See the latest Australian dollar analysis here:
From the excellent Sally Auld at JPM:
Adding a formal QE forecast to the outlook
After a few years of policy stability, 2019 delivered plenty of surprises on the monetary policy front. First was the RBA’s volte-face in February, when the Governor declared that risks to Australian monetary policy had become two-sided. This was followed by consecutive 25bp rate cuts in June and July (the first time the RBA had delivered back-to-back easings since 2012), and then a further 25bp easing in October (the last time the RBA eased policy in the month of October was also 2012).
In 2020, we expect that the policy environment will both tread familiar ground and breach new territory. We continue to expect the RBA to deliver a 25bp easing in February, taking the cash rate to 0.50%. On one level, this easing completes the 100bp of easing we believed was necessary to place a floor under core inflation and to deliver eventual stability in the unemployment rate (see RBA: Now forecasting a
terminal policy rate of 0.5% by mid-2020, S. Auld et. al., 24 May 2019 for more). Admittedly, the labour market outlook is perhaps less acute now than was the case in May when we first made the call for a terminal rate of 0.50%, but this is offset by the fact that the RBA has revised down its estimate of NAIRU since. Nonetheless, risks to the global outlook have shifted in a favourable direction of late and a firmly “on hold” Fed also takes some pressure off the RBA, given the influence of global monetary on domestic policy settings.
But it is worth remembering that the RBA’s current set of forecasts – which see the unemployment rate “stuck” at 5% by mid-2021 and core inflation “stuck” at 1.9% by end-2021 – assume “some chance” of a further rate cut to 0.50% in the first half of 2020. Fundamentals would need to improve markedly for this to be no longer required, and data we have so far for the 3Q GDP tracking exercise are a sobering reminder that the economy hasn’t reached escape velocity just yet. In addition, the prospect of currency strength will keep the RBA on notice, given the importance of currency stability or currency under-performance (preferable) to the broader economic outlook. In light of downward revisions to wages growth and a profile for core inflation that now forecasts a longer period of below target CPI, the November
Statement on Monetary Policy was a clear reminder that the RBA has more work to do. Moreover, it is increasingly clear that a further 25bp rate cut will be insufficient to achieve desired objectives.
Looking ahead to 2020, we now add a formal forecast of QE into our forecast for the RBA next year, on the basis that by year end, the central bank will acknowledge a requirement to do more to achieve its objectives. In 4Q 2020, we think the RBA will deliver 1) a 25bp rate cut to 0.25%; 2) an announcement to purchase AUD35-50bn of nominal ACGBs; and 3) continued use of forward guidance.
From a macro-economic perspective, our call assumes that the RBA should be planning to get core inflation toward the mid-point of the target band. If this is correct, it follows that more stimulus will be required. From a starting point of ~2.0% for core inflation in late 2020, our modelling suggests that around 50-75bp of additional easing is required (beyond our cash rate forecast of 0.50%) to get core CPI
convincingly inside the target band. Taking the ELB to be 0.25% (that is, one more 25bp easing), QE could make up the difference (25-50bp). Indeed, it was interesting that the November SoMP referenced the Board’s conclusion in the October meeting that “…each further cut brings closer the point at which other policy options might come into play.” While this statement can be interpreted in a number of ways, we think the reference to “other options” is significant and suggests that the Bank is perhaps prepared to move towards unconventional policies sooner than expected.
We have selected 4Q20 as the starting point for a QE program on the basis that by then, the RBA will have seen three sets of national accounts since the February rate cut, and will have a sense of whether the economy is sustaining a growth rate that is consistent with trend or better GDP growth. In order to render QE unnecessary, we think the RBA would need to be confident by 3Q20 that the economy has reached a point where quarterly GDP prints of 0.7%q/q are sustainable. Our own GDP forecasts are a little below this for 2020 (quarterly outcomes of around 0.6%q/q).
Our call on QE is based on the following assumptions.
First, that the RBA believes a package of measures to be more impactful. This was one of the key conclusions from the Bank’s study of other QE programs, and the benefits of this approach were outlined clearly by the ECB’s Chief Economist earlier this year:
“In assessing the impact of our policy package, we must also take into account that the instruments were designed to be complementary and mutually reinforcing. Overall, the transmission of our suite of measures to the financial system has been very effective. This is in no small part due to the strong complementarities between instruments, which ensure that all segments of the yield curve are influenced, thereby easing financial conditions more broadly.”
