Investment, Financial Frictions and the Dynamic Effects of Monetary Policy James Cloyne Clodo Ferreira Maren Froemel Paolo Surico UC, Davis Bank of Spain London Business School & BoE ESCB Research Cluster 2 Paris, November 7 th 2018 The views expressed are those of the authors and do not necessarily reflect the views of the Bank of Spain, the Euro-system, Bank of England, MPC, FPC or PRA.
Monetary transmission and financial frictions Which type of firms are more sensitive to interest rate changes? How much do these firms contribute to the aggregate response? How can financial frictions be identified from balance sheet data? Do financial frictions dampen or amplify monetary policy shocks?
Monetary transmission and financial frictions Which type of firms are more sensitive to interest rate changes? How much do these firms contribute to the aggregate response? How can financial frictions be identified from balance sheet data? Do financial frictions dampen or amplify monetary policy shocks?
Empirical challenges and our approach 1 Assess heterogeneity across firms characteristics. Look at firm-level capital expenditure in U.K. and U.S. Explore variation by age, size, growth, leverage and Tobin s Q. 2 Evaluate balance sheet position across groups of firms. Exploit info on dividends and bond issuance decisions, credit scores, and equity prices. 3 Identify a series of monetary policy shocks. U.K.: Gerko and Rey (2017); U.S.: Gertler and Karadi (2015).
Empirical challenges and our approach 1 Assess heterogeneity across firms characteristics. Look at firm-level capital expenditure in U.K. and U.S. Explore variation by age, size, growth, leverage and Tobin s Q. 2 Evaluate balance sheet position across groups of firms. Exploit info on dividends and bond issuance decisions, credit scores, and equity prices. 3 Identify a series of monetary policy shocks. U.K.: Gerko and Rey (2017); U.S.: Gertler and Karadi (2015).
Empirical challenges and our approach 1 Assess heterogeneity across firms characteristics. Look at firm-level capital expenditure in U.K. and U.S. Explore variation by age, size, growth, leverage and Tobin s Q. 2 Evaluate balance sheet position across groups of firms. Exploit info on dividends and bond issuance decisions, credit scores, and equity prices. 3 Identify a series of monetary policy shocks. U.K.: Gerko and Rey (2017); U.S.: Gertler and Karadi (2015).
Empirical challenges and our approach 1 Assess heterogeneity across firms characteristics. Look at firm-level capital expenditure in U.K. and U.S. Explore variation by age, size, growth, leverage and Tobin s Q. 2 Evaluate balance sheet position across groups of firms. Exploit info on dividends and bond issuance decisions, credit scores, and equity prices. 3 Identify a series of monetary policy shocks. U.K.: Gerko and Rey (2017); U.S.: Gertler and Karadi (2015).
Main empirical finding I: heterogeneity Younger firms exhibit significantly larger adjustments in investment after an interest rate change and drive the aggregate response. Within these, the strongest adjustment is recorded among younger firms paying no dividends. Peak effect occurs between two and three years after the shock. Results are robust to controlling for other, more traditional, firms characteristics.
Main empirical finding II: mechanism Younger firms borrowing is more asset-based (than earning-based)......and their investment relies more on external funds (debt) After a contractionary monetary policy shock: interest payments and net worth respond homogeneously for all age groups borrowing,though, drops by a larger and significant amount for younger firms, especially those paying no dividends; sales (demand) responses are less pronounced and more homogenous across age/dividends groups. Consistent with financial frictions playing a quantitatively important role to amplify business cycle fluctuations through collateral values.
Main empirical finding II: mechanism Younger firms borrowing is more asset-based (than earning-based)......and their investment relies more on external funds (debt) After a contractionary monetary policy shock: interest payments and net worth respond homogeneously for all age groups borrowing,though, drops by a larger and significant amount for younger firms, especially those paying no dividends; sales (demand) responses are less pronounced and more homogenous across age/dividends groups. Consistent with financial frictions playing a quantitatively important role to amplify business cycle fluctuations through collateral values.
Main empirical finding II: mechanism Younger firms borrowing is more asset-based (than earning-based)......and their investment relies more on external funds (debt) After a contractionary monetary policy shock: interest payments and net worth respond homogeneously for all age groups borrowing,though, drops by a larger and significant amount for younger firms, especially those paying no dividends; sales (demand) responses are less pronounced and more homogenous across age/dividends groups. Consistent with financial frictions playing a quantitatively important role to amplify business cycle fluctuations through collateral values.
