Credit Spreads and Business Cycle Fluctuations

This 2012 American Economic Review paper by Simon Gilchrist and Egon Zakrajšek constructs a new corporate-bond credit spread index – the “GZ spread” – from a large panel of secondary-market bond prices (346,126 bond-month observations, 5,982 securities, 1,112 US nonfinancial firms, January 1973 to September 2010), matching each bond’s own cash flows to a hypothetical Treasury security with identical cash flows to compute the spread, which avoids the “duration mismatch” that afflicts conventional maturity-bucketed indexes like the Baa-Aaa spread or the paper-bill spread. The GZ spread substantially outperforms these standard indicators as a predictor of future economic activity: in monthly forecasting regressions it is “statistically a highly significant predictor” of payroll employment, unemployment, and industrial production at both short and long horizons, raising adjusted R-squared by 12-15 percentage points at the 12-month horizon, with a 100 basis point increase associated with about a 3 percentage point (annualized) decline in industrial production growth over the next three months; in quarterly regressions a 100 basis point increase in the GZ spread predicts real GDP growth more than 1.25 percentage points lower over the following four quarters. The authors then decompose the spread using a credit-spread pricing model that regresses individual bond spreads on a firm’s Merton/Bharath-Shumway distance-to-default and on bond characteristics (with additional term-structure and volatility controls for callable bonds), defining the “excess bond premium” (EBP) as the gap between the actual average spread and the value predicted by this regression – i.e., the component of pricing unexplained by measured default risk. The central finding is that the EBP, not the predicted (default-risk) component, carries essentially all of the GZ spread’s forecasting power for real activity: in the post-1985 subsample the predicted GZ spread has no forecasting power for real GDP (coefficient -0.023, t = 0.20) while the EBP remains highly significant (t = 6.80), with a 100 basis point EBP increase predicting roughly a 2 percentage point GDP decline over four quarters. In a recursively identified eight-variable quarterly VAR (consumption, business fixed investment, real GDP, GDP deflator inflation, EBP, equity excess return, 10-year Treasury yield, federal funds rate, ordered so EBP shocks affect the real economy only with a lag), a one-standard-deviation EBP shock (about 20 basis points) produces output that bottoms roughly 0.5 percentage points below trend five quarters out, a sharper and more persistent investment decline, a cumulative stock market decline of about 7 percentage points, appreciable disinflation, and monetary policy easing beginning about one quarter after the shock; EBP shocks account for more than 10 percent of output variance and more than 25 percent of business-fixed-investment variance at business-cycle frequencies, proportions the authors note exceed those typically attributed to monetary policy shocks. Additional evidence – a close correlation between the EBP and the Senior Loan Officer Opinion Survey measure of bank credit-standard tightening, an inverse relationship with financial-sector return on assets, and a broker-dealer VAR in which adverse shocks to broker-dealer profitability raise broker-dealer CDS spreads together with the EBP – supports interpreting the EBP as a measure of the financial sector’s risk-bearing capacity and credit-supply conditions, consistent with financial-accelerator and intermediary-asset-pricing theories, though the paper is careful to describe its forecasting results as evidence of predictive content rather than as establishing causality from credit spreads to economic activity, and its recursive VAR ordering is an identifying assumption checked only by limited robustness exercises (alternative orderings, monthly frequency) rather than by formal tests.

Egon Zakrajšek | Macro Paper Warehouse

Credit Spreads and Business Cycle Fluctuations