RDP 2023-09: Does Monetary Policy Affect Non-mining Business Investment in Australia? Evidence from BLADE 3. Methodology

Gulnara Nolan, Jonathan Hambur and Philip Vermeulen

December 2023

Our analysis closely follows Durante et al (2022), who analyse the heterogeneous effects of monetary policy shocks on European firms' investment. We employ a local projections approach (Jordà 2005) of the following form to trace out the effect of monetary policy shocks (shockt) over a number of horizons (h = 0,...,12).[3]

I i,t+h = β h shoc k t + α h I i,t−1 + ∑ j=1 1 γ h X t−j + ν i,t+h

where Ii,t is our measure of investment (IDi,t or Invi,t). When focusing on the extensive margin, the model is effectively a linear probability model and can be interpreted as the effect on the share of firms investing, or the probability of investing. When the intensive margin is used, we are considering the percentage change in investment for those firms continuing to invest from the quarter of the shock.

The vector Xt–j is a set of control variables, including one lag of firm-level revenue growth, GDP and CPI growth, as well as industry and firm size and age (specifically an indicator if the firm was above or below five years old).[4] As the shocks are exogenous, controls are not strictly needed, but they help improve the precision of our estimates, particularly the firm-level controls. We cluster the errors for each period, allowing cross-sectional correlation across firms. This is important given the variable of interest is the same across all firms for a given period (Cameron and Miller 2015). We do not allow for serial correlation as this is addressed by the lagged variables.

Our measure of monetary policy shocks is the measure constructed in Beckers (2020), which covers the sample period up to December quarter 2018. This is a Romer and Romer (2004) style shock that measures the shock as a deviation from a Taylor Rule, augmented with measures of financial conditions and financial market participants' expectations.

In particular, Beckers (2020) estimates an augmented Taylor rule that includes forecasts for economic conditions, as well as a number of indicators of financial conditions (e.g. bond spreads, option-implied volatility). The shocks are then constructed as the deviation of the actual policy rate from that implied by the rule. He produces two main measures: a preferred measure that also accounts for market expectations for the policy rate, and another version that does not. We use the preferred measure for our analysis, though the results are almost identical using the other measure.

We choose the Beckers measure as our preferred measure as it has been shown to resolve the so-called price puzzle in Australian data: that contractionary monetary policy is often estimated to lead to higher prices. Other measures, including those constructed from high-frequency changes in bond yields, are also explored below.

We allow the shock to enter the model directly, similar to Durante et al (2022), rather than using it as an instrument for changes in the cash rate. As such, we are implicitly taking the measure to be a true estimate of the shock, rather than a noisy estimate. That said, the results are reasonably robust to using the shock as an instrument for cash rate changes, which allows for the possibility that the Beckers shock is a noisy proxy for the true shock, though the estimated effects are larger and there is slightly less evidence of significance on the intensive margin (see Figure C6).

Footnotes

For the analysis using BAS data we only use information out to horizon 12 given the slightly shorter sample. For the rest of the analysis we use data out to horizon 16. [3]

The results are generally robust to including more lags of the right-hand side variables, as well as including the contemporaneous controls which imposes the implicit assumption that monetary policy cannot affect current conditions (Ramey 2016). [4]