The Effects of Quantitative Easing on Corporate Investment, Employment, and Financing: Theory and Evidence from the Bond-Lending Channel

Transcription

1 The Effects of Quantitative Easing on Corporate Investment, Employment, and Financing: Theory and Evidence from the Bond-Lending Channel Erasmo Giambona Syracuse University Rafael Matta University of Amsterdam José-Luis Peydró Universitat Pompeu Fabra This Draft: May 20, 2017 Abstract We develop a model in which the markets for safe securities and securitized debt (e.g., mortgage-backed bonds) are complementary in stimulating corporate investment. While a decrease in the net supply of safe debt reduces the yields on corporate bonds, firms are unable to produce safe collateral and scale up investment unless the prices of securitized debt incentivizes intermediaries to invest in private capital. We show that a Quantitative Easing (QE) program based on an increase in demand for both treasury securities and securitized debt boost corporate investment by expanding firms access to bond financing while also reducing the cost of debt: the bond-lending channel. In line with our model predictions, we find that investment and employment increased, respectively, by as much as 11.8 and 3.5 percentage points for firms with access to the bond market (relative to firms that did not have access to such market) following the beginning of QE by the Federal Reserve in November We also find that leverage increased while the cost of debt financing decreased for the affected firms. In our model, firms are able to increase leverage by issuing longer-term (relatively safer) corporate bonds and notes. In line with these additional predictions, we find that senior bonds and notes (as a fraction of total debt) and debt maturity increased for treated firms. As predicted by our theory, these findings suggests that QE stimulated real activities (i.e., investment linked to job creation) through the bond-lending channel. We also find that cash holdings increased for the affected firms, which suggests that by facilitating access to cheaper debt financing, QE might have helped firms build up their cash cushions. All our findings pass the traditional robustness tests, including the parallel trend assumption or the use of alternative measures of investment and leverage. Our theory and empirical evidence help identify the corporate bond market as an important channel for the transmission of monetary policy. : E51, E52, E58, G31, G32, G38. : Quantitative Easing (QE), Bond-Lending Channel, Investment, Employment, Financing..

2 1 Introduction As the Great Recession hit the world economy in the second half of 2007, central banks around the world responded by cutting interest rates to historical lows. By December 16, 2008, the Fed Funds Rate in the U.S. had reached 0.25%, the lowest Fed Funds Rate possible (compared to 4.25% a year earlier). In this environment of effectively zero interest rates, the U.S. Federal Reserve and other central banks around the world quickly turned to Quantitative Easing (QE) to stimulate the economy. By mid-2014, the balance sheet of the Federal Reserve reached the unprecedented level of $4 trillion (including $1.6 trillion of mortgage-backed bonds and $2.4 trillion of treasuries) compared to $0.5 trillion prior to QE (Figure 1). In spite of this massive monetary policy intervention, the jury is still out on whether QE has been able to fuel the U.S. economy. In this paper, we provide theory and evidence suggesting that QE stimulated corporate investment and employment by expanding firms access to the corporate bond market, while also lowering the cost of debt financing: the bond-lending channel. Figure 1: Treasury and Mortgage Backed Securities ($ trillion) held by the Fed (Source: FRED Database) In our model the markets for safe securities and securitized debt (e.g., mortgage-backed bonds) play a complementary role in stimulating corporate investment. If the net supply of safe debt goes down, the excess demand for safety creates a downward pressure on corporate bond yields, which provides firms with an opportunity to fill the gap and reduce borrowing costs. However, firms may 1

3 be unable to respond and produce safe collateral to scale up investment unless the prices of securitized debt incentivizes intermediaries to invest in private capital. We show that a Quantitative Easing (QE) program based on an increase in demand for both treasury securities and securitized debt boost corporate investment by expanding firms access to bond financing while also reducing the cost of debt: the bond-lending channel. We test the predictions of our model by comparing investment, employment, and financing policies of firms with access to the bond market (relative to firms without access to such market) following the beginning of the QE by the Federal Reserve in November At the core of our identification strategy is the assumption that while both groups of firms are exposed to the potential consequences of QE through bank lending (because both groups borrow from banks), only firms with access to the bond market will benefit from the effects of QE on quantity and pricing in the bond market. In line with our model predictions, we find that firms with access to the bond market increased investment and employment, respectively, by as much 11.8 and 3.5 percentage points (relative to firms with no access to the bond market) during the years of the QE policy. This combined evidence suggests that QE stimulated investment linked to job creation. According to our theory, quantitative easing allows firms with access to the bond market to borrow more and at a lower rate. It does so by expanding access to the corporate bond market: the bond-lending channel. In line with these predictions, we find that leverage (market and book) increased for treated firms respectively by 2.4 and 1.3 percentage points (pp), while interest expenses as a fraction of total debt decreased by 2.3 pp. In our model, affected firms are able to increase leverage by issuing longer-term (relatively safer) corporate bonds and notes. In line with these additional predictions, we find that senior bonds and notes (as a fraction of total debt) and debt maturity increased for treated firms. We also find that cash holdings increased for the affected firms, which suggests that by facilitating access to cheaper debt financing, QE might have helped firms build up their cash cushions. All our findings pass the traditional robustness tests, including the parallel trend assumption or the use of alternative measures of investment and leverage. Our paper relates to a growing literature on the effect of government borrowing on corporate policies. Swanson (2011) and Krishnamurthy and Vissing-Jorgensen (2011, 2012) find that changes in the supply of Treasuries affect yields for corporate bonds rated A or better. Using data for public 2

