RDP 2007-07: More Potent Monetary Policy? Insights from a Threshold Model 1. Introduction

The household sector's balance sheet has transformed substantially over the past two decades. The ratio of household debt to disposable income in Australia has risen rapidly from an average of about 40 per cent in the late 1980s to around 160 per cent in 2006 (Figure 1). It has been argued that higher levels of household debt may make the economy more sensitive to a given change in interest rates through its impact on cash flows. However, this effect may have been offset somewhat by the trend decline in business-sector leverage over a similar period, so that the net effect on the interest rate sensitivity of aggregate demand is unclear.

Figure 1: Household and Corporate Debt Indicators

In regard to the cash-flow channel of the transmission mechanism, the impact of monetary policy on aggregate demand may depend on more than just the level of indebtedness.[1] Of potentially greater relevance is the existence of credit-constrained borrowers. If borrowers are severely credit-constrained – that is, they have exhausted all available borrowing opportunities – they will tend to adjust spending substantially in response to even a temporary change in disposable income, including via a change in interest rates. In contrast, unconstrained borrowers (who base their consumption decisions on their expected permanent income) need to adjust their behaviour much less in response to temporary changes in disposable income, since these can be accommodated via an adjustment in saving (that is, via a change in net borrowing). It is easy to see that monetary policy could be more potent in cases where credit constraints are more pervasive. For example, higher interest rates raise the cost of servicing household debt, thereby transferring resources from borrowers (who have a high marginal propensity to consume, particularly if they face credit constraints) to lenders who are likely to have a lower marginal propensity to consume. Aggregate consumption is therefore likely to fall by more than if borrowing constraints were modest. Analogously, a similar story could be told about business investment. Bearing this in mind, this paper asks whether there is evidence that the interest rate sensitivity of the Australian economy depends on credit conditions.

The paper attempts to examine the sensitivity of economic activity to monetary policy by estimating the impact of interest rate shocks on output, conditional on the growth rate of credit. Credit growth provides a simple and readily available indicator of credit conditions. It is used in preference to a measure of indebtedness,[2] since credit constraints could be just as likely to be binding when indebtedness is high or when it is low.[3] Of course, credit growth will depend on the strength of both demand and supply, which means that some caution is needed when interpreting the results.

One way to do this is to identify various episodes when supply considerations were clear drivers of credit growth. One obvious example is financial deregulation, which affected both financial institutions and borrowers. First, there was a removal of interest rate controls in the 1980s. Second, overseas borrowing controls were relaxed and restrictions prohibiting Australian financial institutions from borrowing overseas and restrictions on foreign banks entry were removed. These reforms enhanced Australian households' and firms' ability to access credit by spurring competition as banks sought to expand their lending operations to maintain or expand market share (see, for example, Edey and Gray 1996). As indicated in Figure 2, household and business sector credit growth have moved in slightly different phases over time. Following deregulation in the second half of the 1980s, business credit growth rose significantly, boosted also by sharp increases in prices of commercial property, which was an important source of collateral at the time (Gizycki and Lowe 2000). Business credit fell during the recession in the early 1990s, but here too, supply constraints played a role. Heavy losses on commercial loans by the banks, coupled with a large and prolonged decline in commercial property prices, led to tight credit conditions for businesses extending beyond the recessionary period (Tallman and Bharucha 2000). Moreover, this period coincides with financial institutions reducing their appetite for risk; some even announced explicit goals of reducing their exposure to business lending. For the household sector, credit conditions remained relatively tight for most of the 1980s. It was not until after the 1990s recession that these constraints eased substantially and household credit growth was maintained at a high rate for an extended period. This easing in constraints was associated with increased competitive pressures in the market and a persistent decline in inflation (see Edey and Gray 1996). As well as examining a model with aggregate credit growth and GDP, this paper separately analyses business and household sectors to take advantage of their different experiences in credit conditions over time.

Figure 2: Real Household and Business Credit Yearly Growth Rate

Credit constraints nowadays feature prominently in economic analyses of short-term business-cycle fluctuations. There is considerable macroeconomic literature suggesting that business cycles and credit interact (see Blinder 1987, Bernanke and Gertler 1989, Kiyotaki and Moore 1997 and Azariadis and Smith 1998). Although the theories presented in these studies differ in various dimensions, they all imply that the responsiveness of the economy to shocks might depend on prevailing credit conditions.

There is some indirect empirical evidence supporting the idea that recessions are likely to be periods when borrowers' balance sheets are weak and the availability of credit is tight. The result, that monetary policy shocks have a greater effect on output during recessions than during expansions, appears to hold across a number of countries for a number of different time periods: see for instance, Peersman and Smets (2001) for the euro area; Sensier, Osborn and Öcal (2002) for the United Kingdom; and Garcia and Schaller (2002) and Lo and Piger (2005) for the United States. Moreover, Garcia, Lusardi and Ng (1997) estimate the relationship between consumption and income for households in the US that are identified as credit-constrained and those that can borrow more freely, and find asymmetries in consumption behaviour. Cover (1992) provides a related finding that an expansionary monetary policy has less effect on output than a contractionary policy of the same order of magnitude.

In this paper I test for the presence of asymmetries by employing a threshold model.[4] This nonlinear model endogenously divides the sample into high and low credit-growth regimes and allows for different dynamic responses in each regime. The two regimes are delineated according to whether the moving average of the rate of growth of real credit is above or below an estimated critical threshold; values below the estimated threshold are assumed to represent periods of relatively tight credit and vice versa. The approach is closest in spirit to recent studies by Balke (2000), Atanasova (2003) and Calza and Sousa (2006). Balke concludes that there is evidence of threshold effects in the relationship between measures of credit conditions and economic activity in the US. Atanasova finds similar threshold effects in UK data, as do Calza and Sousa for the euro area. These studies, however, focus only on aggregate credit indicators, whereas this paper also distinguishes between household and business credit indicators.

To preview the findings, it appears that GDP growth in Australia is more responsive to interest rate shocks in the low credit-growth regime. This result is confirmed in separate models for household and business credit growth: consumption is more responsive to interest rate shocks when household credit is growing slowly, and real investment responds similarly in the low business credit-growth regime.

The remainder of the paper is organised as follows. Section 2 describes the econometric methodology. Section 3 presents the empirical specification, explains the test for threshold effects, and generates nonlinear impulse responses functions from the estimated model to examine whether there are any asymmetries. Section 4 concludes.


There are, of course, a number of other ways in which interest rates affect the economy besides the cash-flow channel. For instance, interest rates affect the price of financial assets, such as bonds and shares, and the exchange rate, which affect households and businesses in a variety of ways. [1]

Indicators of indebtedness will tend to split the sample into two parts given that these have typically trended up for household and down for business sectors over the past two decades. [2]

A number of other alternatives have been used in the literature. Balke (2000) identifies ‘tight’ credit regimes as periods when the spread between commercial paper (four to six months) and 6-month Treasury bill rates is abnormally high. He also analyses alternative credit condition proxies such as the mix of bank loans and commercial paper in total external finance, and the difference between the growth rates in the short-term debt of small and large manufacturing firms. Atanasova (2003) considers the 10-year corporate bond spread. The idea being that a high spread on these rates is indicative of periods where firms' financing costs are high. These indicators are less useful for the analysis of both business and household credit conducted in this paper. [3]

Previous studies using Australian data (with some measure of credit) have had a somewhat different focus, aiming to match linear models to the data; see, for instance, Suzuki (2004) and Berkelmans (2005). [4]