flow-based borrowing constraints and macroeconomic fluctuations

Transcription

1 flow-based borrowing constraints and macroeconomic fluctuations Thomas Drechsel (LSE) Annual Congress of the EEA University of Cologne 27 August 2018

2 in a nutshell I What do the dynamics of firm borrowing reveal about macroeconomic fluctuations? I Answering this question depends on correctly capturing the constraints to borrowing I Flow-based borrowing constraints generate empirically plausible business cycle dynamics, other constraints do not 1/22

3 preview of strategy and findings I Present micro evidence: Key features of US corporate debt 1. Lenders ask for collateral 2. Lenders scrutinize flows I Introduce 1. and 2. in a prototype macro model I Responses to various shocks are drastically di erent I Test predictions for investment shock empirically I I Aggregate data (SVAR) suggests flow constraint dominant Firm-level data directly in line with model mechanism 2/22

4 simplified formal intuition for mechanism Resource constraint and capital accumulation: c + i = y k 0 = (1 )k + vi Combine these equations ) 1/v is the relative price: c + k 0 /v = y +(1 )k/v 3/22

5 simplified formal intuition for mechanism Resource constraint and capital accumulation: c + i = y k 0 = (1 )k + vi Combine these equations ) 1/v is the relative price: c + k 0 /v = y +(1 )k/v Borrowing in consumption units (collateral vs. flow): b 0 apple k k 0 /v vs. b 0 apple Di erence: Boom with deleveraging vs. boom with debt buildup in response to v " when constraint binds. 3/22

6 outline for today 1. Micro evidence I Two key features of corporate debt 2. Prototype macro model: I Propose borrowing constraint formulation I Study debt responses to di erent shocks 3. Empirical tests for investment shock: I SVAR I Cross-sectional 4. Conclusion + Remarks on what I am currently working on 4/22

7 1. micro evidence

8 micro evidence overview I What do loan contracts for firms look like? I Use LPC Dealscan data base: Detailed loan-level information covering 75% of US commercial loan market Coverage I Highlight two key features: 1. Collateral 2. Loan covenants 5/22

9 micro evidence what are covenants? I Conditions specified in a loan contract that the borrower is obliged to fulfill while the loan is active I Typically explicit restrictions on financial indicators I Breaches of covenants are frequent and have large economic e ects (see e.g. Chava and Roberts, 2008, Sufi, 2009, Chodorow-Reich and Falato, 2017) 6/22

10 micro evidence loan covenants types Covenant type p25 Median p75 Mean Freq. 1 Max Debt to EBITDA* % 2 Min EBITDA to Interest* % 3 Min EBITDA to Fixed Charge* % 4 Max Leverage Ratio % 5 Max Capex* 6M 19.5M 50M 189M 16.1% 6 Net Worth 44M 123M 325M 3.25B 12.1% 7 Max Snr Debt to EBITDA* % 8 Tangible Net Worth 13M 45M 137M 1.4B 6.4% 9 Min Current Ratio % 10 Min Debt Service Coverage* % EBITDA = earnings before interest, taxes, depreciation, amortization * indicates any flow-based financial indicator 7/22

11 micro evidence covenants and collateral %-share within all loans deals Flow covenants Other covenants Collateral No collateral No info Covenants Collateral %-share within secured/unsecured Covenants if collateral Covenants if no collateral Equal weighted 8/22

12 2.1. propose borrowing limits

13 borrowing constraint Proposed formulation I Based on evidence, propose borrowing constraint in which debt access of firm is restricted by multiple of earnings: b t+1 apple t (also allow lagged and future earnings to enter) I Earnings defined as sales minus labor costs: t = y t w t n t I I compare this to a traditional collateral constraint, where the firm s capital stock serves as collateral: b t+1 apple k E t p k,t+1 k t+1 9/22

14 borrowing constraint important remarks I Study calibration in which constraints are separate I I Helps to derive transparent qualitative predictions To be relaxed in quantitative extension I Constraints are exogenously imposed I In the paper, I discuss microfoundations and provide a formal rationalization based on limited enforcement. I Asset values and flows are directly linked I I Detailed formal discussion in the paper Both timing and definition of flow make a di erence 10 / 22

15 2.2. prototype macro model

16 why set up a model? I Want to know which type of constraint matters most for macroeconomic fluctuations I Empirical challenge: Strong contemporaneous correlation in the data between credit, asset values and flows, so constraints impossible to distinguish from raw data I Proposed solution: 1. Study shocks that imply di erent conditional dynamics 2. Identify these shocks empirically (SVAR) 3. Compare empirical dynamics with model dynamics 11 / 22