Outline 1 Data & approach 2 Heterogeneity 3 Financial frictions 4 Other transmission mechanisms 5 Concluding remarks
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 13 / 44 Firm-level data: Worldscope (U.K.), Compustat (U.S.) Panel of annual (UK) / quarterly (US) data. Sample: 1987-2015. Real variables: capital expenditure, age (years since incorporation or IPO), size (by asset value), growth (by assets), net sales. Financial variables: leverage (debt over assets); Tobin s Q; equity; cash flows; dividends paid; share prices; interest payments, bond issuance. U.K.: 2,435 unique listed firms and around 27,000 (firms x years) obs. U.S.: 11,577 unique listed firms and 623,000 (firms x quarters) obs.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 14 / 44 Investment: National Statistics vs Micro data United Kingdom Levels Growth rates log investment 8.5 9 9.5 10 10.5 Year-on-year growth rates -.1 0.1.2 -.2.3 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 Year Correlation.58 (pvalue = 0) 1985q1 1990q1 1995q1 2000q1 2005q1 2010q1 2015q1 Year investment (ONS) investment (WS) investment (ONS) investment (WS) United States Real Investment (log) 6 6.5 7 7.5 8 % -20-10 0 10-30 -20-10 0 10 20 % 1985 1990 1995 2000 2005 2010 2015 Date 1985 1990 1995 2000 2005 2010 2015 Date National Statistics Aggregated Micro Data National Statistics Aggregated Micro Data
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 15 / 44 Monetary policy shock series High frequency surprises on short rate futures in a 30 minutes window around policy announcements, available since 2001 for the U.K. (Gerko-Rey) and since 1991 for the U.S. (Gertler-Karadi). Monthly macro proxy-svar over 1987-2015 using the high frequency surprises as proxies to extract a SHOCK SERIES for the full sample (see Mertens and Ravn, 2014; Ramey 2016). Firms are matched with monthly interest rate surprises based on their respective filing dates.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 16 / 44 Empirical specification: panel IV-Local Projections G G X j,t+h X j,t 1 = αj h + αg h Dgj,t h + βg h Dgj,t h R t + ɛ j,t+h g=1 g=1 Baseline X j,t+h : capital expenditure over net PPE at horizon h; Dg: dummy for groups of age, size, leverage, paying dividends in previous year; R t: interest rate in quarter t (slightly more convoluted for the U.K. annual data); Instrument: policy shocks in the accounting period t, extracted from proxy-svar. β h g: impulse response for group g at forecast horizon h. Additional firm-level X j,t+h : borrowing, share prices, sales, interest payments; Additional aggregate X t+h : industrial production, stock price index, credit spread. Standard errors clustered by firms and time.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 17 / 44 The average effect: capital expenditure over net PPE United Kingdom United States -.6 -.4 -.2 0.2.4-1 -.5 0.5 0 1 2 3 4 Year 1 4 8 12 16 20 Monetary Policy shock: 25 basis point increase. Standard errors: clustered by firms and time. Confidence band: 90%. Consistent with the MACRO EVIDENCE using data from national statistics. Same message when reporting at the ANNUAL FREQUENCY
Outline 1 Data & approach 2 Heterogeneity 3 Financial frictions 4 Other transmission mechanisms 5 Concluding remarks
DESCRIPTIVE STATISTICS
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 20 / 44 Firms size, growth and revenues as function of AGE Size Asset growth EBITDA United Kingdom Linear Prediction 3.5 4 4.5 5 Size and age: Assets Book value 5 15 25 35 45 55 Age Linear Prediction 5 10 15 20 Asset Growth and Age 5 15 25 35 45 55 Age % 8 9 10 11 12 5 15 25 35 45 55 Age United States log(assets) 4.5 5 5.5 6 6.5 5 15 25 35 45 55 Age % 2 4 6 8 5 15 25 35 45 55 Age % 0.005.01.015.02.025 5 15 25 35 45 55 Age Based on regressions of the variable of interest on age, squared age, sectorsxtime fixed effects (and size).
