RDP 2012-02: The Role of Credit Supply in the Australian Economy 3. A Proxy for Credit Frictions

While the mechanisms through which credit frictions might influence the real economy have been set out from a theoretical perspective, determining the empirical importance of these frictions is challenging. Quantifying the real effects of credit frictions requires identification of exogenous shifts in credit supply, which reflect financial stress, separately from shifts in credit demand, which can reflect other factors. Finding an appropriate variable that isolates shifts in credit supply – particularly for Australia – is difficult:

  • Interest spreads on ‘speculative grade’ corporate bonds have been identified in the literature as a reasonable proxy of the external finance premium and hence credit frictions (Bernanke and Gertler 1995; Gertler and Lown 2000). However, the corporate bond market in Australia has deepened only since the late 1990s and remains dominated by highly-rated institutions.[8]
  • An alternative is to use interest spreads on bank lending. Bank lending is the predominant source of business debt – particularly for smaller firms which are more likely to be credit constrained. However, these spreads are volatile prior to 1990. Moreover, measures of credit spreads might reflect demand conditions and fail to capture changes in the non-price terms of credit contracts, which can be an important aspect of shifts in credit supply.

Instead, our analysis makes use of a variable that has been virtually unused in previous Australian studies – a survey-based measure of the difficulty obtaining business finance.[9] The data are from the ACCI-Westpac Survey of Industrial Trends, the longest, continuous running private-sector survey in Australia. The survey began in 1960, and is conducted quarterly across a wide range of Australian manufacturing firms. It includes the following question, which has been unchanged since June 1966: [10]

‘Do you find it is now harder, easier, or the same as it was three months ago to get finance?’

Figure 3 shows the net balance of responses to this question – that is, the percentage of firms reporting more difficulty obtaining finance over the past three months less the percentage of firms reporting less difficulty obtaining finance.

Figure 3: Difficulty Obtaining Finance

This survey variable offers several key advantages:

  • It has been widely interpreted as an indicator of credit supply in other research, including Lowe and Rohling (1993) and Park (2011). This interpretation is important for our analysis, and is discussed in more detail in Section 3.1.
  • It should capture both price and non-price changes in credit supply. For example, firms should recognise both an increase in lending rates and a change in lending standards (such as increased collateral requirements) as contributing to finance becoming more difficult to obtain.
  • It has a long history, covering a number of business cycles. This history compares favourably to other surveys conducted both in Australia and internationally, most of which commenced in the 2000s.[11]

The ACCI-Westpac survey includes around 300 firms each quarter. The division of respondents between small/medium-sized firms and large firms has remained broadly unchanged over time. Typically, the survey contains more firms employing 1–200 people (around 60 per cent of the sample) than those employing more than 200 people (around 40 per cent). The share of respondents across different manufacturing sub-industries has also remained broadly unchanged over time.[12]

3.1 Difficulty Obtaining Finance as a Gauge of Credit Supply

As noted, the ACCI-Westpac survey variable has been broadly interpreted as an indicator of credit supply in previous research. Accordingly, we interpret innovations to the series as ‘credit supply shocks’ in order to identify the impact of credit supply on real activity. Given its importance, we explore this interpretation further below.

We can see that the variable behaves in a manner over the business cycle that is consistent with broad narratives on developments in credit supply. Respondents report greater difficulty obtaining finance during periods of recession in Australia (and the United States), and less difficulty obtaining finance during periods of economic expansion (Figure 4). Recessions typically result in a deterioration in borrower creditworthiness, prompting lenders to reduce exposures to riskier assets and retract credit supply. Two important episodes of tight credit supply that have been well documented are the early 1990s and the global financial crisis period.[13]

Figure 4: Difficulty Obtaining Finance

If the survey variable was instead driven by cycles in the demand for credit the opposite pattern would be expected over the business cycle. That is, a recession would reduce the demand for credit, and the resulting decline in competition for funds would therefore reduce the difficulty of obtaining finance.

Since the period of financial deregulation in the early–mid 1980s, the availability of credit has been determined largely by lenders allocating funds on the basis of commercial considerations, such as borrower creditworthiness. In contrast, the availability of credit prior to financial deregulation reflected a different set of factors. The Australian financial system was subject to various regulatory controls over the price and quantity of credit, and banks were generally reliant on deposits for funding so that asset growth was constrained. One outcome of these constraints was that increased demand for credit might go unmet, resulting in increased difficulty obtaining finance (see Battellino and McMillan (1989), Lowe and Rohling (1993)).[14] In other words, the survey indicator of the difficulty obtaining finance is likely to be a less reliable measure of changes in credit supply prior to the early–mid 1980s.

