In 2017, the RBA cash rate was just 1.5 per cent. Today it is 4.6 per cent.
Look only at those numbers and monetary policy today appears vastly tighter. But that comparison misses an important part of monetary transmission in a small open economy: the exchange rate.
A higher Australian dollar makes imports cheaper and Australian production relatively more expensive. It lowers imported inflation and shifts demand away from domestic production. A lower dollar works in the opposite direction.
The difficulty is that the Australian dollar moves for many reasons besides monetary policy. Commodity prices are particularly important. A high dollar during a commodity boom does not necessarily tell us that Australian monetary policy is unusually tight.
So can we separate the two?
I estimated a simple long-run relationship between Australia's real trade-weighted exchange rate and commodity prices. The residual gives us a way of asking a useful question: is the Australian dollar unusually expensive or cheap given the commodity prices Australia faces?
The answer produces an interesting history of Australian monetary conditions. It also helps explain how a cash rate of 1.5 per cent could have been too tight in 2017, while a cash rate of 4.6 per cent today does not, by itself, tell us how tight monetary policy is.
Finding an anchor for the dollar
The starting point is a simple relationship:
\[ \ln q_t = \alpha + \beta \ln C_t + \varepsilon_t \]where \(q_t\) is Australia's real trade-weighted exchange rate and \(C_t\) is the RBA's commodity-price index measured in US dollars.
The data are quarterly. The real TWI comes from the RBA's F15 exchange-rate series. The commodity-price index comes from RBA table I2, with the monthly observations averaged to quarters.
I use the real, rather than nominal, TWI because I am interested in the relative price of Australian goods and services, not simply the number of units of foreign currency purchased by an Australian dollar. Over a period of more than three decades, differences in inflation rates matter.
I use commodity prices measured in US dollars for a different reason. Using the Australian-dollar commodity-price index would put the exchange rate mechanically into the explanatory variable. That would partly build the relationship I am trying to estimate into the data itself.
The sample runs from 1993 to June 2026.
The estimated coefficient on commodity prices is:
\[ \hat{\beta} = 0.28 \]So a 10 per cent increase in commodity prices is associated with roughly a 2.8 per cent appreciation in Australia's real TWI.
For a model containing a single explanatory variable, the fit is remarkably strong:
\[ R^2 = 0.82 \]Commodity prices alone account for about 82 per cent of the variation in Australia's real trade-weighted exchange rate over the sample.
That does not, by itself, establish a meaningful long-run relationship. Two persistent economic time series can produce a high \(R^2\) even when there is no stable relationship between them.
So I also tested for cointegration. The Engle-Granger test gives:
\[ p = 0.010 \]The Engle-Granger test rejects the null of no cointegration at conventional significance levels. Although the real TWI and commodity prices both wander considerably through time, their estimated linear combination appears stationary.
That is what makes the residual interesting.
Measuring when the dollar is unusually dear or cheap
The estimated residual is:
\[ \hat{\varepsilon}_t = \ln q_t - \left( \hat{\alpha} + \hat{\beta}\ln C_t \right) \]I convert this into a more intuitive measure:
\[ Gap_t = 100\hat{\varepsilon}_t \]Because this is a log difference, the result can be interpreted approximately as the percentage by which the real TWI is above or below the level implied by commodity prices.
A gap of +5 means the real Australian dollar is about 5 per cent dearer than its commodity-price relationship would suggest. A gap of -10 means it is about 10 per cent cheaper.
The standard deviation ($\sigma$) of the residual over the sample is about 7.1 per cent.
In the chart I leave observations within 0.75 standard deviations, about 5.3 per cent, unshaded. These are relatively ordinary deviations from the historical relationship.
Larger positive gaps are shaded red. The Australian dollar is unusually dear and the exchange-rate channel is adding to monetary tightness.
Larger negative gaps are blue. The dollar is unusually cheap and the exchange-rate channel is offsetting monetary tightness.
The darker areas mark particularly large deviations.
This is deliberately an exchange-rate conditions measure, not an estimate of the overall stance of monetary policy.