4 firms over the last century, Graham, Leary, and Roberts (2014) find a strong negative correlation between government debt and corporate debt and investment. Focusing on debt maturity, Badoer and James (2016) find a negatively significant relation between the maturity of treasury securities and the maturity of corporate debt. Chakraborty, Goldstein, and MacKinlay (2017) focus on bank lending and find that banks that benefitted more from QE increased mortgage origination, while reducing commercial lending (which in turn led to lower investment by their client firms). In a related setting, Rodnyanski and Darmouni (2016) find a similar increase in mortgage origination, but insignificant changes for commercial lending. Lo Duca, Nicoletti, and Martinez (2016) focus on U.S. QE and find a strong positive relation between asset purchase activities by the Federal Reserve and corporate bond issuance. Our paper complements this literature by showing that QE boosted corporate investment and employment by increasing the availability of credit (while also lowering the cost of debt) through the bond-lending channel. To our knowledge, our study is the first to identify the corporate bond market as an important channel for the transmission of monetary policy. Our model is also related to the theoretical literature on the role of firms and intermediaries as providers of safe and liquid securities. Stein (2012) develops a model in which banks exploit the extra utility safe claims generate to households by producing some amount of riskless debt, thereby reducing financing costs. He shows that banks create an excessive amount of safe claims as they do not internalize the fire sales cost in bad states of the world. Krishnamurthy and Vissing-Jorgensen (2015) examine how the supply of treasury affects bank lending in a model where the demand for safe securities by the households generates a safety premium, which banks exploit by issuing safe and liquid claims on risky and illiquid assets. They show that an increase in supply of treasury bonds crowds out bank loans. Greenwood, Hanson, and Stein (2010) study the effect of the government debt maturity structure on the maturity of corporate bonds. They find that an increase in government funding through long-term debt increases the yield on long-term corporate bonds, leading firms to fill the gap by issuing more short-term debt. We contribute to this literature by showing that: an increase in the supply of treasury tilts firms debt structure towards riskier claims; and the resulting effect on corporate investment depends on the government intervention policy on the securitized debt market. The rest of the paper is organized as follows: The theory is in Section 2; Section 3 presents data and main empirical findings; Robustness tests are discussed in Section 4; Section 5 concludes. 3

5 2 Theory and Empirical Implications 2.1 Model Setup Out theory builds on Stein (2012) and Greenwood, Hanson, and Stein (2010). We consider a simple model that lasts for three dates t = 0, 1, 2. The economy is populated by the government and a continuum of firms, intermediaries, investors with absolute demand for safety, and preferred-habitat investors, each of which with mass one. Firms have access to projects that require investment at t = 0 and yield output at t = 2. With probability p, the performance of the projects is good and total output equals f F (K F ) > K F at t = 2 if an investment of K is made at t = 0, where f F ( ) has the following properties: strictly increasing, strictly concave, f F (K F ) 0 as K F, and f F (K F ) as K F 0. With probability 1 p, the performance of the projects is bad and total output equals K F q with q and 0 with probability 1 q. Project performance is i.i.d. across firms such that there is no aggregate uncertainty. At the beginning of t = 1 all agents receive a signal that perfectly reveals whether the performance of the project is good or bad. After the performance of the project becomes common knowledge, firms can liquidate any fraction of their assets. Liquidated assets are sold to intermediaries at the beginning of t = 1 for a price of P A, which is endogenously determined in the model. The firm finances the project with either riskless bonds or risky debt that mature at t = 2. The interest rate i on risky debt is exogenous, with expected value µ E [i] and variance σ 2 V ar [i]. Riskless bonds promise an endogenously determined rate of return of r at t = 2 for each unit invested at t = 0. The fraction φ F of the investment that firms can finance with safe bonds is bounded by the price of their assets at t = 1: φ F K F (1 + r) P A K F. Intermediaries buy assets from firms and start the production of securitized debt at beginning of t = 1. They have access to a securitization technology that, with probability θ, transforms K I units of firms assets into f I (K I ) > K I units of marketable securitized debt, where f I ( ) satisfies the same properties assumed for f F ( ). With probability 1 θ, securitization yields K I γ marketable securitized debt with probability γ and 0 units with probability 1 γ. units of Marketable securitized debt is sold at t = 2 for an endogenously determined price P M. The marketability of securitized debt is i.i.d. across intermediaries such that there is no aggregate uncertainty. At the end of t = 1 and before t = 2, all participants receive a signal that perfectly reveals 4