As an illustration of this, Table 1 calculates the average of GDP growth, business credit growth and a measure of business confidence during periods of increasing difficulty obtaining finance, and compares these averages to those for periods of decreasing difficulty obtaining finance. Prior to deregulation, increasing difficulty obtaining finance tended to be associated with periods of stronger economic growth, stronger credit growth and improving business confidence, whereas the opposite has been the case in the period since deregulation.[15]

Table 1: Difficulty Obtaining Finance and Economic Conditions(a)
Average of economic conditions over the past 12 months
Before deregulation pre-1984(b) After deregulation 1984–2011
Increasing difficulty Decreasing difficulty Difference Increasing difficulty Decreasing difficulty Difference
GDP growth 4.1 1.9 2.1   2.7 3.8 −1.1
Business credit growth 19.4 14.8 4.6   9.5 9.9 −0.4
Business confidence 14.3 −1.7 16.0   −11.8 1.1 −12.9

Notes: (a) Year-ended growth in GDP and business credit; average over the past year of business confidence as measured by the ACCI-Westpac survey; difficulty obtaining finance is adjusted for a structural break in the mean in September 2002, as discussed in Section 4.2.
(b) Sample periods are based on longest available series: GDP growth (1966); business credit growth(1976); business confidence (1966).

Sources: ABS; ACCI-Westpac; RBA

While the behaviour of the survey variable strongly suggests that since deregulation it has been dominated by developments in credit supply, we cannot rule out some influence of credit demand. For example, a positive shock to credit demand might see borrowers increase leverage, creditors tighten lending standards and pricing, and borrowers report greater difficulty obtaining finance. Nonetheless, by restricting our analysis to the post-deregulation period, we expect the influence of credit demand factors on the survey variable to be of a second order of importance.

3.2 Other Caveats in the Use of the Survey Variable

Notwithstanding its advantages, there are also a number of limitations in using the survey variable. Most importantly, the ACCI-Westpac survey covers only manufacturing firms. As a result, we are only assessing financing conditions for one component of total investment. Credit frictions are also likely to operate through other components of business and household investment as well as consumption and trade (Pagan and Robinson 2011).

However, to the extent that credit conditions are correlated across sectors, the ACCI-Westpac variable should act as a proxy for credit frictions more broadly. This correlation is apparent when we compare the ACCI-Westpac variable to broader information about credit conditions, such as in the National Australia Bank's Monthly Business Survey (although this is only available more recently).[16] The ACCI-Westpac variable also exhibits a relationship with international surveys of credit conditions, reflecting the globalised nature of capital markets (Figure 5).[17]

Figure 5: International Credit Conditions Surveys

Nonetheless, credit conditions are unlikely to be perfectly correlated across sectors, and the variable is likely to reflect some factors unique to the Australian manufacturing sector. In particular, over the past few decades there has been a structural shift in the economy away from the manufacturing sector, exacerbated in recent years by the mining boom and associated sharp rise in the exchange rate. Manufacturing profits as a share of total profits (excluding manufacturing and mining) have continued to decline over the past decade, and heightened competitive pressures during this period may have contributed to tighter financing conditions in the manufacturing sector relative to the rest of the economy (Figure 6). This appears evident in the ACCI-Westpac variable, with the net balance of firms reporting difficulty obtaining finance trending upwards from the early part of the 2000s. For this reason, we test and control for a structural break in the survey variable, which is discussed in more detail in Section 4.2.

Figure 6: Manufacturing Sector Indicators

Finally, as with any survey, there are also potential issues in interpreting responses to the survey question. In particular, past work has suggested that respondents may report the level rather than change in difficulty obtaining finance (Lowe and Rohling 1993; Park 2011).

3.3 Related Literature

Previous papers have used a range of techniques to identify and estimate the importance of credit supply shocks, including: sign- and zero-restricted structural VARs (Halvorsen and Jacobsen 2009; Meeks 2009; Busch, Scharnagl and Scheithauer 2010; Helbling et al 2010; Tamási and Világi 2011); DSGE models (Queijo von Heideken 2009; Gilchrist and Zakrajšek 2011); and panel regressions using firm and industry level data (Terrones, Scott and Kannan 2009; Becker and Ivashina 2010). Most of these papers use financial indicators such as credit quantities, prices, spreads, default rates and/or net worth to identify structural shocks to credit supply.