That distinction matters.
Why leave interest rates out?
A conventional model of the real exchange rate might include relative real interest rates as well as commodity prices.
Schematically, we might write:
\[ \ln q_t = \alpha +\beta\ln C_t +\gamma(r_t-r_t^*) +u_t \]where \(r_t-r_t^*\) represents the Australian real interest rate relative to the relevant foreign rate.
But I have deliberately left the interest-rate differential out.
Why?
Because it is partly what I want to see.
If the fuller relationship above were the correct model, then the residual from my commodity-only equation would contain something like:
\[ \hat{\varepsilon}_t \approx \gamma(r_t-r_t^*) + u_t \]The first term is exactly the monetary-policy mechanism of interest. If Australian rates rise relative to foreign rates, Australian-dollar assets become relatively more attractive. Other things equal, that tends to raise the exchange rate.
Removing commodity prices first therefore gives us a way of seeing exchange-rate movements that cannot readily be explained by Australia's commodity cycle.
But the second term is important too.
The residual also contains changes in risk appetite, global capital flows, expectations and anything else affecting the Australian dollar that is not captured by commodity prices.
So red does not mean "the RBA was running tight monetary policy", and blue does not mean "the RBA was running loose monetary policy".
It means the exchange rate was adding to, or subtracting from, monetary tightness.
The distinction is easy to see during the Global Financial Crisis. In late 2008 the real TWI fell to around 22 per cent below the level implied by commodity prices. That enormous blue observation was not simply the result of easy RBA policy. Global risk aversion and capital flows were moving violently.
But whatever caused the fall, a much cheaper Australian dollar was nevertheless easing Australian monetary conditions through the exchange-rate channel.
That is what this measure is designed to capture.
The 2010s look different through this lens
The most striking part of the chart is the 2010s.
The RBA progressively reduced the cash rate. It fell from 4.75 per cent in 2011 to 1.5 per cent by 2016.
Looking only at the cash rate, monetary policy became steadily easier.
But the exchange-rate channel tells a different story.
For much of the decade, the real Australian dollar remained expensive relative to the level suggested by commodity prices. The chart is persistently red.
That meant the exchange rate was working against the easing delivered through lower mortgage rates, cheaper business finance and the other domestic interest-rate channels.
This becomes particularly interesting around 2017.
The cash rate was just 1.5 per cent. In retrospect, monetary policy at the time looks too tight relative to economic conditions. Inflation remained below target and unemployment was above estimates of full employment.
Yet a 1.5 per cent cash rate sounds extraordinarily low by today's standards.
The exchange rate helps reconcile those facts.
The Australian dollar was unusually dear relative to its commodity-price anchor. The exchange-rate channel was adding materially to monetary tightness even while the nominal cash rate was sitting at what then looked like an exceptionally low level.
A low cash rate did not necessarily mean loose monetary conditions.
Then the relationship reversed
The pandemic produced almost the opposite configuration.
The cash rate was cut to essentially zero, and the Australian dollar became unusually cheap relative to commodity prices. Both the domestic interest-rate channel and the exchange-rate channel were therefore pushing in an expansionary direction.
More interestingly, the exchange rate remained relatively cheap through much of the subsequent tightening cycle.
The RBA began raising the cash rate, eventually delivering hundreds of basis points of tightening. For mortgage borrowers, the change was enormous.
But the exchange-rate channel did not initially reinforce that tightening to anything like the same extent.
By early 2023, the real TWI was around 12 per cent cheaper than its commodity-price relationship suggested.
So the RBA could be raising the cash rate aggressively while one important monetary transmission channel was still working in the opposite direction.
Again, the cash rate alone did not describe the stance experienced by the economy.
What is it saying now?
The picture has changed substantially.
From around 12 per cent cheap in early 2023, the real TWI had moved to about 5 per cent dear by June 2026. That is a swing of roughly 17 percentage points relative to the commodity-price benchmark.
At the same time, the cash rate has reached 4.6 per cent.
The June exchange-rate gap was still within the normal range of the model, but it was approaching the levels seen around 2017.
