RDP 2022-05: The Real Effects of Debt Covenants: Evidence from Australia 5. Aggregate Effects of Covenant Composition

The differential responses by covenant configuration beg the question: has the effectiveness of monetary policy changed due to changes in the structure of covenants over time? According to Figure 3, the prevalence of ICC relative to other types of covenants has declined over the past decade. As such, the role of covenants in amplifying monetary policy transmission post-GFC might have been dampened by the shift away from ICC towards NICC. Moreover, the overall usage of covenants has increased, which could also affect transmission.

In this section, I quantify the effect of the shift in the composition of covenants by using the coefficients from the full sample regression in Equation (4) and information on the composition of covenants to capture the average partial response to a monetary policy shock. I examine how this has changed over time to quantify how changes in composition affected monetary policy pass-through. Specifically, I compute the average partial responses^[3] of business activity in each year of the data as follows:

That is, the average partial response in time t is the sum of coefficients ${\beta }_{1,Cov}^h$, estimated for each covenant configuration using Equation (4), across configurations and weighted by the composition of covenants in the previous year (as a share of all firms). I then compare the average partial response in any given year with that in 2003.

Focusing first on investment, Figure 8 shows that the responsiveness of investment to a monetary policy shock has declined due to shifts in the covenant composition. The response of business investment one year after a 100 basis point monetary policy shock is estimated to be 0.35 percentage points smaller in 2020, compared to the case where the composition of covenants remained unchanged. To put this in context, the RBA's macroeconometric model MARTIN predicts that business investment declines by about 1.3 per cent over one year following a 100 basis point cash rate increase (Ballantyne et al 2020). As such, the shift in the composition of covenants is estimated to have lowered the responsiveness of firms' investment by around 15 to 25 per cent.

Focusing on non-mining firms only, I find the estimates to remain virtually unchanged (Figure F1), suggesting that changes in the composition of covenants could potentially account for part of the weakness in non-mining investment observed over the 2010s (Debelle 2017; Lowe 2018; Dynan 2021). Applying the results from my sample to the full economy seems reasonable, given my sample is made up of large firms, and large firms account for a large proportion of non-mining investment (Dynan 2021).

The compositional change in covenants includes both the increased reporting of covenants and the shift from ICC to NICC within firms subject to covenants. To the extent that the lower share of NC firms reflects the trend in reporting standards rather than a true increase in covenant exposure, the previous calculation may overstate the effects of the compositional change over time. Therefore I also take a more conservative approach in estimating the effect by keeping the share of NC firms unchanged from the level in 2003 and just allowing the ICC and NICC split within firms with covenants to vary. Results suggest that the responsiveness of investment to monetary policy shock is 0.2 percentage points smaller in 2020 (Figure F2).

Similar results are found for staff expenses, with the average response to a monetary policy shock after one year being 0.4 percentage points lower in 2020, compared to the case where the composition of covenants remained unchanged. However, total staff expenses in the sample account for less than 10 per cent of total private sector labour costs according to the national accounts. As such, it is a stronger assumption that the changes can be extrapolated to the broader economy. However, to the extent that the same shifts have occurred amongst non-listed firms, this might still suggest some moderate aggregate effects.

One potential concern is that the decline in reported ICC relative to NICC might have been driven by a lower breaching rate of ICCs due to the low interest rate environment in Australia. In such a case, there may not be an underlying shift in the composition of covenants. However, the rate of breaches amongst firms with ICC has increased over the past decade, and at a much faster pace compared with NICC firms. Secondly, having removed breaching instances and keeping only firms routinely reporting covenants, the shift in the composition of covenants away from ICC towards NICC remains apparent (Figure G1). As such, there is no evidence of the trend being driven by breaches. Nevertheless, it is possible that the increase in the relative NICC share reflects changes in the reporting standard that caused the overall rise, to the extent that they disproportionately affected the NICC. In that case, the counterfactual analysis results should still be treated as an upper bound on the effect of the shifting composition of covenants.

More broadly, it is important to note that the estimates take a relatively simple, partial equilibrium approach to understanding the effect of changes in the composition of covenants. As such they should be thought of as indicative, not precisely estimated. Still, they do suggest that increased use of NICC could help explain part of the weakness in non-mining investment over the 2010s.

Footnote

Impulse responses of business activity to monetary policy shocks, abstracting from the transmission of shocks via firms' characteristics. [3]