More closely related to our research, a small number of papers have used survey data to identify credit shocks. Bassett et al (2010) use the Federal Reserve's quarterly Senior Loan Officer Opinion Survey (SLOOS) to create a measure of credit supply shocks. The authors regress banks' individual responses to changes in their lending standards on a range of bank-specific and macroeconomic variables.

The residuals from this panel regression are then aggregated and treated as an exogenous credit shock series in a VAR model of the economy. They find that a one standard deviation tightening in lending standards results in a decline in real GDP of 0.4 per cent in the first year after the shock. A tightening in lending standards also results in an increase in the credit spread and a decline in the federal funds rate.

In a similar vein, Bayoumi and Darius (2011) include an aggregate measure of lending standards from the SLOOS in a standard VAR model with a range of other macroeconomic variables, interest rates and financial asset prices. Structural shocks are identified by a Choleski decomposition. They estimate that a one standard deviation shock to lending standards results in a 0.3 per cent decline in GDP after one year, and 0.4 per cent after two years. The authors also consider an alternative measure of credit conditions from the NFIB Small Business Survey (SBS). One of the interesting differences between the SLOOS and SBS is that the SLOOS examines credit conditions from the lenders' perspective, whereas the SBS examines them from the borrowers' perspective (as with the ACCI-Westpac survey). The results from the model using SBS data are broadly consistent with the SLOOS model, although the magnitudes of the effects of credit shocks are smaller.

Rather than using an aggregate measure of credit conditions, Lown and Morgan (2006) focus only on changes in commercial and industrial lending standards from the SLOOS. The authors also use a VAR model with a Choleski decomposition to identify structural credit shocks, and they find that an 8 percentage point increase in the net balance of banks reporting a tightening in standards leads to a 3 per cent decline in the quantity of lending and a 0.5 per cent decline in GDP.

Finally, Ciccarelli, Maddaloni and Peydró (2010) use responses from the ECB's and Federal Reserve's bank lending surveys to construct measures of both credit supply and credit demand. These credit variables are included in VAR models (a panel VAR in the case of the euro area) along with real GDP, the GDP deflator and policy interest rates. The credit channel of monetary policy is found to be operational in both the euro area and the United States, with a monetary policy shock affecting credit availability. In terms of decomposing GDP growth during the financial crisis, the authors find that restrictions to the supply of credit to firms played an important role in reducing output growth in the euro area and restrictions to credit availability for mortgages played an important role in explaining changes in GDP growth in the United States.

Footnotes

See Debelle (2011). [8]

As discussed in footnote 3, Suzuki (2004) uses the ACCI-Westpac variable in a structural VAR framework. However, the aim of that work is somewhat different, looking to test the role of a bank lending channel of monetary policy transmission in Australia, as opposed to identifying shifts in credit supply. [9]

The ACCI-Westpac survey period has changed a few times throughout its history, but in recent years it has been conducted over four weeks around the middle month of each quarter. Prior to 1993, respondents were able to choose from the responses: harder, easier, no change or not applicable (NA). Post-1993, the NA response was removed from the survey. The option of the NA response prior to 1993 may have increased the volatility of the series in the earlier part of the sample. [10]

Other Australian surveys that include questions on credit conditions include the National Australia Bank's Monthly Business Survey (its question on difficulty obtaining finance started in 2008) and the UBS Loan Officers' Survey, which started in 2009. International surveys with a long history on credit conditions include the Federal Reserve's Senior Loan Officer Opinion Survey (which started in 1967, but there are periods in its history where survey data were not collected) and the US National Federation of Independent Business' Small Business Survey, which started in 1986. [11]

One exception to this is the noticeable decline in the proportion of respondents from textiles, fabrics, floor coverings, felt, canvas & rope and clothing & footwear sub-industries between the 1960s and 1980s. [12]

For the 1990s period see Kent and D'Arcy (2000) and Macfarlane (2006), and for the global financial crisis period see Debelle (2009), Battellino (2010, 2011) and Davis (2011). [13]

Also see Battellino (2007) and references contained therein. [14]

This analysis does not address the possibility of endogeneity between credit supply conditions and economic variables (hence the need for the structural VAR analysis presented later in this paper). Nonetheless, for the purposes of Table 1, a simple means of addressing this endogeneity issue is to use lagged observations of economic conditions, which yields qualitatively similar results. [15]

This correlation is likely to be boosted by the fact that financial institutions are themselves subject to credit frictions. A deterioration in one sector of the economy can result in a deterioration in the balance sheets of intermediaries, resulting in a tightening of credit conditions more broadly. [16]

Note that some of the international surveys are based on information from lenders, whereas the ACCI-Westpac survey is from the perspective of borrowers. [17]