Does credit matter for the business cycle? A vast literature in macroeconomics suggests that it does. The well-known financial accelerator mechanism describes how adverse shocks can be propagated through falling prices of collateralised assets which then deteriorate the balance sheets of firms, inhibiting their borrowing and investment (Bernanke and Gertler 1989, Kiyotaki and Moore 1997). Another more recent literature promotes the idea that disruptions in the financial sector (‘financial shocks’) can induce a tightening of collateral constraints that spills over to the real sector (e.g. Jermann and Quadrini 2012).
In our recent paper we call into question the importance of the standard collateral channel for business cycle dynamics (Azariadis et al. 2015). When firm debt is decomposed into secured and unsecured components, we find that it is the unsecured component of firm debt which correlates strongly and positively with US GDP, while the secured component of firm debt is acyclical. Figure 1 illustrates this finding – the contemporaneous correlation coefficient between unsecured debt and GDP over the period 1981-2012 is 0.7, whilst the between secured debt and GDP it is -0.05.
Figure 1. Secured and unsecured firm credit over the business cycle (US data, 1981-2012)
Figure 1 is based on firm-level data from Compustat. A natural question is how sensitive the result is to the particular firm sample. Indeed, we show that unsecured borrowing is much more prominent for larger firms – when we remove the largest 5% of firms, the share of unsecured debt in total debt drops from 83% to 67%. Also the growth rate of unsecured debt is much higher in larger firms. However, the cyclical properties remain similar – unsecured debt is strongly procyclical while secured debt is acyclical, regardless of whether the largest firms are included in the sample or not. We also confirm this finding for aggregate data from the flow of funds accounts – mortgages of non-financial firms (as a proxy for secured borrowing) are acyclical, whereas corporate bonds (as a proxy for unsecured borrowing) are procyclical.
Another important observation is that unsecured firm debt leads GDP by about a year. In contrast, secured debt tends to lag GDP whenever the contemporaneous correlation is weakly positive (see Figure 2). To obtain some indication of causality, we conduct a Granger causality test to explore if secured or unsecured debt contains information to predict output. We find that unsecured debt helps predict future GDP movements, while this is not the case for secured debt. This result suggests that in the period 1981-2012 the so-called ‘credit cycle’ and its intimate relation to the business cycle is not driven by movements in secured debt or the value of collateral.
Figure 2. Lead-lag correlations between debt at year t+j with GDP at year t for j=-4,…,0,…,4
To explain the role of unsecured credit for the business cycle, we develop a canonical macroeconomic model in which firms differ in productivity and borrow secured and unsecured debt. Self-enforcing limits on unsecured debt depend on borrowers’ and lenders’ expectations about future credit conditions. We show that those expectations may be subject to sunspot shocks – self-fulfilling expectations about future credit conditions induce fluctuations of credit, output and aggregate productivity. Based on a model calibration, we conduct a ‘horse race’ between those sunspot shocks, shocks to collateral, and shocks to aggregate technology. All three shocks together do a good job of describing the fluctuations of major macroeconomic variables. In a variance decomposition, we find that sunspot shocks account for about half of all output volatility, whereas aggregate technology shocks are unimportant.
Sunspot shocks also generate a stronger and more persistent response than the other two shocks (see Figure 3). On the other hand, if the endogenous influence of sunspots on credit conditions is excluded a priori, our results show that too much output volatility would be incorrectly attributed to exogenous movements in aggregate technology – a standard result in the literature. We conclude that self-fulfilling and endogenously propagated credit shocks are important in US business cycles.
Figure 3. Impulse responses to three shocks in the business-cycle model
Azariadis, C, L Kaas, and Y Wen (2015), “Self-Fulfilling Credit Cycles”, Working Paper No. 2015-005A, Federal Reserve Bank of St. Louis
Bernanke, B and M Gertler (1989), “Agency costs, net worth, and business fluctuations”, The American Economic Review, 79, 14–31.
Jermann, U and V Quadrini (2012), “Macroeconomic Effects of Financial Shocks”, The American Economic Review, 102, 238–271.
Kiyotaki, N and J Moore (1997), “Credit cycles”, Journal of Political Economy, 105, 211–248.