Updating Australia’s Business Cycle Reference Dates
When the de-trend is not your friend
This week’s release of the Q4 national accounts is a good excuse to update a previous post, in which I applied the NBER approach to business cycle dating to Australian real GDP. The classical Burns-Mitchell approach to business cycle dating aligns very nicely with nominal GDP level targeting (NGDPLT), which would have monetary policy stabilise the level of nominal GDP around a long-run target growth path. The classical approach to business cycle dating is also consistent with NGDPLT in not requiring us to remove a stochastic trend from the data in order to identify the business cycle.
I apply the Bry-Boschan-Pagan-Harding business cycle dating algorithm to Australian real GDP since Q3 1959. The resulting business cycle reference dates and metrics are shown below.
The algorithm identifies the final quarter of 2019 as the business cycle peak, implying the recession ‘began’ in 2019. Similarly, note how the early 1990s recession ‘began’ in Q2 1990. Contractions (recessions) start at the peak of a business cycle and end at the trough. We shouldn’t make too much of this nuance, but it underscores the fact that Australia was already primed for a recession in 2020 on the back of the Q1 bushfires and associated disruption, even before the pandemic hit. Using monthly data, the NBER says the US recession began in February 2020. The most recent Australian recession ‘ends’ with the trough in Q2 2020, but as of Q4 2020, real output is still 1.1% below its previous peak in Q4 2019, much less a no-pandemic counterfactual growth path.
One thing that stands out from this analysis is the extraordinary longevity of Australia’s expansions, with a mean duration of 30 quarters, although the most recent expansion is obviously an influential observation in that calculation. When Adrian Pagan and Don Harding proposed this approach in a 2002 paper using data through to Q1 1997, the mean duration for Australian expansions was 20.6 quarters, with contractions only slightly longer than shown above at 3.3 quarters.
In contrast to the classical approach to business cycle dating focused on levels, ‘gap’-based methods would have us believe that the output gap is all but closed. For example, Kamber, Morley and Wong’s ‘nowcast’ of their Beveridge–Nelson decomposition based on a mixed-frequency Bayesian VAR has the US output gap for Q4 2020 at only -0.4%, while Q1 2021 is nowcast at +0.8%. A positive output gap would have a forward-looking Fed already tapping the breaks in a conventional Taylor rule framework.
Nowcast of US output gap using the Beveridge–Nelson decomposition based on a mixed-frequency Bayesian VAR
As noted last week, real yields have done much of the heaving-lifting relative to inflation expectations in raising nominal bond yields in recent weeks. Joe Gagnon and Madi Sarsenbayev find that bond yields have not been a great predicator of inflation in the past, although this is just another way of saying that markets have been placing bets against the secular downtrend in interest rates for a long time now. Even some of the bonds funds that rode that downtrend to considerable fame and fortune in previous decades have bet against that trend more recently.
Reconceptualising the business cycle in terms of levels rather than growth rates (which is itself a form of de-trending) might just lead to less biased forecasts on the part of everyone concerned.
ICYMI
Heritage Foundation drops Hong Kong from its Index of Economic Freedom, where it previously occupied first place, on the basis it is no longer an independent jurisdiction.
Peter Costello’s Future Fund is ranked 10th in the Peterson Institute’s Sovereign Wealth Fund scorecard, below the State Oil Fund of the Republic of Azerbaijan.
The FT editorialises against the RBNZ’s new housing remit.
Your tweets:
Create your profile
Only paid subscribers can comment on this post
Check your email
For your security, we need to re-authenticate you.
Click the link we sent to , or click here to sign in.