described as “the definitive analysis of the socioeconomic transition of modern China and its precarious journey into the future."

The content in this piece is partly based on proprietary analysis that Exante Data does for institutional clients as part of its full macro strategy and flow analytics services. The content offered here differs significantly from Exante Data’s full service and is less technical as it aims to provide a more medium-term policy relevant perspective. The opinions and analytics expressed in this piece are those of the author alone and may not be those of Exante Data Inc. or Exante Advisors LLC. The content of this piece and the opinions expressed herein are independent of any work Exante Data Inc. or Exante Advisors LLC does and communicates to its clients.

Exante Advisors, LLC & Exante Data, Inc. Disclaimer

Exante Data delivers proprietary data and innovative analytics to investors globally. The vision of exante data is to improve markets strategy via new technologies. We provide reasoned answers to the most difficult markets questions, before the consensus.

This communication is provided for your informational purposes only. In making any investment decision, you must rely on your own examination of the securities and the terms of the offering.  The contents of this communication does not constitute legal, tax, investment or other advice, or a recommendation to purchase or sell any particular security.  Exante Advisors, LLC, Exante Data, Inc. and their affiliates (together, "Exante") do not warrant that information provided herein is correct, accurate, timely, error-free, or otherwise reliable. EXANTE HEREBY DISCLAIMS ANY WARRANTIES, EXPRESS OR IMPLIED.