Second, that credit markets are functional at the time of implementation. This means that the RBA can purchase sovereign bonds and have confidence that stable to lower spreads will deliver lower yields on non-sovereign product too.
Third, the government delivers some modest fiscal easing in the May budget, but nothing that materially changes the outlook for growth or inflation.
Fourth, bank funding costs remain contained, negating the need for explicit programs supporting domestic banks (such as cheap funding, or increased provision of liquidity in the RBA’s OMOs).
Finally, that the global backdrop improves in 1H20 (as per the J.P. Morgan forecasts), allowing the RBA plenty of time to pause after the February rate cut before implementing QE later in the year.
Calibrating asset purchases
Asset purchases work by signaling future intent on monetary policy and by compressing risk and term premia. For the RBA, the former will likely be the focus, given that term premium is already negative in the longer end of the curve, credit markets are functional and curve spreads are relatively flat. In the past decade, most central banks have conducted asset purchases in terms of a fixed amount of purchases over a specific period, but occasionally they have been open ended, normally until some macro objective has been met. Only in Japan have we seen an explicit objective for longer-term yields.
Academic literature studying the impact of QE programs in the US and Europe suggests that purchases somewhere between 9-13% of GDP are required to deliver an easing that is consistent with 100bp of rate cuts.3 So one simple rule of thumb might be that the RBA needs to do around half of this (another 50bp of easing), to deliver a positive impulse to core inflation. Asset purchases equivalent to 5.5% of GDP would be worth around AUD30bn.
One firm conclusion of the academic research is that asset purchases need to be sizeable, in order to have an impact; indeed, empirical estimates for the US suggest asset purchases of around 9-13% of GDP are worth just +0.2% on core inflation.
These estimates show the inherent difficulty in getting a meaningful shift in macroeconomic variables without large scale asset purchases.
Another way to calibrate the level of asset purchases is to quantify the sensitivity of ACGB yields to shifts in demand (in this case, central bank purchases). We do this by estimating both supply and demand curves for ACGBs, and then assume a QE program that is executed uniformly across the curve (that is, equal dollar purchases\ across all maturity buckets).
Our analysis (summarized in Table 1) suggests that effects are larger on the front of the yield curve, relative to the back. If we assume that the objective of ACGB purchases is to deliver an equivalent 25-50bp in “cash rate space”, the size of the program then dictates the extent of the RBA’s commitment to sustaining low rates.
Lowering yields across the curve gives credibility today to the RBA’s claim they will eventually return inflation to levels consistent with the target. A larger program signals more urgency.
Our model (key results outlined in Table 1) suggests:
A$100bn of purchases across the curve (rounding to the nearest whole number) lowers 1Y rates by 140bps, 10Y rates by 70bps. We note these are risk neutral rates, to convey monetary policy expectations (see notes in Table 1), and would not be realized fully in market yields due to the ELB. While the effects may seem large (particularly for the front end), in reality we do not know what other central banks would have experienced had they executed their programs in Australianlike conditions; in most other cases of QE, policy rates were already at zero and financial markets were in stress upon introduction.
Rearranging, we can find the size of the total QE program required to deliver a 50bp rate reduction (per the above), at any given tenor. If the RBA wants to reduce 3Y-4Y rates by 50bp, they need to buy AUD45-50bn. If they are committed to a stronger signal, and want a 50bp reduction in 10Y rates, they need to buy AUD74bn.
Coincidentally, a 50bp reduction of 3Y rates would eliminate almost all of the current implied increase embedded in current market pricing (per the Treasury’s estimate of risk neutral rates). The current risk neutral rate for 3Y bonds is ~125bp; a spread of 50bp over the current cash rate (0.75%); see Figure 2. Keeping a flat path out to 3 years would thus be consistent with QE extending the guidance that rates are likely to be low for some time. If we use a 50bp reduction in 1Y-4Y yields as a starting point for sizing the QE program, this would require a program of AUD35- 50bn of purchases of nominal ACGBs, a little larger than the back of the envelope number above (~AUD30bn) but still under 10% of ACGBs outstanding.
