Australian federal election predictions

Rank punditry versus efficient markets

May 20

We previously made the case for efficient markets over rank punditry as the basis for election predictions. With Australia’s federal election held tomorrow, I thought we should re-visit the market-implied probabilities. My former boss, Professor Simon Jackman, has done the coding so we don’t have to. Here is his summary of the main outcomes:

This is how we got here:

There is no shortage of model-based forecasts which, strictly speaking, should be in the betting market price, but are in any event to be preferred over rank punditry. If you think in pays to be Bayes, here is Mark the Ballot’s model. Other model-based forecasts include Armarium Interreta, Buckleys and None, and AEF. There is a certain consistency to the predictions, although they condition on similar information sets.

We will check-in next week to see how the markets and models did. There is, of course, no definitive empirical test of market efficiency due to the model selection problem, although betting market prices pass standards tests for a random walk.

If you want to do a compare and contrast with the rank punditry, you will have to follow Andrew Leigh’s lead and see if you can track down some falsifiable pundit predictions. Andrew’s research suggests a 13/20 pundit strike rate derived from 32 hours of Insiders and Meet the Press transcripts. Pity the research assistant who read them all. Note that betting market forecasts are not only efficient in the technical sense of discounting available information, they are also efficient in the sense of economising on the consumption of punditry. A few lines of code are probably more informative than dozens of column inches.

We also suggested that changes in financial market prices provide the best gauge of the economic implications of the election outcome. For the record, today’s prices are:

AUD-USD 0.7016

10-year bond yield 3.31

5-yr bond yield 2.99

2-yr bond yield 2.45

June 22 IB futures 99.45

ASX200 7145

There are no ABS data releases scheduled for Monday, so we should get a reasonably clean read from the markets then. I would suggest benchmarking against the Friday close for US markets. No significant change in financial market prices implies no partisan differences in macroeconomic outcomes, at least in expectation.

There is a lot of schadenfreude out there about the turmoil in crypto markets but my sense is that, as a risky asset on the periphery of global markets, crypto is the canary in the coal mine of the global business cycle. There was a time when crypto traded counter-cyclically, reflecting its origins as an alternative asset that was meant to provide a refuge from the depredations of government, including capital controls.

More recently, the positive beta on crypto is actually a good sign that it was being mainstreamed and I still think that is the case. DLT is a great innovation with a lot of applications that will ultimately depend on some form of digital currency. The global payments system has been a giant hold-out from digital innovation and is ripe for transformation. A lot of crypto business models will hit the wall before the market sifts through what works and what doesn’t. My view is that we will end up in a world of CBDCs and stablecoins with bank-like regulation, which will probably be sub-optimal in lots of ways, but will still capture many of the prospective benefits of crypto. I think BTC is sufficiently differentiated that it survives in the long-run, but have no view on its long-run relative price.

The other sign of stress in the global financial system is high-yield spreads, one of the real-time inputs into our US labour market models. Volatility-implied high yields spreads are quite a bit higher than actuals. That’s not entirely new, but no less concerning for that. Before indulging in too much schadenfreude over crypto, you want might to check on the resilience of your own portfolio.

The minutes of the May RBA Board meeting were interesting for their reflection on the implications of the effects of QT:

The minutes seem to characterise the effects of QT as operating in parallel with the increase in the cash rate, in contrast to Lowe’s claim at the post-meeting press conference that QT was already in the price. We could of course make exactly the same claim about future increases in the cash rate also being in the price, but I doubt Lowe would state that quite so explicitly. The seeming disconnect between the characterisation of the effects of QT between Lowe’s press conference and the minutes I think speaks to Lowe’s aversion to thinking about monetary policy in terms other than the cash rate.

What matters is how the RBA’s policy instruments evolve relative to these expectations. The run-off in the bond portfolio is perhaps more predictable than future changes in the cash rate as things stand, but there will always be some residual uncertainty about the future size of the RBA’s balance sheet. Even a small negative shock at this point would likely see the RBA expanding its balance sheet again. Assistant Governor Chris Kent will be speaking next week on the topic of ‘From QE to QT – The next phase in the Reserve Bank's Bond Purchase Program,’ which may yield a bit more insight into how the RBA is thinking about QT, but I think that will be more in terms of implementation that macroeconomic implications.

The first quarter Australian wage price index came in 0.7% q/q and 2.4% over the year. This was a little weaker than the 0.8% q/q and 2.5% y/y implied by our model, but still the fastest wages growth at an annual rate since the December quarter 2018. The RBA’s May Statement on Monetary Policy is forecasting wages growth at 2.7% by the middle of 2022 and 3% by the end of the year. Wages growth is forecast to peak at 3.7% for the year-ended June 2024. For all the angst about falling real wages, the real producer wage (COE/hour worked, GDP deflated) is still behaving very much in line with long-run productivity trends.

It is interesting that the RBA has started referencing nominal (non-farm) average earnings per hour as one of its wages benchmarks, which was running at only 3.3% for the year-ended 2021. As we have suggested previously, growth in national accounts-based measures of labour income can be viewed as a proxy for nominal GDP that removes some of the volatility associated with movements in the terms of trade. The RBA’s forecasts have this measure spiking to 6% by mid-year, before falling back to 4.4% by the end of 2022. This is only a little above the 3.8% long-run trend growth rate for compensation of employees per hour worked, which has a stable long-run relationship with nominal GDP. We can only hope the increased interest in these measures is a steppingstone to nominal income targeting.

ICYMI