That comparison is revealing.
In 2017, the cash rate was 1.5 per cent and the exchange rate was adding materially to monetary tightness. In 2026, the cash rate is 4.6 per cent, but by June the exchange-rate gap was approaching the range observed in 2017.
This does not mean monetary policy today is only as tight as it was in 2017.
The mortgage channel is obviously very different. Household debt is repriced against a much higher cash rate. Business borrowing costs, saving incentives, asset prices and credit conditions also matter.
It means something narrower, but important:
Radically different cash rates can coexist with surprisingly similar degrees of tightness through the exchange-rate channel.
That is one reason the level of the cash rate cannot, by itself, tell us whether monetary policy is tight or loose.
How robust is the relationship?
A natural concern is that a relationship estimated across more than three decades may depend heavily on where the sample begins.
Australia's economy changed substantially over this period. The mining boom changed the importance of commodity exports. Commodity pricing arrangements changed. Financial markets deepened. China's importance to Australian trade increased dramatically.
So I re-estimated the model using different starting dates.
The result is reassuring.
Moving the beginning of the sample from 1983 through to 2002 leaves the estimated commodity-price elasticity in a relatively narrow range of about 0.22 to 0.28.
The estimated latest exchange-rate gap remains between roughly +4.8 and +6.0 per cent.
And the Engle-Granger cointegration test continues to reject at the 5 per cent level across those specifications.
That does not prove the coefficient has been structurally constant. Nor does it establish that this is the unique model of Australia's equilibrium real exchange rate.
But the central result does not appear to be an artefact of choosing 1993 as the starting point.
What the model does not tell us
There are several reasons not to push the model further than it deserves.
First, zero is not some deep estimate of the Australian dollar's economic equilibrium value. The regression contains a constant, so the residual is centred on the historical relationship over the estimation period. "Fair", in this context, means close to what commodity prices would historically have implied, not an estimate of purchasing-power parity or a fundamental equilibrium exchange rate.
Second, the model uses the full sample. The historical gaps therefore contain information that would not have been available to a policymaker in real time. Re-estimating the equation as new observations arrive can revise the historical relationship.
Third, the coefficient on commodity prices is fixed. A 0.28 elasticity is imposed on the whole sample even though the structure of the Australian economy and commodity markets has changed.
Fourth, this is a static long-run equation. It does not estimate how quickly the exchange rate adjusts after commodity prices move. Some apparent gaps may therefore reflect differences in timing rather than genuine departures from the long-run relationship.
Fifth, cointegration provides evidence for a stable long-run relationship, but it does not tell us the speed at which deviations correct. Estimating that would require an error-correction model.
Finally, the residual is not a monetary-policy shock. Interest-rate differentials are in it, deliberately, but so are risk sentiment, capital flows and other omitted influences.
Those limitations are why I call the result an exchange-rate conditions measure, rather than a monetary-policy index.
The cash rate is the instrument, not the whole stance
Central banks set interest rates. Economies respond through transmission channels.
Mortgage payments are one. Business borrowing is another. Asset prices, saving incentives, bank credit and expectations all matter.
For a small open economy such as Australia, the exchange rate matters too.
And the exchange-rate channel depends not simply on whether the RBA raises or lowers rates, but on Australian rates relative to those elsewhere, together with commodity prices and global financial conditions.
That produces configurations that can look strange if the cash rate is treated as synonymous with monetary policy.
A cash rate of 1.5 per cent can coexist with an exchange rate adding substantially to tightness.
A rapidly rising cash rate can coexist with an unusually cheap exchange rate offsetting some of that tightening.
And a cash rate of 4.6 per cent can coexist with an exchange-rate channel that, at least by June, remained within its historically normal range.
The chart therefore isn't an alternative measure of the overall stance of monetary policy.
It is a reminder that the stance cannot be read directly from the policy instrument.
The cash rate tells us what the RBA has done. The exchange rate helps tell us how some of that policy, together with the rest of the world, is actually reaching the Australian economy.
No comments:
Post a Comment