Banking Industry Risk and Macroeconomic Implications April 2014 Francisco Covas a Emre Yoldas b Egon Zakrajsek c Extended Abstract There is a large body of literature that focuses on the financial system either as a propagation mechanism for macroeconomic shocks, or as an independent source of fluctuations in real economic activity, see Brunnermeier and Sannikov (2013), Gilchrist and Zakrajsek (2012), and He and Krishnamurty (2012) for recent examples. In addition, there is a growing recent literature that explores implications of uncertainty shocks for economic activity. Arellano et al. (2012), Bloom et al. (2012), and Christiano et al. (2013) develop dynamic stochastic general equilibrium models featuring uncertainty shocks, and Chauvet et al. (2013), Allen et al. (2012), and Bakshi et al. (2011) investigate the predictive content of financial volatility measures for key macroeconomic aggregates. In this paper, we build on these two strands of the literature and analyze time-varying risk in the banking industry using stock return data and explore its connections with macroeconomic activity. In particular, we focus on the residuals obtained from rolling-window regressions of bank stock returns on standard risk factors. We then analyze those residuals and test for the presence of latent risk factors that are intended to capture banking industry specific risks. We use daily stock price data from CRSP over July 1973-December 2013 period and use the top 50 banks with respect to market capitalization to form balanced panels over two-year rolling windows. Then, we estimate the workhorse three-factor Fama-French asset-pricing model (Fama and French, 1992) for each sample. 1 Using the statistical procedure proposed by Bai and Ng (2002) we find that there is at least one common factor in banks return residuals beginning The views expressed are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors or the Federal Reserve System. a Federal Reserve Board, Division of Monetary Affairs, Phone: (202) 452-3497, E-mail: francisco.b.covas@frb.gov. b Federal Reserve Board, Division of Monetary Affairs, Phone: (202) 973-7302, E-mail: emre.yoldas@frb.gov. c Federal Reserve Board, Division of Monetary Affairs, Phone: (202) 728-5864, E-mail: egon.zakrajsek@frb.gov. 1 Results are robust to alternative factor model specifications or estimation windows. 1
in 1990 (see Figure 1-Panel a). For parsimony, we focus on the first principal component in the remainder of the analysis. The explanatory power of the latent banking risk factor trended up since early 1990s and increased dramatically after the passage of the Gramm-Leach-Bliley Act in 1999 (see Figure 1-Panel b). It also increased notably during the 2007-2008 financial crisis. We find that the estimated latent banking risk factor exhibits substantial volatility clustering (see Figure 2-Panel a). To arrive at a broad measure of risk we construct an index that reflects three key components of risk: (i) time variation in the second moment of the latent risk factor, (ii) the changing explanatory power of the latent risk factor in the cross section of bank return residuals, (iii) changing exposure of the average bank to the latent risk factor. Specifically, we simply use the product of the estimated volatility process of the latent risk factor, the fraction of total residual variance explained by the latent factor, and the median bank s exposure to it (Figure 2-Panel b). The proposed measure of uncertainty in the banking sector leads broad measures of uncertainty, such as the VIX, during the 2007-2008 financial crisis (see Figure 3). To explore macroeconomic implications of the banking industry uncertainty index, we estimate a vector autoregression (VAR) using monthly data from January 1990-December 2013. The ordering of the variables in the VAR assumes that shocks to the banking industry uncertainty index affects other financial variables, such as stock prices, contemporaneously while employment and output respond with a lag. We find that in response to one-standard deviation shock to the banking industry uncertainty index, industrial production growth slows by about 1.3 percent and employment grows 0.6 percent slower over a 3-year period (see Figure 4). Moreover, both the investment-grade corporate bond spread and the VIX respond significantly to shocks to the banking industry uncertainty index, which underscores the role played by banks in the propagation of shocks from the banking sector to the broad financial system. 2