6 the amount of marketable securitized debt that intermediaries would be able to deliver at t = 2. At this point, intermediaries can choose to liquidate any fraction of their assets, which will be sold at t = 2 for P M. Intermediaries can finance their asset purchase with either riskless bonds or risky debt. The fraction φ I of their asset purchase financed with riskless bonds is such that φ I K I P A (1 + r) P M K I. To close the model, we specify the exogenous components of demand and supply of securities. The government offers treasury bonds at t = 0 that promise the riskless rate of return r at t = 2. The net supply of treasury securities (overall supply net of securities held by the Fed) equals S G B. Investors with absolute demand for safety inelastically demand D AS B units of riskless bonds at t = 0, so that the net exogenous demand is D B DB AS SG B. Lastly, the government exogenously demand D G M units of securitized debt at t = 2, while preferred-habitat investors inelastically demand DP H M. This yields an aggregate exogenous demand for marketable securitized debt of D M D G M + DP H M. 2.2 Equilibrium and Empirical Predictions We start with the firms maximization problem. Each firm s expected profit is given by: Π F (K F, φ F ) = pf F (K F ) + (1 p) P A K F K F (1 + µ) + φ F K F (µ r), (1) where the first three terms on the right-hand side represent the NPV of the project if it is fully financed with risky debt, and the last term is the reduction in financing costs resulting from issuing a fraction φ F of safe bonds. Each firm solves: max Π F (K F, φ F ) (2) K F,φ F subject to φ F K F (1 + r) P A K F. Since Π F (K F, φ F ) is increasing in φ F, it is easy to see that the constraint binds. This implies that firms choose φ F (P A, r) = P A 1+r. Substituting this into the objective function yields the following first-order condition for K F : f F (K F (P A, r)) = P A µ p ( 1 P ) A. 1 + r 5

7 Now we turn to the problem of intermediaries. Their expected profit is given by: Π I (K I, φ I ) = θp M f I (K I ) + (1 θ) P M K I K I P A (1 + µ) + φ I K I P A (µ r), (3) where the first three terms on the right-hand side represent the NPV of the project if it is fully financed with risky debt, and the last term is the reduction in financing costs resulting from issuing a fraction φ I of safe bonds. Each intermediary solves: max Π I (K I, φ I ) (4) K I,φ I subject to φ I K I P A (1 + r) P M K I. Since Π I (K I, φ I ) is increasing in φ I, it is easy to see that the constraint binds. This implies that firms choose φ I (P A, P M, r) = first-order condition for K I : P M P A (1+r). Substituting this into objective function yields the following f I (K I (P A, P M, r)) = µ θ The market clearing conditions imply ( PA 1 ). (5) P M 1 + r φ I (P A, P M, r) K I (P A, P M, r) + φ F (P A, r) K F (P A, r) = D B (Safe Market), (6) which allows us to derive the following proposition. (1 p) K F (P A, r) = K I (P A, P M, r) (Private Capital Market), (7) f I (K I (P A, P M, r)) = D M (Securitized Market), (8) Proposition 1 (Bond-Lending Channel). The following is true concerning government interventions in the treasury and securitized debt markets: (i) A decrease in the government net supply of treasury bonds, along with no change in the government demand for securitized debt, increases corporate bonds issuance and does not change corporate investment. (ii) An increase in the government demand for securitized debt, along with no change in the government net supply of treasury bonds, increases corporate investment and does not change corporate bonds issuance. 6

8 (ii) A decrease in the government net supply of treasury bonds, along with an increase in the government demand for securitized debt, increases both corporate bonds issuance and corporate investment. Proposition 1 sheds light on several important aspects of Quantitative Easing. The first result is that firms fill the gap left by the government in the market for safe securities following a decrease in the net supply of treasury bonds. However, the increase in the issuance of safe corporate debt in response to lower yields is not accompanied by an increase in investment. Firms simply increase the proportion of current investment funded with safe bonds. Without a rise in the prices of securitized debt, intermediaries do not increase their investments in private assets. As a result, firms are unable to produce safe assets that would allow them to scale up their investments. According to the second result, an increase in the government demand for securitized debt raises the price of long-term assets and stimulates intermediaries to invest in private capital. This increases corporate investment and raises the amount of assets available to produce riskless bonds. However, this increases safe debt yields and induces firms to increase the proportion of investments financed with risky debt. Lastly, a decrease in the net supply of treasuries, together with an increase in the demand for securitized debt, provides the highest incentive to corporate investments. Higher securitized debt prices channel intermediaries investments towards private capital. This increases the value of assets that corporations can use to produce riskless bonds, increasing their incentives to supply safety. The government s higher demand for treasuries holds down the yields on safe securities, which allow firms to fully exploit the debt capacity channel provided by the bond-lending channel. 3 Investment, Employment, and Financing of Firms with Access to the Bond Market During the QE Period: Empirical Evidence We obtain our data from several sources: (1) data on investment, employment, capital structure, firm characteristics, and ratings come from COMPUSTAT; (2) data on debt composition are from Capital IQ; (3) aggregate data on treasury and mortgage backed securities held by the Federal Reserve are from the FED Database; (4) data on Treasury bond yields are from the Department of the Treasury. We restrict our sample period to the years ; an equal split between the pre- 7