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 21 / 44 Firms financial characteristics as function of AGE Credit scores Paying dividends Leverage United Kingdom Linear Prediction 60 65 70 75 80 Credit Score and Age 5 15 25 35 45 55 Age Prob(paid dividends) 0.2.4.6.8 1 Probabilities of Paying Dividends Last Year 5 15 25 35 45 55 Age Linear Prediction 16 17 18 19 20 Leverage and age: Book value 5 15 25 35 45 55 Age United States probability 0.2.4.6.8 1 5 15 25 35 45 55 Age low rating high rating probability 0.2.4.6.8 1 5 15 25 35 45 55 Age % 20 25 30 35 40 5 15 25 35 45 55 Age Based on regressions of the variable of interest on age, squared age, sectorsxtime fixed effects (and size).
Summary: younger firms tend on average to be smaller in size grow faster (in assets) have less internal funds have lower credit scores (and probability of issuing bonds) probability of paying dividends have lower leverage have higher (average) Tobin s Q
IMPULSE RESPONSE ANALYSIS
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 24 / 44 Dynamic effects of monetary policy on investment Younger Middle-aged Older United Kingdom -2-1 0.5 Years -2-1 0.5 Years -2-1 0.5 Years United States -.5-1.5-1 0.5 -.5-1.5-1 0.5 -.5-1.5-1 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 25 / 44 Investment response by AGE & DIVIDENDS: U.K. Younger Older NO dividends -4-2 -1 0.5-4 -2-1 0.5 Years Years Dividends -4-2 -1 0.5-4 -2-1 0.5 Years Years Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 26 / 44 Investment response by AGE & DIVIDENDS: U.S. Younger Older NO dividends -1.5-1 -.5 0.5-1.5-1 -.5 0.5 Dividends -1.5-1 -.5 0.5-1.5-1 -.5 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 27 / 44 Heterogeneity by age and dividends is ROBUST to... 1 Size charts 2 Firm s growth charts 3 Leverage charts 4 Tobin s Q charts
Outline 1 Data & approach 2 Heterogeneity 3 Financial frictions 4 Other transmission mechanisms 5 Concluding remarks
UNCONDITIONAL CORRELATIONS
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks Borrowing: asset-based vs. earning-based b LT i,t = G G β 1,g xdg i,t COLLATERAL i,t 1 + β 2,g xdg i,t EBITDA i,t 1 +X i,tγ+ɛ i,t g=1 g=1 U.K. U.S. Young / nodiv Old / div Young / nodiv Old / div 0.0245*** 0.0117 0.0628*** 0.0376** COLLATERAL t 1 (0.0092) (0.0093) (0.0131) (0.0141) -0.0126 0.0694*** 0.0068 0.0484** EBITDA t 1 (0.0114) (0.0191) (0.0160) (0.0183) R 2 0.4128 0.4128 N 14 921 16 149 Dependent variable: long-term debt Note: the regression includes timexsector and firm-level fixed effects. Standard errors are clustered by time and firm. 30 / 44
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks Borrowing: asset-based vs. earning-based b LT i,t = G G β 1,g xdg i,t COLLATERAL i,t 1 + β 2,g xdg i,t EBITDA i,t 1 +X i,tγ+ɛ i,t g=1 g=1 U.K. U.S. Young / nodiv Old / div Young / nodiv Old / div 0.0245*** 0.0117 0.0628*** 0.0376** COLLATERAL t 1 (0.0092) (0.0093) (0.0131) (0.0141) -0.0126 0.0694*** 0.0068 0.0484** EBITDA t 1 (0.0114) (0.0191) (0.0160) (0.0183) R 2 0.4128 0.4128 N 14 921 16 149 Dependent variable: long-term debt Note: the regression includes timexsector and firm-level fixed effects. Standard errors are clustered by time and firm. 31 / 44
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks Borrowing: asset-based vs. earning-based b LT i,t = G G β 1,g xdg i,t COLLATERAL i,t 1 + β 2,g xdg i,t EBITDA i,t 1 +X i,tγ+ɛ i,t g=1 g=1 U.K. U.S. Young / nodiv Old / div Young / nodiv Old / div 0.0245*** 0.0117 0.0628*** 0.0376** COLLATERAL t 1 (0.0092) (0.0093) (0.0131) (0.0141) -0.0126 0.0694*** 0.0068 0.0484** EBITDA t 1 (0.0114) (0.0191) (0.0160) (0.0183) R 2 0.4128 0.4128 N 14 921 16 149 Dependent variable: long-term debt Note: the regression includes timexsector and firm-level fixed effects. Standard errors are clustered by time and firm. 32 / 44
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 33 / 44 Financing investment: external vs. internal funds Based on a regression of investment on net debt and net equity issuances, cash flows, sectorxtime dummies (and firms controls).