Market impact: ACGB yields lower, credit spreads unchanged, steeper ASW curve and AUD/USD eventually lower
In this section, we make some brief and high level observations on the likely market impact of our QE call. For readers interested in more detail on this topic, we will explore the outlook for AUD rates and FX in 2020 at length in our forthcoming outlook publications.
If the RBA aims to signal a protracted period of unchanged interest rates, then lowering the risk neutral 3Y rate by around 50bp seems like a good place to start.
What happens to the shape of the curve beyond that will depend upon the exact framework that the central bank uses to determine purchases across the curve. In reality, we doubt the RBA would purchase dollar amounts evenly across the curve, given that the end objective might be to accumulate a portfolio with a given duration target, and/or that purchases could be likely weighted by issue size or some other factor. Generally speaking, history would suggest that QE programs have flattened yield curves (2s/10s), either because investors move further out the curve to obtain additional yield, or because programs have explicitly targeted longer dated yields (such as Operation Twist in the US, and YCC in Japan). A similar dynamic is likely in Australia.
Our underlying assumption of credit market functionality and the influence of a portfolio balance effect would suggest that yields on other debt should fall by similar magnitudes too, leaving longer end credit spreads broadly unchanged. But as we note below, externalities associated with QE may drive a sharper compression in front end swap spreads relative to longer end swap spreads, and so we are biased to think the ASW curve would steepen in a QE environment.
For currency impact, we refer readers to a research note we published on unconventional monetary policy in Australia in 2016 (see It’s never too soon to ask – what would unconventional monetary policy look like in Australia? S. Auld et. al., 31 August 2016). In that piece, we tested the sensitivity of the exchange rate (AUD/USD) to sovereign yield differentials (AUD less USD) at different maturity points on the curve. We estimated uncovered interest rate parity (UIP) conditions for AUD/USD, using weekly data on 1Y increments of ACGB par rates, starting in 2013. We matched ACGB yields with an equivalent tenor UST par rate to form the AUD-USD yield spread. Then we estimated simple univariate regressions of the level of AUD/USD on those spreads. Generally, we found that a 10bp narrower AUD-USD spread is – across the curve – consistent with AUD being lower by 1.5 to 3.0 US cents. For the 3Y tenor, a 50bp reduction in yields (holding UST yields constant) is worth 7c lower on AUD/USD.
If this decline seems large, then three caveats to the extent of FX response to QE may be worth considering. First, positioning and valuation are important, in that QE announcements usually have their greatest impact on the currency when it is overvalued and investors are long. This is not the case with AUD/USD at present.
Second, Australia’s fiscal and current account status is such that there is no longer an onerous funding task, potentially requiring less currency depreciation to offset a given decline in yields. Finally, as we have observed before (see Same, same but different: Trading unconventional monetary policy in the Antipodes, S. Auld et. al., 30 September 2019) the distribution of ACGBs across the foreign investor base may prevent liquidation of existing securities under a QE program, minimizing the currency impact.
Of course, the broader backdrop to Australian QE should support an idiosyncratic weakness for AUD/USD, simply because the RBA’s first foray into QE comes at a time when other central banks are closer to reaching the limit of such policies (ECB and BoJ), or no longer pursuing balance sheet expansion that is driven by a desire to expand the asset side of their balance sheet (Federal Reserve and BoE).
Positive externalities of QE
As the discussion above suggests, QE programs are often thought of in terms of their influence on macro-economic variables. But in Australia, we think there some positive externalities associated with a QE program that expands the RBA’s balance sheet via ACGB purchases. These externalities relate to the rise in ES balances that occur as a consequence of any QE program and what this potentially implies for banks’ participation in the domestic repo market. We first alluded to this dynamic in a research note earlier in the year (see Australian QE: Liquidity flows the same way down under, B. Jarman, 31 July 2019), noting that QE could also help to compress front end funding spreads.