9 QE period ( ) and the during-qe period ( ). We exclude firm-year observations for which assets are less than $50 million. Finally, to avoid undue influence of outliers, we winsorize all continuous variables at the 1st and 99th percentiles of their distributions. We analyze fifteen different corporate policies: (1) Investment is the ratio of capital expenditures (COMPUSTAT s item capx) to lagged property, plant, & equipment (COMPUSTAT s item ppent); (2) Employment is the growth in the number of employees defined as the ratio between number of employees at t (COMPUSTAT s item emp) minus the number of employees at t 1 to the number of employees at t 1; (3) Cash is the ratio of cash and marketable securities (COMPUSTAT s item che) to book assets (COMPUSTAT s item at); (4) Market Leverage is the ratio of total debt (COMPUSTAT s items dlt+dlc) to market assets (COMPUSTAT s items at+prcc f csho ceq txditc); (5) Book Leverage is the ratio of total debt to book assets; (6) Interest Expenses/ is the ratio of interest expenses (COMPUSTAT s item xint) to total debt; (7) Senior Bonds & Notes/ is the ratio of senior bond and notes (from Capital IQ) to total debt; (8) Subordinated Bonds & Notes/ is the ratio of subordinated bond and notes (from Capital IQ) to total debt; (9) Bank / is the ratio of outstanding credit lines and term loans (from Capital IQ) to total debt; (10) > 1 Year is the ratio of debt maturing in more than one year (COMPUSTAT s item dlt) to total debt; (11) in 2 Years is the ratio of debt maturing in two years (COMPUSTAT s item dd2) to total debt; (12-14) in 3 5 Years are defined similarly; (15) > 5 Years is the ratio of debt maturing in more than five years (COMPUSTAT s items dlt + dlc dlc dd2 dd3 dd4 dd5) to total debt. For each of these variables, we carry out difference-in-difference estimations to study whether the particular corporate policy changed for firms with access to the bond market during the QE period relative to control firms. We do so by estimating the following model: Corporate P olicy i,t = β 1 (Bond Market Access QE P eriod) i,t + β 2 Bond Market Access i,t + Controls i,tγ + y i + z t + ɛ i,t, (9) where Corporate P olicy i,t is one of the fifteen corporate policies defined above for firm i in year t, Bond Market Access is an indicator that equals 1 if firm i has either a bond rating (COMPUS- TAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm), QE P eriod is an indicator that equals 1 if year t belongs to fiscal years , and y i and z t are respectively 8

10 firm and year fixed-effects. Our control variables includes the following company characteristics: (1) Log of Sales is the natural logarithm of sales (COMPUSTAT s item sale measured in billions of 2011 dollars using the Producer Price Index published by the U.S. Department of Labor as the deflator); (2) q is the ratio of market assets to book assets; (3) Profitability is the ratio of earnings before interest, taxes, depreciation and amortization (COMPUSTAT s item oibdp) to book assets; (4) EarningsVolatility is the ratio of the standard deviation of earnings before interest, taxes, depreciation and amortization using 4 years of consecutive observations to the average book value of total assets estimated over the same time horizon; (5) Tangibility is the ratio of property, plant, & equipment to book assets. This selection of control variables comes from some of the most cited corporate finance papers, including Titman and Wessels (1988), Rajan and Zingales (1995), Barclay and Smith (1995), Baker and Wurgler (2002), Johnson (2003), MacKay and Phillips (2005), Faulkender and Petersen (2006), Flannery and Rangan (2006), Billett, King, and Mauer (2007), Lemmon, Roberts, and Zender (2008), Byoun (2008), and Sibilkov (2009). Table 1 shows that treated firms differ from control firms in terms of corporate policies and control variables. To mitigate the concern that these differences could bias our results, we: (1) control for firm characteristics throughout all regressions; and (2) assess the robustness of our findings to the so-called parallel trend assumption. Table 1 About Here Our theory predicts that in the years following the QE policy put in place by the Federal Reserve, firms with access to the bond market invest more relative to control firms. Tables 2 reports results from difference-in-difference investment regressions based on Eq. (1). Our variable of interest is the interaction between the Bond Market Access indicator and the QE Period indicator. Across all seven estimations in Table 2, the coefficient estimate on the interaction term is positive, highly significant and economically large. Focusing on column (7), the coefficient of (statistically significant at the 1 percent level) suggests that during quantitative easing, firms with access to the bond market increased investment by 7.5 percentage points (pp) (or by 36.1% relative to the average investment of 20.9% for treated firms) relative to firms that did not have access to the bond market. Table 2 About Here 9

11 In Table 3, we re-estimate the investment model specifications for employment. Focusing on column (7), the positive coefficient on the interaction term of interest indicates that employment increased by a sizable 2.3 percentage points for treated firms during QE. The combined evidence of Tables 2 and 3 suggest that QE stimulated investment linked to job creation. Table 3 About Here Although our model does not have an explicit prediction for cash, one can expect that cheaper cost of financing might have helped firms build their cash cushions. In line with this prediction, Table 4, column (7) shows that cash holdings for treated firms increased by 2.1 percentage points in the QE period relative to control firms. Table 4 About Here What are the channels that allow firms with access to the bond market to invest more during quantitative easing? According to our theory, QE allows firms with access to the bond market to borrow more and at a lower yield. It does so by expanding access to the corporate bond market: the bond-lending channel. In line with this prediction, Figure 2 shows that while yields between corporate and treasury bonds diverged at the beginning of 2008 (the peak of the Great Recession), they started to converge as soon as the Fed started QE at the end of Notably, by mid-2009, yields on AAA corporate bonds were practically identical to yields on 10-year treasury bonds. In such environment, we should expect leverage to increase and cost of debt to decrease for treated firms. Table 6 reports evidence supporting each of these predictions. Focusing on columns (2) and (4), we find that market and book leverage increased, respectively, by 2.4 and 1.3 percentage points (or 8.9% and 3.7% relative to the average market and book leverage) for treated firms during the QE period. We also find that interest expenses decreased by 2.3 pp (or 28.6% relative to the average of 7.3%) for treated firms during QE. Figure 2 About Here Table 5 About Here Our theory also predicts that treated firms are able to borrow more by issuing (relatively safer) senior bonds and notes. Therefore, we should expect the composition of corporate debt for treated 10