CONDITIONAL CORRELATIONS
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 35 / 44 EQUITY (MKT. VALUE) responds for textbfall firms Younger & NO dividends Older & Paying dividends United Kingdom -8-4 -2 0 2-8 -4-2 0 2 Year end (Q4) Year end (Q4) United States -8-6 -4-2 0 2-8 -6-4 -2 0 2 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 36 / 44 More homogenous INTEREST PAYMENTS responses Younger & NO dividends Older & Paying dividends United Kingdom percent -10-5 -2 0 2 5 year end percent -10-5 -2 0 2 5 year end United States -4-2 0 2 4-4 -2 0 2 4 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 37 / 44 Larger BORROWING response for Younger/No Div. Younger & NO dividends Older & Paying dividends United Kingdom -10-5 -2 0 2 Years -10-5 -2 0 2 Years United States -1.5-1 -.5 0.5 1-1.5-1 -.5 0.5 1 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Outline 1 Data & approach 2 Heterogeneity 3 Financial frictions 4 Other transmission mechanisms 5 Concluding remarks
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 39 / 44 More homogenous SALES responses Younger & NO dividends Older & Paying dividends United Kingdom -.5 -.25 0.25.5 -.5 -.25 0.25.5 Years Years United States -3-2 -1 0 1 2-3 -2-1 0 1 2 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Outline 1 Data & approach 2 Heterogeneity 3 Financial frictions 4 Other transmission mechanisms 5 Concluding remarks
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 41 / 44 Our contribution: SIX NEW FINDINGS... 1 Younger firms respond more than any other group and drive the aggregate response of investment to a monetary policy shock. 2 Results are more pronounced among young firms paying no dividends and robust to controlling for other firms characteristics. 3 Younger firms capex relies more on debt (than internal funds). 4 Younger firms debt is more asset-based (than earning-based). 5 Net worth and interest payments move for all firms. 6 Borrowing move most among younger firms. 7 Sales responses are homogeneous.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 42 / 44 Our contribution: SIX NEW FINDINGS... 1 Younger firms respond more than any other group and drive the aggregate response of investment to a monetary policy shock. 2 Results are more pronounced among young firms paying no dividends and robust to controlling for other firms characteristics. 3 Younger firms capex relies more on debt (than internal funds). 4 Younger firms debt is more asset-based (than earning-based). 5 Net worth and interest payments move for all firms. 6 Borrowing move most among younger firms. 7 Sales responses are homogeneous.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 43 / 44 Our contribution: SIX NEW FINDINGS... 1 Younger firms respond more than any other group and drive the aggregate response of investment to a monetary policy shock. 2 Results are more pronounced among young firms paying no dividends and robust to controlling for other firms characteristics. 3 Younger firms capex relies more on debt (than internal funds). 4 Younger firms debt is more asset-based (than earning-based). 5 Net worth and interest payments move for all firms. 6 Borrowing move most among younger firms. 7 Sales responses are homogeneous.
Data & approach Heterogeneity Financial frictions Other transmission mechanisms Concluding remarks 44 / 44 Our contribution:...and AN INTERPRETATION Younger firms tend to use external finance (mostly debt) to fund capital expenditure, and tend to borrow against the value of the assets used as collateral. A contractionary monetary policy shock raises credit spreads, affecting most firms relying on external finance macro evidence A contractionary monetary policy shock pushes down asset prices & tighten their borrowing constraint, leading to a fall in investment. YOUNG firms face significant financial frictions & financial accelerator plays a key role in the transmission of monetary policy to investment.
