RDP 2022-01: MARTIN Gets a Bank Account: Adding a Banking Sector to the RBA's Macroeconometric Model 6. Conclusion

We have built an extension to MARTIN that incorporates one of the key financial accelerator mechanisms – a banking sector that endogenously and nonlinearly changes credit supply in response to loan losses and/or changes to their funding costs. By combining a large and complex macroeconometric model with a micro-simulation model, and nonlinear stress testing and funding costs frameworks, our approach moves beyond the existing macroeconometric frontier.

We have shown that the ability to increase lending rates across new and outstanding loans – as is the case in Australia – provides a powerful capital restoration mechanism that means credit supply does not need to be reduced as considerably as if banks were confined to restricting new loans only. This finding highlights the importance of having a model that is designed specifically to capture the features of the Australian banking system (instead of relying on the experiences of stressed foreign banking systems), and has important implications for assessing the potential macroeconomic costs of banking system stress in Australia.

To ensure the features of the Australian system are captured as best as possible, an important avenue for future research is the explicit incorporation of business loan losses into BA-MARTIN (Section 3.2 explains why this is not yet possible). In recent history, business loans have been a larger driver of banks' losses than household loans (Rodgers 2015). By calibrating our model to APRA's stress testing results, we are able to capture business loan losses only to the extent that they are correlated with household loan losses. Our current approach is therefore far from ideal.

Two questions we have begun exploring in this paper are: ‘How can the banking sector amplify macroeconomic shocks in Australia?’ and ‘How can the banking sector change the pass-through of monetary policy?’ But the results presented here are only the beginning. The COVID-19 downside scenario from the May 2020 Statement on Monetary Policy is just one scenario out a wide range of plausible sources of stress. To properly evaluate how the banking sector can amplify macroeconomic shocks, a much wider range of possible shocks should be assessed.

Our analysis of how the banking sector can change the pass-through of monetary policy is similarly narrow. We assume the economy starts from a benign economic situation before it is ‘shocked’ in a way that stresses the banking sector and requires monetary policy to respond. The results are promising; monetary policy is more effective precisely when it needs to be. But this may not always be the case.

The amplified pass-through occurs solely due to the ability of policy to reduce losses. If the banking system begins in a stressed state, the fact that lower interest rates slow the capital restoration process (by decreasing NIMs and increasing credit demand) may dominate the beneficial effect of reducing further losses. In extreme cases, it may be that a lower cash rate slows capital restoration by so much that banks' endogenous credit supply responses more than offset the macroeconomic effect of the cash rate cut. The point at which this occurs is known as the ‘reversal rate’ (Brunnermeier and Koby 2018). Investigating the conditions that may give rise to a reversal rate in Australia is an important avenue for future research.

Having a banking sector within a macroeconomic model permits assessment of the potential complementarities or trade-offs between the RBA's inflation and full employment objectives and its financial stability objective. BA-MARTIN also provides an additional tool to help calibrate macroprudential policies. For example, a countercyclical capital buffer could be modelled as a temporary decline in banks' capital target $\left(\overline{e}\right)$, which would reduce the capital shortfall that banks try to make-up during the period of peak stress. Alternatively, APRA could give banks relief with respect to how quickly they expect capital to be restored following a crisis (i.e. reducing the speed of adjustment parameter $\left(\lambda \right))$.^[31] In fact, by helping banks coordinate on a lower $\lambda$ when losses are at their highest, APRA would be fulfilling the classic regulatory role of reducing negative externalities (see Section 3.8.2).

We look forward to seeing how BA-MARTIN is used in the future, but hope that its accuracy is never tested in reality.

Footnote

A pre-emptive example of such a policy was APRA's announcement that, following COVID-19, ‘its future expectations for capital will allow banks to rebuild their capital buffers in an orderly manner’ (RBA 2020a). [31]