12 firms to change during QE. The evidence in Table 6 supports these additional predictions. We find that the proportion of senior bonds and notes increase by as much as 8.3 pp for treated firms (column 2), while subordinated bonds and notes and bonds and bank debt decrease respectively by 1 pp and 4.3 pp (columns 4 and 6). Table 6 About Here Overall, the findings in Tables 5 and 6 support our model predictions that the channel through which firms with access to the bond market invest more is by increasing leverage through the issuance of senior bonds and notes and by reducing their cost of debt: the bond-lending channel. We also test whether corporate debt maturity changed during QE. According to the gap-filling theory of Greenwood, Hanson, and Stein (2010), if the availability of long-term treasury securities is reduced, corporate issuers respond to the shock by supplying longer-term (relatively safe) debt securities. Because through QE the Federal Reserve purchases long-term treasuries and MBS, we predict corporate issuers to fill the gap by issuing longer-term securities. Therefore, we predict an increase in debt maturity for treated firms during QE. Table 7 supports this prediction. Column 1 shows that the percentage of debt maturing in more than 1 year increased by 3.7 pp for treated firms during QE. More importantly, columns 4 and 5 show that most of the increase was driven by the longer maturity tranches (namely, debt maturing in 4 years and debt maturing in 5 years). Table 7 About Here Figure 3 provides a visual representation of our main findings in Tables Robustness Tests Figure 3 About Here In Table 2, we measure investment as the ratio of capital expenditures to lagged property, plant, & equipment. While this is one of the most used measure of investment, the literature has proposed several other measures. For robustness, we re-estimate our results in Table 2 using the three following measures: (1) the ratio of capital expenditures to contemporaneous property, plant, & equipment; (2) the ratio of capital expenditures to contemporaneous assets; and (3) the ratio of 11

13 capital expenditures to lagged assets. Table 8 reports results using these different measures of investment. Independently from the measure of investment, the coefficient on the interaction term of interest (Bond Market Access QE Period) is positive and highly statistically significant. Table 8 About Here In Table 9, we replace the QE Period indicator with dummies for each year after the start of QE (namely, 2008, 2009, 2010, and 2011). The advantage of this specification vis-a-vis our basic approach is that it allows us to link the change in corporate policies to the different stages of QE. For example, focusing on investment, column 1 shows that investment increased only by 2.8 pp in the fiscal year 2008, while the effect was a very sizable 12.3 pp in 2009, and still sizable but smaller in 2010 (8.7 pp) and 2011 (6.7 pp). This pattern is consistent with the implementation of the QE policy by the Fed. It is not surprising that the effect is relatively small in fiscal year 2008 because QE only started in November The full effect of the policy on investment was in 2009, but started to decrease afterwards perhaps because the number of positive net present value projects started to decline for treated firms. Table 9 About Here In Table 10, we replace the Bond Market Access indicator with two indicators, one for whether the firm has access to the Investment Grade Market and the other for whether the firm has access to the Speculative Market. Our theory suggests that investment grade firms should benefit the most from QE because investment grade bonds are better substitutes for treasury bonds and agency MBS. In line with this prediction, we find that investment increased by 10.1 pp for investment grade firms relative to 5.6 pp for speculative grade firm. The effect on leverage is less clear. While book leverage increased more for investment grade firms relative to speculative grade firms (1.5 pp vs. 1.1 pp), the opposite is true for market leverage. The reduction on interest expenses is slightly larger for investment grade firm, while the effect on cash is practically identical for the two groups. Overall, we also find a larger increase in senior bonds and notes for investment grade firms. Bank debt also decreased more for investment grade firms. Interestingly, while subordinated bonds and notes decreased for speculative grade firms, they actually increased slightly for investment grade firms. Finally, while debt maturing in more than 1 year remained unchanged for speculative grade 12

14 firms, it increased by a sizable 6.4 pp for investment grade firms (column 9). Notably, this increase was driven by debt maturing in more than 5 years (column 14). Table 10 About Here Although we have accounted for firm heterogeneity and secular trends in both treated and control groups, a trend in the corporate polices specific to the treated firms could bias our results. The presence of such specific trend would constitute a violation of the parallel-trend assumption, which requires that the outcome variable for the treated and control firms move in parallel before the treatment takes place. Such violation would be problematic, because the difference-in-difference estimates could be capturing a treated-specific trend rather than the QE effect on corporate polices. Table 11 assesses the robustness of our findings to the parallel trend assumption. The table re-estimates the corporate policy models of Tables 2-7, but with an additional interaction variable: Bond Market Access Trend, where Trend is a linear time trend variable (see, Angrist and Pischke, 2009). The coefficient estimate on this variable is statistically significant only in two of the fifteen regressions, but it is economically small. More important, the estimates on our key variable Bond Market Access QE Period remain about the same as those reported in Tables 2 7. Therefore, trends specific to treated firms are not the reason for the corporate policy changes during QE. 5 Concluding Remarks Table 11 About Here In this paper, we buind on the gap filling theory of Greenwood, Hanson, and Stein (2010) and Stein (2012) and propose a model in which a Quantitative Easing (QE) program based on an increase in demand for both treasury securities and securitized debt boost corporate investment by expanding firms access to bond financing while also reducing the cost of debt: the bond-lending channel. We test the predictions of the model by comparing investment, employment, and financing of firms with access to the bond market (relative to firms that do not have access to such market) in the period of the QE policy initiated by the Federal Reserve in November In line with our model s predictions, we find that investment and employment increased significantly for the affected firms during the QE period. We also find that leverage increased, while interest expenses 13