Extra Slides 45 / 44
Monetary policy surprises and shocks High-frequency Surprises Policy Shocks 0.08 3 United Kingdom 0.06 0.04 0.02 0-0.02-0.04-0.06 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Date 2 1 0-1 -2-3 -4 1985 1990 1995 2000 2005 2010 2015 Date United States % -.3 -.2 -.1 0.1 1990m1 1993m1 1996m1 1999m1 2002m1 2005m1 2008m1 2011m1 2014m1 Date % -3-2 -1 0 1 2 3 1987m1 1991m1 1995m1 1999m1 2003m1 2007m1 2011m1 2015m1 Date Back 46 / 44
deviation deviation 12 24 36 48 60 deviation The response of selected macro variables Investment Industrial production United Kingdom 0.6 0.4 0.2 0-0.2-0.4-0.6-0.8-1 -1.2 0.2 0-0.2-0.4-0.6-0.8-1 -1.4 0 2 4 6 8 10 12 14 16 Months 0.4 0.2 United States 0.2 0-0.2-0.4-0.6-0.8-1 -1.2-1.4 0-0.2-0.4-0.6-0.8-1 -1.2-1.6 0 2 4 6 8 10 12 14 16-1.4 1 6 11 16 21 26 31 36 41 46 Months Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. 47 / 44
deviation deviation 12 24 36 48 60 12 24 36 48 60 The response of selected macro variables cont ed Employment Credit Spread United Kingdom 0.05 0-0.05-0.1-0.15-0.2-0.25 0.25 0.2 0.15 0.1 0.05 0-0.05 Months Months 0.1 0.4 United States 0-0.1-0.2-0.3-0.4 0.3 0.2 0.1 0-0.1-0.5 0 2 4 6 8 10 12 14 16-0.2 1 6 11 16 21 26 31 36 41 46 Months Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. Back to average effect Back to summary of results 48 / 44
49 / 44 The U.S. average effect reported at annual frequency Quarterly Annual -1 -.5 0.5-1 -.5 0.5 End year (Q4) Monetary Policy shock: 25 basis point increase. Standard errors: clustered by firms and time. Confidence band: 90%. Back to average effect
50 / 44 Investment responses by SIZE groups Smaller Medium Larger United Kingdom -1 -.5 0.5 Year end (Q4) -1 -.5 0.5 Year end (Q4) -1 -.5 0.5 Year end (Q4) United States -1.5-1 -.5 0.5-1.5-1 -.5 0.5-1.5-1 -.5 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
51 / 44 Controlling for (SMALLER) size NO dividends & Younger No dividends & Older United Kingdom -4-2 -1 0.5-4 -2-1 0.5 Years Years United States -2-1 0 1 2 3-2 -1 0 1 2 3 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. Back to robustness summary
Investment responses by ASSET GROWTH groups Faster-growing Slower-growing United Kingdom -1 -.5 0.5 Years -1 -.5 0.5 Years United States -2-1 0.5-2 -1 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. 52 / 44
53 / 44 Controlling for (FASTER) asset growth NO dividends & Younger No dividends & Older United Kingdom -4-2 -1 0.5-4 -2-1 0.5 Years Years United States -2-1 0.5-2 -1 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. Back to robustness summary
54 / 44 Investment responses by LEVERAGE groups Lower Medium Higher United Kingdom -4-2 -1 0 2 Year end (Q4) -4-2 -1 0 2 Year end (Q4) -4-2 -1 0 2 Year end (Q4) United States -1.5-1 -.5 0.5-1.5-1 -.5 0.5-1.5-1 -.5 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%.
Controlling for (LOWER) leverage NO dividends & Younger No dividends & Older United Kingdom -4-2 -1 0.5-4 -2-1 0.5 Years Years United States -2-1 0.5-2 -1 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. Back to robustness summary 55 / 44
Investment responses by TOBIN S Q groups Higher Lower United Kingdom -1 0.5 Years -1 0.5 Years United States -1.5-1 -.5 0.5-1.5-1 -.5 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. 56 / 44
57 / 44 Controlling for (HIGHER) Tobin s Q NO dividends & Younger No dividends & Older United Kingdom -4-2 -1 0.5-4 -2-1 0.5 Years Years United States -2-1 0.5-2 -1 0.5 Monetary Policy shock: 25 basis point increase. Standard errors clustering: by firms and time. Confidence band: 90%. Back to robustness summary