17 model environment simplified overview I Stylized model with a relatively realistic characterization of the firm sector I Approach: Standard neoclassical model... + Tax advantage for firm debt (see e.g. Hennessy and Whited, 2005) + Investment adjustment cost (p k is a ected by 1/v and by variation in Tobin s q) + Borrowing constraint (collateral or flow-based) 12 / 22

18 model results irf of debt to different shocks 5 4 TFP shock Model with collateral constraint Model with flow-based constraint Financial shock 3-2 % 2 % % Investment shock (permanent) % Aggregate demand shock / 22

19 3.1. empirical test: svar

20 empirical test: svar I For the empirical tests, I focus on one of the shocks that gives di erent predictions: The investment shock I This shock has a direct empirical counterpart: The inverse relative price of investment goods I I use two alternative identification schemes: 1. Long-run restrictions (Fisher, 2006) 2. Medium-horizon restrictions (Barsky and Sims, 2012) I Below I present the results for / 22

21 svar: specification and lr restrictions Consider the MA(1) representation of an SVAR: Y t = B(L) 1 u t, with Y t =[dlog(p kt ) dlog(y t /n t ) log(n t )] 0. Long-run restrictions on B(1) 1 =[B 0 B 1... B p ] 1 : 1. p k t only a ected by first shock 2. y t /n t only a ected by first and second shock Then: Add other variables of interest: earnings, capital, debt. Leave remaining shocks unidentified, so that ordering irrelevant. Data and sample 15 / 22

22 svar results ist shock 0 Price of capital 2 Labor productivity 1 Hours % % % Earnings 0 Capital 4 Debt % 5 % -2 % / 22

23 3.2. empirical test: firm-level

24 specification of local projection Estimate the horizon h IRF of total debt of firm i by running the regression log(b i,t+h )= h + h û IST,t + controls i,t + i,t+h and obtaining estimates of h, h =0, 1, 2, / 22

25 specification of local projection Estimate the horizon h IRF of total debt of firm i by running the regression log(b i,t+h )= h + h û IST,t + controls i,t + i,t+h and obtaining estimates of h, h =0, 1, 2,... Furthermore, can interact shock with dummies that capture whether firm is flow borrower of collateral borrower log(b i,t+h ) = h + h û IST,t + controls i,t + flow h + coll h 1 i,t,flow û IST,t + flow h 1 j,t,flow 1 i,t,coll û IST,t + coll h 1 i,,t,coll + i,t+h 17 / 22

26 cross-sectional projections more details I The Compustat-Dealscan merged data set: 200, 000 firm-quarter obs for 4, 500 distinct firms, I Since the loan issuance information is sparse sample is reduced when introducing 1 i,t,flow and 1 i,t,coll I As controls I use firm size and two-digit industry. I I include two lags of u IST to mop up serial correlation I When expanding h, I keep the firm composition constant I In what follows I show 90% bands 18 / 22

27 projection results (1/2) average response of debt to ist panel data: average debt irf to ist % / 22

28 projection results (2/2) interacted responses of debt to ist 15 With collateral 15 With earnings covenants % % With both 15 With neither % % / 22

29 take-aways from empirics I Proposed mechanism: Expansionary investment shock rises debt levels if borrowing constraint is relaxed by the shock I I False with collateral constraint True with flow constraint I Aggregate dynamics suggest that the flow-based constraint more relevant for the economy as a whole I Firm-level responses to the shocks are directly supportive of the suggested theoretical mechanism 21 / 22

30 4. conclusion

31 conclusion I Can firm credit tell us something about the macroeconomy? I Micro evidence suggests feedback between firms earnings flows and their access to debt I I formulate a theory and verify its predictions in aggregate and firm-level US data I Flow-based borrowing constraints capture dynamics correctly 22 / 22

32 quantitative model what i am currently work on I So far qualitative predictions and tests, with main focus on the sign of debt dynamics I Now: Put both the collateral and the flow-based constraint on equal footing, and embed them in a New Keynesian DSGE which is estimated on US data I Allows the data to speak about the relevance of the constraints and the way they a ect the aggregate dynamics. I In essence, I would like to find out how the answer to the question What drives US business cycles? changes when capturing the two constraints, jointly and individually I Marginal likelihood, variance decompositions, IRFs,...

33 appendix slides

34 relation to the literature overview I Financial frictions in business cycles: Kiyotaki and Moore (1997, 2012), Bernanke, Gertler, and Gilchrist (1999), Azariadis, Kaas, and Wen (2016), Greenwald (2016), many more. I Empirical corporate finance literature on covenants: Chava and Roberts (2008), Sufi (2009), Chodorow-Reich and Falato (2017), many more. I Existing work with flow-constraints (on di erent questions): Kiyotaki (1998), Jappelli and Pagano (1989), Mendoza (2006), Bianchi (2011), Korinek (2011), Schmitt-Grohe and Uribe (2016a, 2016b). I IST shocks: Greenwood, Hercowitz, and Krusell (2000), Fisher (2006), Justiniano, Primiceri, and Tambalotti (2010, 2011).