15 decreased for treated firms. In our theory, firms increase leverage by issuing longer-term (relatively safer) corporate bonds and notes. In line with these additional predictions, we find that the proportion of senior bonds and notes and debt maturity increased for treated firms. Finally, we also find that cash holdings increased for firms with access to the bond market, which suggests that by facilitating access to cheaper debt financing, QE helped firms build up their cash reserves. There is a growing literature on the importance of QE for lending. The common theme of this research is that the effects of QE propagate to the real economy through the bank-lending channel. Our paper complements existing studies by showing that QE boosted corporate investment and employment by increasing the availability of credit (while also lowering the cost of debt) through the bond-lending channel. Future research should investigate additional implications for the real economy of the corporate bond market as a channel for the transmission of monetary policy. 14

16 References [1] Angrist, J., and J. Pischke, 2009, Mostly Harmless Econometrics An Empiricist s Companion, 1st Ed., Princeton, NJ: Princeton University Press. [2] Badoer, D., and C. James, 2016, The Determinants of Long-Term Corporate Issuances, Journal of Finance 71, [3] Baker, M., and J. Wurgler, 2002, Market Timing and Capital Structure, Journal of Finance 57, [4] Barclay, M., and C. Smith, 1995, The Priority Structure of Corporate Liabilities, Journal of Finance 50, [5] Billett, M., T. King, and D. Mauer, 2007, Growth Opportunities and the Choice of Leverage, Maturity, and Covenants, Journal of Finance 62, [6] Byoun, S., 2008, How and When Do Firms Adjust their Capital Structures towards Targets? Journal of Finance 63, [7] Chakraborty, I., I. Goldstein, and A. MacKinlay, 2017, Monetary Stimulus and Bank Lending, Working paper, Wharton Research Paper. [8] Faulkender, M., and M. Petersen, 2006, Does the Source of Capital Affect Capital Structure? Review of Financial Studies 19, [9] Flannery, M., and K. Rangan, 2006, Partial Adjustment Toward Target Capital Structure, Journal of Financial Economics 79, [10] Graham, Leary, and Roberts, 2014, How Does Government Borrowing Affect Corporate Financing and Investment? Working paper. [11] Greenwood, Robin, Samuel Hanson, and Jeremy C. Stein, 2010, A Gap-Filling Theory of Corporate Maturity Choice, Journal of Finance 65, [12] Johnson, S., 2003, Maturity and the Effects of Growth Opportunities and Liquidity Risk on Leverage, Review of Financial Studies 16, [13] Krishnamurthy, Arvind, and Annette Vissing-Jorgensen, 2011, The Effects of Quantitative Easing on Interest Rates: Channels and Implications for Policy, Brookings Papers on Economic Activity Fall, [14] Krishnamurthy, Arvind, and Annette Vissing-Jorgensen, 2012, The Aggregate Demand for Treasury, Journal of Political Economy 120, [15] Krishnamurthy, Arvind, and Annette Vissing-Jorgensen, 2015, The Impact of Treasury Supply on Financial Sector Lending and Stability, Journal of Financial Economics 118,

17 [16] Lo Duca, M., G. Nicoletti, and A. Martinez, 2016, Global Corporate Bond Issuance: What Role for U.S. Quantitative Easing? Journal of International Money and Finance 60, [17] Lemmon, M., M. Roberts, and J. Zender, 2008, Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure, Journal of Finance 63, [18] MacKay, P., and G. Phillips, 2005, How Does Industry Affect Firm Financial Structure? Review of Financial Studies 18, [19] Rajan, R., and L. Zingales, 1995, What Do We Really Know About Capital Structure? Some Evidence From International Data, Journal of Finance 50, [20] Rodnyanski, A., and O. Darmouni, 2016, The Effects of Quantitative Easing on Bank Lending Behavior, Working paper, Princeton Research Paper. [21] Sibilkov, V., 2009, Asset Liquidity and Capital Structure, Journal of Financial and Quantitative Analysis 44, [22] Stein, J., 2012, Monetary Policy as Financial Stability Regulation, Quarterly Journal of Economics 127, [23] Swanson, Eric T., 2011, Let s Twist Again: A High-Frequency Event-Study Analysis of Operation Twist and its Implications for QE2, Brookings Papers on Economic Activity, spring, [24] Titman, S., and R. Wessels, 1988, The Determinants of Capital Structure Choice, Journal of Finance 43,