As a starting point, we refer readers to a recent research note by our colleagues in US interest rate derivatives strategy, which investigates why, in a world of ample excess reserves held at the Fed, Banks were not able to respond to strong economic incentives when the overnight GC repo rate spiked towards 10% (see What is preventing banks from policing the repo market, J. Younger, 31 October 2019). The broad conclusion of this note is that banks’ desire to maintain conservative liquidity buffers relative to required reserve requirements, together with significant volatility in intra-day payment and settlement activity, have limited banks’ ability to tactically deploy cash to repo markets. In turn, this has kept front end funding spreads elevated for longer, ultimately requiring a liquidity injection from the Federal Reserve.
One could argue that regulatory requirements in Australia have also delivered a similar outcome, such that even a small QE program could have an impact on front end funding spreads. We observe that the correlation between short-run changes in ESA balances and the repo-cash spread is 0 (Figure 3). The same is true of the RBA’s repo quantity offering ($) vs the repo-cash spread. This supports the notion that banks do not lift their demand for cash when GC rates are elevated to redeploy into the repo market. Presumably this is because they strongly prioritize maintaining their liquidity buffers (the costs associated with not doing so are high, both financially and reputationally).
We also observe that shocks to ESA balances are strongly mean-reverting: 2/3 of shocks unwind within a month (Figure 4). The implication is that banks only need to worry about short-run shocks to liquidity, not sustained shocks. This means they should be very sensitive to any permanent increase in reserves that is sufficient to cover any foreseeable short-term liquidity shocks. Since the new framework for ESA balances was introduced in 2013, there has never been a 1 week withdrawal greater than A$6bn (Figure 5). The key point is that even a very small QE program would deliver sufficient excess ESA balances to remove any risk of banks being short funding at the end of the day. All else equal, this should help to narrow GC-OIS and FRA-OIS spreads. This is an important externality when rates approach their lower bound, especially given some positive correlation between the shape of the OIS curve and the GC-OIS spread (Figure 6).
A word of warning
There is a degree of discomfort around forecasting QE in Australia. Some of this stems from the fact that the RBA has been tight-lipped on the topic, and so we have very limited knowledge of the Bank’s views on what structure of QE would deliver optimal outcomes in Australia. While this doesn’t by any means prevent us from trying, it does somewhat reduce the conviction level in our characterization of unconventional policy.
Another source of discomfort is the Governor’s view that returning inflation to the mid-point of the target is not an end in itself; rather, it is a broader welfare maximization objective that ultimately determines the extent to which the RBA will pursue policies designed to boost inflation. We have assumed in the above analysis that the RBA ultimately does want to get core inflation towards the middle of the target band, but we are wary that this might not reflect the Bank’s preferences, particularly if trade-offs with other objectives (such as financial stability) become more acute.
In the end, the path of inflation expectations will ultimately determine the Bank’s trajectory on this issue, as it represents a medium term binding constraint on the RBA’s flexibility around the inflation target. Various measures of inflation expectations have been trending lower of late (Figure 7); the RBA will be watching to see whether recent rate cuts can achieve stabilization or lift in these measures.
Another caveat concerns fiscal policy. On this front, our expectations are low given 1) we are early in the political cycle; 2) the government is comfortably ahead in the polls; 3) we have a forecast of stabilization and modestly better macro-economic outcomes in 1H20, both globally and locally, and 4) there is little historical precedent for pre-emptive fiscal easing (both in Australia and in other DM economies).
But in the event we were to be surprised, and the government did deliver a fiscal package which was strongly supportive of growth (for example, by bringing forward the stage 2 personal income tax cuts), then there would be little need for a QE package in 4Q20. Instead, QE would be more likely to be delivered as a response to the next global recession, rather than as a response to more chronic shortfalls on inflation and unemployment.
Conclusion – less is not more
It is clear from the November SoMP that current policy settings are unable to effect macro-economic outcomes that are consistent with the RBA’s objectives. Indeed, a forecast set that largely assumes another 25bp easing in 1H20 (on the back of 75bp of easing so far) still reflects an unemployment rate above NAIRU and core inflation outcomes that are inconsistent with the inflation target (given almost four years of below target outcomes for core CPI already, with another two forecast). This leaves the door wide open for the RBA to move towards the next phase of easing, via unconventional monetary policy. The Governor’s speech on 26 November, titled “Unconventional Monetary Policy: Some Lessons from Overseas”, will be an important marker in gauging how close the RBA is to breaching the realm of unconventional monetary policy.