References Allen, L., T. Bali, and Y. Tang, 2012, Does Systemic Risk in the Financial Sector Predict Future Economic Downturns?, Review of Financial Studies, 25, 3000-3036. Arellano, C., Y. Bai, and P.J. Kehoe, 2012, Financial Frictions and Fluctuations in Volatility, Federal Reserve Bank of Minneapolis Research Department Staff Report 466. Bai, J. and S. Ng, 2002, Determining the Number of Factors in Approximate Factor Models, Econometrica, 70, 191-221. Bakshi, G., G. Panayotov, and G. Skoulakis, 2011, Improving the Predictability of Real Economic Activity and Asset Returns with Forward Variances Inferred from Option Portfolios, Journal of Financial Economics, 100, 475-495. Bloom, N. M. Floetotto, N. Jaimovich, I. Saporta-Eksten, and S.J. Terry, 2012, Really Uncertain Business Cycles, NBER Working Paper 18245. Brunnermeier, M.K. and Y. Sannikov, 2013, A Macroeconomic Model with a Financial Sector, American Economic Review, forthcoming. Chauvet, M., Z. Senyuz, and E. Yoldas, 2013, What Does Financial Volatility Tell Us About Macroeconomic Fluctuations?, FEDS Working Paper 2013-61. Engle, R.F., 2002, Dynamic Conditional Correlation: A Simple Class of Multivariate Generalized Autoregressive Conditional Heteroskedasticity Models, Journal of Business and Economic Statistics, 20, 339-350. Fama, E.F. and K.R. French, 1992, The Cross-section of Expected Stock Returns, the Journal of Finance, 47, 427-465. Christiano, L.J., R. Motto, and M. Rostagno, 2013, Risk Shocks, American Economic Review, Forthcoming. Gilchrist, S. and E. Zakrajsek, 2012, Credit Spreads and Business Cycle Fluctuations, American Economic Review, 102, 1692-1720. He, Z. and A. Krishnamurthy, 2012, A Macroeconomic Framework for Quantifying Systemic Risk, Working Paper, Kellogg School of Management, Northwestern University. 3
Figure 1: Latent Factors in Bank Return Residuals Panel a: Estimated Number of Latent Risk Factors Note: IC1-IC3 correspond to the number of factors selected by the three asymptotically consistent information criteria proposed in Bai and Ng (2002, p. 201). For each information criterion, the mode of daily data in a given month is presented. Panel b: Fraction of Variance Explained by the First Principal Component Note: Monthly values are arithmetic averages of daily data. 4
Figure 2: Volatility Clustering in the Latent Risk Factor Panel a: Volatility of the Latent Risk Factor Note: Volatility is obtained from an integrated GARCH specification with parameter equal to 0.94 as in the RiskMetrics model for daily data and the aggregated to monthly frequency by taking the maximum value in each month. Panel b: Banking Industry Risk Index Note: Latent risk factor volatility (see Panel a) is scaled by the fraction of variance explained by the first principal component (Figure 1-Panel b) and the median exposure to the latent risk factor. June 1975 is normalized to 1. 5
Figure 3: Latent Risk Factor Volatility and VIX Note: Aggregation to monthly frequency is implemented by taking maximum value in a given month for both variables. January 1990 is normalized to 1. Figure 4: Response of Macroeconomic Aggregates to Shocks to the Latent Risk Factor based Uncertainty Measure 0.5 Industrial Production Growth 0.2 Employment Growth 0.0 0.0-0.5-0.2-1.0-0.4-1.5-0.6-2.0-0.8-2.5-1.0 PCE Inflation Core Loan Growth.1 0.5.0 0.0 -.1-0.5-1.0 -.2-1.5 -.3 Note: Monthly VAR with three lags is estimated using data from January 1990 to December 2013 on the following variables: Industrial production growth, growth in payroll employment, core loan growth, PCE inflation, latent risk factor based uncertainty measure, VIX, BAA spread, and the 2-year T-Note yield. Identification of orthogonal shocks is obtained via Cholesky decomposition. 6-2.0