18 Table 1 - Descriptive Statistics: Bond Market Access (Treated Firms) and No Bond Market Access (Control Firms) This table reports sample average and standard deviation for the main variables used in the paper. The sample includes non-financial firms for the period Firm level data are from the COMPUSTAT and Capital IQ. Bond Market Access firms are those with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). Investment is the ratio of capital expenditures (COMPUSTAT s item capx) to lagged property, plant, & equipment (COMPUSTAT s item ppent). Employment is the ratio between number of employees at t (COMPUSTAT s item emp) minus the number of employees at t-1 to the number of employees at t-1. Cash is the ratio of cash and marketable securities (COMPUSTAT s item che) to book assets (COMPUSTAT s item at). Market Leverage is the ratio of total debt to market assets (COMPUSTAT s items at + prcc_f*csho ceq txditc). Book Leverage is the ratio of total debt (COMPUSTAT s items dlt + dlc) to book assets. Interest Expenses/ is the ratio of interest expenses (COMPUSTAT s item xint) to total debt. Senior Bonds & Notes/ is the ratio of senior bond and notes (from Capital IQ) to total debt. Subordinated Bonds & Notes/ is the ratio of subordinated bond and notes (from Capital IQ) to total debt. Bank / is the ratio of outstanding credit lines and term loans (from Capital IQ) to total debt. > 1 Year is the ratio of debt maturing in more than one year (COMPUSTAT s item dlt) to total debt. in 2 Years is the ratio of debt maturing in two years (COMPUSTAT s item dd2) to total debt. in 3-5 Years are defined similarly. > 5 Years is the ratio of debt maturing in more than five years (COMPUSTAT s items dlt + dlc dlc dd2 dd3 dd4 dd5) to total debt. Log of Sales is the natural logarithm of sales (COMPUSTAT s item sale measured in billions of 2011 dollars using the Producer Price Index published by the U.S. Department of Labor as the deflator). q is the ratio of market assets to book assets. Profitability is the ratio of earnings before interest, taxes, depreciation and amortization (COMPUSTAT s item oibdp) to book assets. EarningsVolatility is the ratio of the standard deviation of earnings before interest, taxes, depreciation and amortization using 4 years of consecutive observations to the average book value of total assets estimated over the same time horizon. Tangibility is the ratio of property, plant, & equipment to book assets. Standard errors are in parentheses. Treated: Bond Market Access Control: No Bond Market Access Mean St. Dev. Obs. Mean St. Dev. Obs. Difference in Mean between Samples Investment , , *** (0.007) Employment , , *** (0.003) Cash , , *** (0.003) Market Leverage , , *** (0.002) Book Leverage , , *** (0.003) Interest Expenses/ , , ** (0.001) Senior Bonds & Notes/ , , *** (0.005) Subordinated Bonds & Notes/ , , *** (0.002) Bank / , , *** (0.006) > 1 Year , , *** (0.004) in 2 Years , , (0.003) in 3 Years , , *** (0.003) in 4 Years , , (0.003) in 5 Years , , *** (0.003) > 5 Years , , *** (0.005) Log of Sales , , *** (0.022) q , , *** (0.019) Profitability , , *** (0.002) EarningsVolatility , , *** (0.002) Tangibility , , *** (0.003) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.

19 Table 2 Change in Investment for Firms with Bond Market Access in the Quantitative Easing Period This table presents estimates from investment regressions. The sample includes non-financial firms for the period All firm level data are from COMPUSTAT industrial database. Bond Market Access is an indicator for firms with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). QE Period is a dummy variable equal to one for the years , and zero for the years Refer to Table 1 for detailed variable definitions. Standard errors reported in parentheses are clustered at the firm level. Dependent variable: Investment (1) (2) (3) (4) (5) (6) (7) Bond Market Access QE Period 0.118*** 0.110*** 0.094*** 0.118*** 0.097*** 0.119*** 0.075*** (0.011) (0.011) (0.012) (0.011) (0.011) (0.011) (0.011) Bond Market Access *** *** * *** *** *** (0.022) (0.021) (0.025) (0.022) (0.020) (0.022) (0.023) Log of Sales *** *** (0.018) (0.020) q 0.049*** 0.025*** (0.008) (0.008) Profitability 0.250*** 0.263*** (0.064) (0.071) EarningsVolatility 1.208*** 1.077*** (0.138) (0.148) Tangibility (0.091) (0.092) Firm Fixed Effects Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Obs. 26,776 26,772 22,569 26,757 26,739 26,772 22,542 R-2 (within) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.

20 Table 3 Change in Employment for Firms with Bond Market Access in the Quantitative Easing Period This table presents estimates from employment regressions. The sample includes non-financial firms for the period All firm level data are from COMPUSTAT industrial database. Bond Market Access is an indicator for firms with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). QE Period is a dummy variable equal to one for the years , and zero for the years Refer to Table 1 for detailed variable definitions. Standard errors reported in parentheses are clustered at the firm level. Dependent variable: Employment (1) (2) (3) (4) (5) (6) (7) Bond Market Access QE Period 0.033*** 0.035*** 0.022*** 0.033*** 0.029*** 0.033*** 0.023*** (0.005) (0.005) (0.006) (0.005) (0.005) (0.005) (0.006) Bond Market Access (0.011) (0.011) (0.012) (0.011) (0.011) (0.011) (0.012) Log of Sales 0.024*** 0.020*** (0.006) (0.007) q 0.026*** 0.021*** (0.003) (0.003) Profitability 0.187*** 0.112*** (0.025) (0.028) EarningsVolatility 0.299*** 0.235*** (0.038) (0.038) Tangibility ** 0.038* (0.036) (0.038) Firm Fixed Effects Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Obs. 24,459 24,446 21,294 24,436 24,405 24,452 21,232 R-2 (within) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.