35 relation to the literature greenwald (2016) I It is interesting that in the case of households, flow-based constraints are also present I In fact, there are regulatory constraints on payment-to-income (PTI) ratios for mortgages in many countries I Greenwald (2016) explores the role of these constraints for the transmission of macro shocks through the mortgage market I My work is similar in nature, but I want to bring the corporate sector to the center stage

36 relation to the literature lian and ma (2017) I Another closely related paper is Lian and Ma (2017) I These authors also investigate the di erence between flow-based and secured firm borrowing and their evidence is well in line with what I present I They focus on empirically testing the influence of earnings on borrowing in firm panel data I I embed these insights into a macroeconomic model and study their consequences for business cycles

37 loan-level data sample coverage Borrowing corporations Loan deals Loan facilities Back

38 more on empirical evidence ii equal-weighted shares %-share within all loans deals Flow covenants Other covenants Collateral No collateral No info Covenants Collateral %-share within secured/unsecured Covenants if collateral Covenants if no collateral Back

39 svar details data and sample I I use data for the nonfinancial business sector I Retrieve nominal data and deflate with the consumption deflator for nondurable goods and services I For the capital price, use the deflator for equipment investment, as this is the one which admits the long-run trend o which my IST shock is identified I I show results for the sample 1952:Q1-2016:Q4, with 4 lags Back

40 Bibliography I Azariadis, C., L. Kaas, and Y. Wen (2016): Self-Fulfilling Credit Cycles, The Review of Economic Studies, 83, Barsky, R. B. and E. R. Sims (2012): Information, Animal Spirits, and the Meaning of Innovations in Consumer Confidence, American Economic Review, 102, Bernanke, B. S., M. Gertler, and S. Gilchrist (1999): The financial accelerator in a quantitative business cycle framework, Handbook of Macroeconomics, 1, Bianchi, J. (2011): Overborrowing and Systemic Externalities in the Business Cycle, American Economic Review, 101, Chava, S. and M. R. Roberts (2008): How Does Financing Impact Investment? The Role of Debt Covenants, The Journal of Finance, 63, Chodorow-Reich, G. and A. Falato (2017): The Loan Covenant Channel: How Bank Health Transmits to the Real Economy, Working Paper 23879, NBER. Fisher, J. D. (2006): The Dynamic E ects of Neutral and InvestmentSpecific Technology Shocks, Journal of Political Economy, 114, Greenwald, D. L. (2016): The mortgage credit channel of macroeconomic transmission, Working Paper. Greenwood, J., Z. Hercowitz, and P. Krusell (2000): The role of investment-specific technological change in the business cycle, European Economic Review, 44, Hennessy, C. A. and T. M. Whited (2005): Debt dynamics, The Journal of Finance, 60, Jappelli, T. and M. Pagano (1989): Consumption and capital market imperfections: An international comparison, The American Economic Review, Justiniano, A., G. E. Primiceri, and A. Tambalotti (2010): Investment shocks and business cycles, Journal of Monetary Economics, 57, (2011): Investment shocks and the relative price of investment, Review of Economic Dynamics, 14, ,specialissue:SourcesofBusinessCycles. Kiyotaki, N. (1998): Credit and business cycles, The Japanese Economic Review, 49, Kiyotaki, N. and J. Moore (1997): Credit Cycles, Journal of Political Economy, 105,

41 Bibliography II (2012): Liquidity, Business Cycles, and Monetary Policy, Working Paper 17934, National Bureau of Economic Research. Korinek, A. (2011): Excessive dollar borrowing in emerging markets: Balance sheet e ects and macroeconomic externalities, Working paper. Lian, C. and Y. Ma (2017): Anatomy of Corporate Borrowing Constraints,. Mendoza, E. G. (2006): Lessons from the Debt-Deflation Theory of Sudden Stops, The American Economic Review, 96, Schmitt-Grohe, S. and M. Uribe (2016a): Adjustment to Small, Large, and Sunspot Shocks in Open Economies With Stock Collateral Constraints, Working Paper 22971, National Bureau of Economic Research. (2016b): Multiple Equilibria in Open Economy Models with Collateral Constraints: Overborrowing Revisited, Working Paper 22264, National Bureau of Economic Research. Sufi, A. (2009): Bank Lines of Credit in Corporate Finance: An Empirical Analysis, Review of Financial Studies, 22,