21 Table 4 Change in Cash Holdings for Firms with Bond Market Access in the Quantitative Easing Period This table presents estimates from cash holding regressions. The sample includes non-financial firms for the period All firm level data are from COMPUSTAT industrial database. Bond Market Access is an indicator for firms with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). QE Period is a dummy variable equal to one for the years , and zero for the years Refer to Table 1 for detailed variable definitions. Standard errors reported in parentheses are clustered at the firm level. Dependent variable: Cash (1) (2) (3) (4) (5) (6) (7) Bond Market Access QE Period 0.027*** 0.019*** 0.026*** 0.026*** 0.025*** 0.027*** 0.021*** (0.003) (0.003) (0.004) (0.003) (0.003) (0.003) (0.003) Bond Market Access *** *** *** *** *** *** *** (0.006) (0.005) (0.007) (0.005) (0.006) (0.006) (0.006) Log of Sales *** *** (0.004) (0.004) q 0.016*** 0.011*** (0.002) (0.001) Profitability 0.032** 0.087*** (0.013) (0.013) EarningsVolatility 0.087*** 0.031** (0.016) (0.013) Tangibility *** *** (0.024) (0.021) Firm Fixed Effects Yes Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Obs. 27,915 27,859 22,922 27,830 26,739 27,907 22,542 R-2 (within) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.

22 Table 5 Change in Leverage and Interest Expenses for Firms with Bond Market Access in the Quantitative Easing Period This table presents estimates from leverage (market and book) and interest expenses regressions. The sample includes non-financial firms for the period All firm level data are from COMPUSTAT industrial database. Bond Market Access is an indicator for firms with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). QE Period is a dummy variable equal to one for the years , and zero for the years Refer to Table 1 for detailed variable definitions. Standard errors reported in parentheses are clustered at the firm level. Dependent variable: Market Leverage Book Leverage Interest Expenses/ (1) (2) (3) (4) (5) (6) Bond Market Access QE Period 0.016*** 0.024*** 0.011*** 0.013*** ** ** (0.004) (0.004) (0.004) (0.005) (0.009) (0.010) Bond Market Access 0.079*** 0.066*** 0.103*** 0.095*** *** ** (0.008) (0.008) (0.009) (0.010) (0.014) (0.019) Log of Sales 0.021*** 0.019*** ** (0.004) (0.004) (0.008) q *** *** 0.020*** (0.002) (0.001) (0.006) Profitability *** *** 0.056* (0.014) (0.018) (0.034) EarningsVolatility 0.021* (0.011) (0.013) (0.052) Tangibility 0.134*** 0.138*** (0.024) (0.026) (0.042) Firm Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Obs. 22,830 22,598 27,294 22,341 22,592 18,034 R-2 (within) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.

23 Table 6 Change in Composition for Firms with Bond Market Access in the Quantitative Easing Period This table presents estimates from leverage (market and book) and interest expenses regressions. The sample includes non-financial firms for the period All firm level data are from COMPUSTAT industrial database. Bond Market Access is an indicator for firms with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). QE Period is a dummy variable equal to one for the years , and zero for the years Refer to Table 1 for detailed variable definitions. Standard errors reported in parentheses are clustered at the firm level. Dependent variable: Senior Bonds & Notes/ Subordinated Bonds & Notes/ Bank / (1) (2) (3) (4) (5) (6) Bond Market Access QE Period 0.068*** 0.083*** *** ** *** *** (0.009) (0.010) (0.004) (0.005) (0.010) (0.011) Bond Market Access 0.138*** 0.111*** 0.035*** 0.038*** * (0.018) (0.022) (0.009) (0.011) (0.018) (0.023) Log of Sales 0.018** ** (0.007) (0.003) (0.008) q *** *** (0.003) (0.001) (0.004) Profitability *** (0.030) (0.012) (0.039) EarningsVolatility * (0.041) (0.019) (0.054) Tangibility *** (0.051) (0.018) (0.052) Firm Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Obs. 25,182 19,699 25,182 19,699 25,182 19,699 R-2 (within) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.

24 Table 7 Change in Maturity for Firms with Bond Market Access in the Quantitative Easing Period This table presents estimates from debt maturity regressions. The sample includes non-financial firms for the period All firm level data are from COMPUSTAT industrial database. Bond Market Access is an indicator for firms with either a bond rating (COMPUSTAT s item splticrm) or a commercial paper rating (COMPUSTAT s item spsticrm). QE Period is a dummy variable equal to one for the years , and zero for the years Refer to Table 1 for detailed variable definitions. Standard errors reported in parentheses are clustered at the firm level. Dependent variable: > 1 Year in 2 Years in 3 Years in 4 Years in 5 Years > 5 Years (1) (2) (3) (4) (5) (6) Bond Market Access QE Period 0.037*** *** 0.029*** (0.009) (0.006) (0.006) (0.006) (0.007) (0.012) Bond Market Access 0.075*** *** *** *** *** (0.015) (0.012) (0.012) (0.013) (0.013) (0.024) Log of Sales * * ** 0.020** (0.007) (0.006) (0.006) (0.005) (0.006) (0.010) q *** ** (0.004) (0.003) (0.003) (0.003) (0.003) (0.006) Profitability 0.075** (0.037) (0.037) (0.026) (0.026) (0.027) (0.042) EarningsVolatility 0.060* ** (0.032) (0.033) (0.021) (0.021) (0.025) (0.037) Tangibility ** * (0.042) (0.037) (0.030) (0.031) (0.033) (0.060) Firm Fixed Effects Yes Yes Yes Yes Yes Yes Year Fixed Effects Yes Yes Yes Yes Yes Yes Obs. 18,759 16,357 16,342 16,455 16,192 16,246 R-2 (within) Note: ***, ** and * indicate statistical significance at the 1%, 5%, and 10% (two-tail) test levels, respectively.